Merge pull request #840 from MODSetter/dev

feat: Google Drive HITL tools, team management UI, Daytona sandboxes, indexing pipeline hardening & test infrastructure

.cursor/skills/tdd/SKILL.md

@@ -0,0 +1,112 @@
+---
+name: tdd
+description: Strict Python TDD workflow using pytest (Red-Green-Refactor).
+---
+
+---
+name: tdd
+description: Test-driven development with red-green-refactor loop. Use when user wants to build features or fix bugs using TDD, mentions "red-green-refactor", wants integration tests, or asks for test-first development.
+---
+
+# Test-Driven Development
+
+## Philosophy
+
+**Core principle**: Tests should verify behavior through public interfaces, not implementation details. Code can change entirely; tests shouldn't.
+
+**Good tests** are integration-style: they exercise real code paths through public APIs. They describe _what_ the system does, not _how_ it does it. A good test reads like a specification - "user can checkout with valid cart" tells you exactly what capability exists. These tests survive refactors because they don't care about internal structure.
+
+**Bad tests** are coupled to implementation. They mock internal collaborators, test private methods, or verify through external means (like querying a database directly instead of using the interface). The warning sign: your test breaks when you refactor, but behavior hasn't changed. If you rename an internal function and tests fail, those tests were testing implementation, not behavior.
+
+See [tests.md](tests.md) for examples and [mocking.md](mocking.md) for mocking guidelines.
+
+## Anti-Pattern: Horizontal Slices
+
+**DO NOT write all tests first, then all implementation.** This is "horizontal slicing" - treating RED as "write all tests" and GREEN as "write all code."
+
+This produces **crap tests**:
+
+- Tests written in bulk test _imagined_ behavior, not _actual_ behavior
+- You end up testing the _shape_ of things (data structures, function signatures) rather than user-facing behavior
+- Tests become insensitive to real changes - they pass when behavior breaks, fail when behavior is fine
+- You outrun your headlights, committing to test structure before understanding the implementation
+
+**Correct approach**: Vertical slices via tracer bullets. One test → one implementation → repeat. Each test responds to what you learned from the previous cycle. Because you just wrote the code, you know exactly what behavior matters and how to verify it.
+
+```
+WRONG (horizontal):
+  RED:   test1, test2, test3, test4, test5
+  GREEN: impl1, impl2, impl3, impl4, impl5
+
+RIGHT (vertical):
+  RED→GREEN: test1→impl1
+  RED→GREEN: test2→impl2
+  RED→GREEN: test3→impl3
+  ...
+```
+
+## Workflow
+
+### 1. Planning
+
+Before writing any code:
+
+- [ ] Confirm with user what interface changes are needed
+- [ ] Confirm with user which behaviors to test (prioritize)
+- [ ] Identify opportunities for [deep modules](deep-modules.md) (small interface, deep implementation)
+- [ ] Design interfaces for [testability](interface-design.md)
+- [ ] List the behaviors to test (not implementation steps)
+- [ ] Get user approval on the plan
+
+Ask: "What should the public interface look like? Which behaviors are most important to test?"
+
+**You can't test everything.** Confirm with the user exactly which behaviors matter most. Focus testing effort on critical paths and complex logic, not every possible edge case.
+
+### 2. Tracer Bullet
+
+Write ONE test that confirms ONE thing about the system:
+
+```
+RED:   Write test for first behavior → test fails
+GREEN: Write minimal code to pass → test passes
+```
+
+This is your tracer bullet - proves the path works end-to-end.
+
+### 3. Incremental Loop
+
+For each remaining behavior:
+
+```
+RED:   Write next test → fails
+GREEN: Minimal code to pass → passes
+```
+
+Rules:
+
+- One test at a time
+- Only enough code to pass current test
+- Don't anticipate future tests
+- Keep tests focused on observable behavior
+
+### 4. Refactor
+
+After all tests pass, look for [refactor candidates](refactoring.md):
+
+- [ ] Extract duplication
+- [ ] Deepen modules (move complexity behind simple interfaces)
+- [ ] Apply SOLID principles where natural
+- [ ] Consider what new code reveals about existing code
+- [ ] Run tests after each refactor step
+
+**Never refactor while RED.** Get to GREEN first.
+
+## Checklist Per Cycle
+
+```
+[ ] Test describes behavior, not implementation
+[ ] Test uses public interface only
+[ ] Test would survive internal refactor
+[ ] Code is minimal for this test
+[ ] No speculative features added
+```

.cursor/skills/tdd/deep-modules.md

@@ -0,0 +1,33 @@
+# Deep Modules
+
+From "A Philosophy of Software Design":
+
+**Deep module** = small interface + lots of implementation
+
+```
+┌─────────────────────┐
+│   Small Interface   │  ← Few methods, simple params
+├─────────────────────┤
+│                     │
+│                     │
+│  Deep Implementation│  ← Complex logic hidden
+│                     │
+│                     │
+└─────────────────────┘
+```
+
+**Shallow module** = large interface + little implementation (avoid)
+
+```
+┌─────────────────────────────────┐
+│       Large Interface           │  ← Many methods, complex params
+├─────────────────────────────────┤
+│  Thin Implementation            │  ← Just passes through
+└─────────────────────────────────┘
+```
+
+When designing interfaces, ask:
+
+- Can I reduce the number of methods?
+- Can I simplify the parameters?
+- Can I hide more complexity inside?

.cursor/skills/tdd/interface-design.md

@@ -0,0 +1,33 @@
+# Interface Design for Testability
+
+Good interfaces make testing natural:
+
+1. **Accept dependencies, don't create them**
+```python
+# Testable
+def process_order(order, payment_gateway):
+    pass
+
+# Hard to test
+def process_order(order):
+    gateway = StripeGateway()
+
+```
+
+
+2. **Return results, don't produce side effects**
+```python
+# Testable
+def calculate_discount(cart) -> float:
+    return discount
+
+# Hard to test
+def apply_discount(cart) -> None:
+    cart.total -= discount
+
+```
+
+
+3. **Small surface area**
+* Fewer methods = fewer tests needed
+* Fewer params = simpler test setup

.cursor/skills/tdd/mocking.md

@@ -0,0 +1,69 @@
+
+# When to Mock
+
+Mock at **system boundaries** only:
+
+* External APIs (payment, email, etc.)
+* Databases (sometimes - prefer test DB)
+* Time/randomness
+* File system (sometimes)
+
+Don't mock:
+
+* Your own classes/modules
+* Internal collaborators
+* Anything you control
+
+## Designing for Mockability
+
+At system boundaries, design interfaces that are easy to mock:
+
+**1. Use dependency injection**
+
+Pass external dependencies in rather than creating them internally:
+
+```python
+import os
+
+# Easy to mock
+def process_payment(order, payment_client):
+    return payment_client.charge(order.total)
+
+# Hard to mock
+def process_payment(order):
+    client = StripeClient(os.getenv("STRIPE_KEY"))
+    return client.charge(order.total)
+
+```
+
+**2. Prefer SDK-style interfaces over generic fetchers**
+
+Create specific functions for each external operation instead of one generic function with conditional logic:
+
+```python
+import requests
+
+# GOOD: Each function is independently mockable
+class UserAPI:
+    def get_user(self, user_id):
+        return requests.get(f"/users/{user_id}")
+
+    def get_orders(self, user_id):
+        return requests.get(f"/users/{user_id}/orders")
+
+    def create_order(self, data):
+        return requests.post("/orders", json=data)
+
+# BAD: Mocking requires conditional logic inside the mock
+class GenericAPI:
+    def fetch(self, endpoint, method="GET", data=None):
+        return requests.request(method, endpoint, json=data)
+
+```
+
+The SDK approach means:
+
+* Each mock returns one specific shape
+* No conditional logic in test setup
+* Easier to see which endpoints a test exercises
+* Type safety per endpoint

.cursor/skills/tdd/refactoring.md

@@ -0,0 +1,10 @@
+# Refactor Candidates
+
+After TDD cycle, look for:
+
+- **Duplication** → Extract function/class
+- **Long methods** → Break into private helpers (keep tests on public interface)
+- **Shallow modules** → Combine or deepen
+- **Feature envy** → Move logic to where data lives
+- **Primitive obsession** → Introduce value objects
+- **Existing code** the new code reveals as problematic

.cursor/skills/tdd/tests.md

@@ -0,0 +1,60 @@
+# Good and Bad Tests
+
+## Good Tests
+
+**Integration-style**: Test through real interfaces, not mocks of internal parts.
+
+```python
+# GOOD: Tests observable behavior
+def test_user_can_checkout_with_valid_cart():
+    cart = create_cart()
+    cart.add(product)
+    result = checkout(cart, payment_method)
+    assert result.status == "confirmed"
+
+```
+
+Characteristics:
+
+* Tests behavior users/callers care about
+* Uses public API only
+* Survives internal refactors
+* Describes WHAT, not HOW
+* One logical assertion per test
+
+## Bad Tests
+
+**Implementation-detail tests**: Coupled to internal structure.
+
+```python
+# BAD: Tests implementation details
+def test_checkout_calls_payment_service_process():
+    mock_payment = MagicMock()
+    checkout(cart, mock_payment)
+    mock_payment.process.assert_called_with(cart.total)
+
+```
+
+Red flags:
+
+* Mocking internal collaborators
+* Testing private methods
+* Asserting on call counts/order
+* Test breaks when refactoring without behavior change
+* Test name describes HOW not WHAT
+* Verifying through external means instead of interface
+
+```python
+# BAD: Bypasses interface to verify
+def test_create_user_saves_to_database():
+    create_user({"name": "Alice"})
+    row = db.query("SELECT * FROM users WHERE name = ?", ["Alice"])
+    assert row is not None
+
+# GOOD: Verifies through interface
+def test_create_user_makes_user_retrievable():
+    user = create_user({"name": "Alice"})
+    retrieved = get_user(user.id)
+    assert retrieved.name == "Alice"
+
+```

Dockerfile.allinone

@@ -260,6 +260,10 @@ ENV NEXT_PUBLIC_FASTAPI_BACKEND_URL=http://localhost:8000
 ENV NEXT_PUBLIC_FASTAPI_BACKEND_AUTH_TYPE=LOCAL
 ENV NEXT_PUBLIC_ETL_SERVICE=DOCLING
 
+# Daytona Sandbox (cloud code execution — no local server needed)
+ENV DAYTONA_SANDBOX_ENABLED=FALSE
+# DAYTONA_API_KEY, DAYTONA_API_URL, DAYTONA_TARGET: set at runtime for production.
+
 # Electric SQL configuration (ELECTRIC_DATABASE_URL is built dynamically by entrypoint from these values)
 ENV ELECTRIC_DB_USER=electric
 ENV ELECTRIC_DB_PASSWORD=electric_password

docker-compose.yml

@@ -65,6 +65,11 @@ services:
       - ELECTRIC_DB_PASSWORD=${ELECTRIC_DB_PASSWORD:-electric_password}
       - AUTH_TYPE=${AUTH_TYPE:-LOCAL}
       - NEXT_FRONTEND_URL=${NEXT_FRONTEND_URL:-http://localhost:3000}
+      # Daytona Sandbox – uncomment and set credentials to enable cloud code execution
+      # - DAYTONA_SANDBOX_ENABLED=TRUE
+      # - DAYTONA_API_KEY=${DAYTONA_API_KEY:-}
+      # - DAYTONA_API_URL=${DAYTONA_API_URL:-https://app.daytona.io/api}
+      # - DAYTONA_TARGET=${DAYTONA_TARGET:-us}
     depends_on:
       - db
       - redis

scripts/docker/entrypoint-allinone.sh

@@ -232,6 +232,7 @@ echo "  Auth Type:       ${NEXT_PUBLIC_FASTAPI_BACKEND_AUTH_TYPE}"
 echo "  ETL Service:     ${NEXT_PUBLIC_ETL_SERVICE}"
 echo "  TTS Service:     ${TTS_SERVICE}"
 echo "  STT Service:     ${STT_SERVICE}"
+echo "  Daytona Sandbox: ${DAYTONA_SANDBOX_ENABLED:-FALSE}"
 echo "==========================================="
 echo ""
 

surfsense_backend/.env.example

@@ -167,37 +167,12 @@ LANGSMITH_ENDPOINT=https://api.smith.langchain.com
 LANGSMITH_API_KEY=lsv2_pt_.....
 LANGSMITH_PROJECT=surfsense
 
-# Uvicorn Server Configuration
+# Agent Specific Configuration
-# Full documentation for Uvicorn options can be found at: https://www.uvicorn.org/#command-line-options
+# Daytona Sandbox (secure cloud code execution for deep agent)
-UVICORN_HOST="0.0.0.0"
+# Set DAYTONA_SANDBOX_ENABLED=TRUE to give the agent an isolated execute tool
-UVICORN_PORT=8000
+DAYTONA_SANDBOX_ENABLED=TRUE
-UVICORN_LOG_LEVEL=info
+DAYTONA_API_KEY=dtn_asdasfasfafas
-
+DAYTONA_API_URL=https://app.daytona.io/api
-# OPTIONAL: Advanced Uvicorn Options (uncomment to use)
+DAYTONA_TARGET=us
-# UVICORN_PROXY_HEADERS=false
+# Directory for locally-persisted sandbox files (after sandbox deletion)
-# UVICORN_FORWARDED_ALLOW_IPS="127.0.0.1"
+SANDBOX_FILES_DIR=sandbox_files
-# UVICORN_WORKERS=1
-# UVICORN_ACCESS_LOG=true
-# UVICORN_LOOP="auto"
-# UVICORN_HTTP="auto"
-# UVICORN_WS="auto"
-# UVICORN_LIFESPAN="auto"
-# UVICORN_LOG_CONFIG=""
-# UVICORN_SERVER_HEADER=true
-# UVICORN_DATE_HEADER=true
-# UVICORN_LIMIT_CONCURRENCY=
-# UVICORN_LIMIT_MAX_REQUESTS=
-# UVICORN_TIMEOUT_KEEP_ALIVE=5
-# UVICORN_TIMEOUT_NOTIFY=30
-# UVICORN_SSL_KEYFILE=""
-# UVICORN_SSL_CERTFILE=""
-# UVICORN_SSL_KEYFILE_PASSWORD=""
-# UVICORN_SSL_VERSION=""
-# UVICORN_SSL_CERT_REQS=""
-# UVICORN_SSL_CA_CERTS=""
-# UVICORN_SSL_CIPHERS=""
-# UVICORN_HEADERS=""
-# UVICORN_USE_COLORS=true
-# UVICORN_UDS=""
-# UVICORN_FD=""
-# UVICORN_ROOT_PATH=""

surfsense_backend/.gitignore

@@ -6,6 +6,7 @@ __pycache__/
 .flashrank_cache
 surf_new_backend.egg-info/
 podcasts/
+sandbox_files/
 temp_audio/
 celerybeat-schedule*
 celerybeat-schedule.*

surfsense_backend/alembic/versions/102_add_enable_summary_to_connectors.py

@@ -0,0 +1,46 @@
+"""102_add_enable_summary_to_connectors
+
+Revision ID: 102
+Revises: 101
+Create Date: 2026-02-26
+
+Adds enable_summary boolean column to search_source_connectors.
+Defaults to False for all existing and new connectors so LLM-based
+summary generation is opt-in.
+"""
+
+from __future__ import annotations
+
+from collections.abc import Sequence
+
+import sqlalchemy as sa
+
+from alembic import op
+
+# revision identifiers, used by Alembic.
+revision: str = "102"
+down_revision: str | None = "101"
+branch_labels: str | Sequence[str] | None = None
+depends_on: str | Sequence[str] | None = None
+
+
+def upgrade() -> None:
+    conn = op.get_bind()
+    existing_columns = [
+        col["name"] for col in sa.inspect(conn).get_columns("search_source_connectors")
+    ]
+
+    if "enable_summary" not in existing_columns:
+        op.add_column(
+            "search_source_connectors",
+            sa.Column(
+                "enable_summary",
+                sa.Boolean(),
+                nullable=False,
+                server_default=sa.text("false"),
+            ),
+        )
+
+
+def downgrade() -> None:
+    op.drop_column("search_source_connectors", "enable_summary")

surfsense_backend/app/agents/new_chat/chat_deepagent.py

@@ -6,10 +6,14 @@ with configurable tools via the tools registry and configurable prompts
 via NewLLMConfig.
 """
 
+import asyncio
+import logging
+import time
 from collections.abc import Sequence
 from typing import Any
 
 from deepagents import create_deep_agent
+from deepagents.backends.protocol import SandboxBackendProtocol
 from langchain_core.language_models import BaseChatModel
 from langchain_core.tools import BaseTool
 from langgraph.types import Checkpointer
@@ -25,6 +29,8 @@ from app.agents.new_chat.tools.registry import build_tools_async
 from app.db import ChatVisibility
 from app.services.connector_service import ConnectorService
 
+_perf_log = logging.getLogger("surfsense.perf")
+
 # =============================================================================
 # Connector Type Mapping
 # =============================================================================
@@ -128,6 +134,7 @@ async def create_surfsense_deep_agent(
     additional_tools: Sequence[BaseTool] | None = None,
     firecrawl_api_key: str | None = None,
     thread_visibility: ChatVisibility | None = None,
+    sandbox_backend: SandboxBackendProtocol | None = None,
 ):
     """
     Create a SurfSense deep agent with configurable tools and prompts.
@@ -167,6 +174,9 @@ async def create_surfsense_deep_agent(
                          These are always added regardless of enabled/disabled settings.
         firecrawl_api_key: Optional Firecrawl API key for premium web scraping.
                           Falls back to Chromium/Trafilatura if not provided.
+        sandbox_backend: Optional sandbox backend (e.g. DaytonaSandbox) for
+                        secure code execution. When provided, the agent gets an
+                        isolated ``execute`` tool for running shell commands.
 
     Returns:
         CompiledStateGraph: The configured deep agent
@@ -205,32 +215,41 @@ async def create_surfsense_deep_agent(
             additional_tools=[my_custom_tool]
         )
     """
+    _t_agent_total = time.perf_counter()
+
     # Discover available connectors and document types for this search space
-    # This enables dynamic tool docstrings that inform the LLM about what's actually available
     available_connectors: list[str] | None = None
     available_document_types: list[str] | None = None
 
+    _t0 = time.perf_counter()
     try:
-        # Get enabled search source connectors for this search space
         connector_types = await connector_service.get_available_connectors(
             search_space_id
         )
         if connector_types:
-            # Convert enum values to strings and also include mapped document types
             available_connectors = _map_connectors_to_searchable_types(connector_types)
 
-        # Get document types that have at least one document indexed
         available_document_types = await connector_service.get_available_document_types(
             search_space_id
         )
     except Exception as e:
-        # Log but don't fail - fall back to all connectors if discovery fails
-        import logging
-
         logging.warning(f"Failed to discover available connectors/document types: {e}")
+    _perf_log.info(
+        "[create_agent] Connector/doc-type discovery in %.3fs",
+        time.perf_counter() - _t0,
+    )
 
     # Build dependencies dict for the tools registry
     visibility = thread_visibility or ChatVisibility.PRIVATE
+
+    # Extract the model's context window so tools can size their output.
+    _model_profile = getattr(llm, "profile", None)
+    _max_input_tokens: int | None = (
+        _model_profile.get("max_input_tokens")
+        if isinstance(_model_profile, dict)
+        else None
+    )
+
     dependencies = {
         "search_space_id": search_space_id,
         "db_session": db_session,
@@ -241,6 +260,7 @@ async def create_surfsense_deep_agent(
         "thread_visibility": visibility,
         "available_connectors": available_connectors,
         "available_document_types": available_document_types,
+        "max_input_tokens": _max_input_tokens,
     }
 
     # Disable Notion action tools if no Notion connector is configured
@@ -269,35 +289,61 @@ async def create_surfsense_deep_agent(
         modified_disabled_tools.extend(linear_tools)
 
     # Build tools using the async registry (includes MCP tools)
+    _t0 = time.perf_counter()
     tools = await build_tools_async(
         dependencies=dependencies,
         enabled_tools=enabled_tools,
         disabled_tools=modified_disabled_tools,
         additional_tools=list(additional_tools) if additional_tools else None,
     )
+    _perf_log.info(
+        "[create_agent] build_tools_async in %.3fs (%d tools)",
+        time.perf_counter() - _t0,
+        len(tools),
+    )
 
     # Build system prompt based on agent_config
+    _t0 = time.perf_counter()
+    _sandbox_enabled = sandbox_backend is not None
     if agent_config is not None:
-        # Use configurable prompt with settings from NewLLMConfig
         system_prompt = build_configurable_system_prompt(
             custom_system_instructions=agent_config.system_instructions,
             use_default_system_instructions=agent_config.use_default_system_instructions,
             citations_enabled=agent_config.citations_enabled,
             thread_visibility=thread_visibility,
+            sandbox_enabled=_sandbox_enabled,
         )
     else:
         system_prompt = build_surfsense_system_prompt(
             thread_visibility=thread_visibility,
+            sandbox_enabled=_sandbox_enabled,
         )
+    _perf_log.info(
+        "[create_agent] System prompt built in %.3fs", time.perf_counter() - _t0
+    )
 
-    # Create the deep agent with system prompt and checkpointer
+    # Build optional kwargs for the deep agent
-    # Note: TodoListMiddleware (write_todos) is included by default in create_deep_agent
+    deep_agent_kwargs: dict[str, Any] = {}
-    agent = create_deep_agent(
+    if sandbox_backend is not None:
+        deep_agent_kwargs["backend"] = sandbox_backend
+
+    _t0 = time.perf_counter()
+    agent = await asyncio.to_thread(
+        create_deep_agent,
         model=llm,
         tools=tools,
         system_prompt=system_prompt,
         context_schema=SurfSenseContextSchema,
         checkpointer=checkpointer,
+        **deep_agent_kwargs,
+    )
+    _perf_log.info(
+        "[create_agent] Graph compiled (create_deep_agent) in %.3fs",
+        time.perf_counter() - _t0,
     )
 
+    _perf_log.info(
+        "[create_agent] Total agent creation in %.3fs",
+        time.perf_counter() - _t_agent_total,
+    )
     return agent

surfsense_backend/app/agents/new_chat/sandbox.py

@@ -0,0 +1,275 @@
+"""
+Daytona sandbox provider for SurfSense deep agent.
+
+Manages the lifecycle of sandboxed code execution environments.
+Each conversation thread gets its own isolated sandbox instance
+via the Daytona cloud API, identified by labels.
+
+Files created during a session are persisted to local storage before
+the sandbox is deleted so they remain downloadable after cleanup.
+"""
+
+from __future__ import annotations
+
+import asyncio
+import contextlib
+import logging
+import os
+import shutil
+from pathlib import Path
+
+from daytona import (
+    CreateSandboxFromSnapshotParams,
+    Daytona,
+    DaytonaConfig,
+    SandboxState,
+)
+from daytona.common.errors import DaytonaError
+from deepagents.backends.protocol import ExecuteResponse
+from langchain_daytona import DaytonaSandbox
+
+logger = logging.getLogger(__name__)
+
+
+class _TimeoutAwareSandbox(DaytonaSandbox):
+    """DaytonaSandbox subclass that accepts the per-command *timeout*
+    kwarg required by the deepagents middleware.
+
+    The upstream ``langchain-daytona`` ``execute()`` ignores timeout,
+    so deepagents raises *"This sandbox backend does not support
+    per-command timeout overrides"* on every first call.  This thin
+    wrapper forwards the parameter to the Daytona SDK.
+    """
+
+    def execute(self, command: str, *, timeout: int | None = None) -> ExecuteResponse:
+        t = timeout if timeout is not None else self._timeout
+        result = self._sandbox.process.exec(command, timeout=t)
+        return ExecuteResponse(
+            output=result.result,
+            exit_code=result.exit_code,
+            truncated=False,
+        )
+
+    async def aexecute(
+        self, command: str, *, timeout: int | None = None
+    ) -> ExecuteResponse:  # type: ignore[override]
+        return await asyncio.to_thread(self.execute, command, timeout=timeout)
+
+
+_daytona_client: Daytona | None = None
+_sandbox_cache: dict[str, _TimeoutAwareSandbox] = {}
+THREAD_LABEL_KEY = "surfsense_thread"
+
+
+def is_sandbox_enabled() -> bool:
+    return os.environ.get("DAYTONA_SANDBOX_ENABLED", "FALSE").upper() == "TRUE"
+
+
+def _get_client() -> Daytona:
+    global _daytona_client
+    if _daytona_client is None:
+        config = DaytonaConfig(
+            api_key=os.environ.get("DAYTONA_API_KEY", ""),
+            api_url=os.environ.get("DAYTONA_API_URL", "https://app.daytona.io/api"),
+            target=os.environ.get("DAYTONA_TARGET", "us"),
+        )
+        _daytona_client = Daytona(config)
+    return _daytona_client
+
+
+def _find_or_create(thread_id: str) -> _TimeoutAwareSandbox:
+    """Find an existing sandbox for *thread_id*, or create a new one.
+
+    If an existing sandbox is found but is stopped/archived, it will be
+    restarted automatically before returning.
+    """
+    client = _get_client()
+    labels = {THREAD_LABEL_KEY: thread_id}
+
+    try:
+        sandbox = client.find_one(labels=labels)
+        logger.info("Found existing sandbox %s (state=%s)", sandbox.id, sandbox.state)
+
+        if sandbox.state in (
+            SandboxState.STOPPED,
+            SandboxState.STOPPING,
+            SandboxState.ARCHIVED,
+        ):
+            logger.info("Starting stopped sandbox %s …", sandbox.id)
+            sandbox.start(timeout=60)
+            logger.info("Sandbox %s is now started", sandbox.id)
+        elif sandbox.state in (
+            SandboxState.ERROR,
+            SandboxState.BUILD_FAILED,
+            SandboxState.DESTROYED,
+        ):
+            logger.warning(
+                "Sandbox %s in unrecoverable state %s — creating a new one",
+                sandbox.id,
+                sandbox.state,
+            )
+            sandbox = client.create(
+                CreateSandboxFromSnapshotParams(language="python", labels=labels)
+            )
+            logger.info("Created replacement sandbox: %s", sandbox.id)
+        elif sandbox.state != SandboxState.STARTED:
+            sandbox.wait_for_sandbox_start(timeout=60)
+
+    except Exception:
+        logger.info("No existing sandbox for thread %s — creating one", thread_id)
+        sandbox = client.create(
+            CreateSandboxFromSnapshotParams(language="python", labels=labels)
+        )
+        logger.info("Created new sandbox: %s", sandbox.id)
+
+    return _TimeoutAwareSandbox(sandbox=sandbox)
+
+
+async def get_or_create_sandbox(thread_id: int | str) -> _TimeoutAwareSandbox:
+    """Get or create a sandbox for a conversation thread.
+
+    Uses an in-process cache keyed by thread_id so subsequent messages
+    in the same conversation reuse the sandbox object without an API call.
+
+    Args:
+        thread_id: The conversation thread identifier.
+
+    Returns:
+        DaytonaSandbox connected to the sandbox.
+    """
+    key = str(thread_id)
+    cached = _sandbox_cache.get(key)
+    if cached is not None:
+        logger.info("Reusing cached sandbox for thread %s", key)
+        return cached
+    sandbox = await asyncio.to_thread(_find_or_create, key)
+    _sandbox_cache[key] = sandbox
+    return sandbox
+
+
+async def delete_sandbox(thread_id: int | str) -> None:
+    """Delete the sandbox for a conversation thread."""
+    _sandbox_cache.pop(str(thread_id), None)
+
+    def _delete() -> None:
+        client = _get_client()
+        labels = {THREAD_LABEL_KEY: str(thread_id)}
+        try:
+            sandbox = client.find_one(labels=labels)
+        except DaytonaError:
+            logger.debug(
+                "No sandbox to delete for thread %s (already removed)", thread_id
+            )
+            return
+        try:
+            client.delete(sandbox)
+            logger.info("Sandbox deleted: %s", sandbox.id)
+        except Exception:
+            logger.warning(
+                "Failed to delete sandbox for thread %s",
+                thread_id,
+                exc_info=True,
+            )
+
+    await asyncio.to_thread(_delete)
+
+
+# ---------------------------------------------------------------------------
+# Local file persistence
+# ---------------------------------------------------------------------------
+
+
+def _get_sandbox_files_dir() -> Path:
+    return Path(os.environ.get("SANDBOX_FILES_DIR", "sandbox_files"))
+
+
+def _local_path_for(thread_id: int | str, sandbox_path: str) -> Path:
+    """Map a sandbox-internal absolute path to a local filesystem path."""
+    relative = sandbox_path.lstrip("/")
+    return _get_sandbox_files_dir() / str(thread_id) / relative
+
+
+def get_local_sandbox_file(thread_id: int | str, sandbox_path: str) -> bytes | None:
+    """Read a previously-persisted sandbox file from local storage.
+
+    Returns the file bytes, or *None* if the file does not exist locally.
+    """
+    local = _local_path_for(thread_id, sandbox_path)
+    if local.is_file():
+        return local.read_bytes()
+    return None
+
+
+def delete_local_sandbox_files(thread_id: int | str) -> None:
+    """Remove all locally-persisted sandbox files for a thread."""
+    thread_dir = _get_sandbox_files_dir() / str(thread_id)
+    if thread_dir.is_dir():
+        shutil.rmtree(thread_dir, ignore_errors=True)
+        logger.info("Deleted local sandbox files for thread %s", thread_id)
+
+
+async def persist_and_delete_sandbox(
+    thread_id: int | str,
+    sandbox_file_paths: list[str],
+) -> None:
+    """Download sandbox files to local storage, then delete the sandbox.
+
+    Each file in *sandbox_file_paths* is downloaded from the Daytona
+    sandbox and saved under ``{SANDBOX_FILES_DIR}/{thread_id}/…``.
+    Per-file errors are logged but do **not** prevent the sandbox from
+    being deleted — freeing Daytona storage is the priority.
+    """
+    _sandbox_cache.pop(str(thread_id), None)
+
+    def _persist_and_delete() -> None:
+        client = _get_client()
+        labels = {THREAD_LABEL_KEY: str(thread_id)}
+
+        try:
+            sandbox = client.find_one(labels=labels)
+        except Exception:
+            logger.info(
+                "No sandbox found for thread %s — nothing to persist", thread_id
+            )
+            return
+
+        # Ensure the sandbox is running so we can download files
+        if sandbox.state != SandboxState.STARTED:
+            try:
+                sandbox.start(timeout=60)
+            except Exception:
+                logger.warning(
+                    "Could not start sandbox %s for file download — deleting anyway",
+                    sandbox.id,
+                    exc_info=True,
+                )
+                with contextlib.suppress(Exception):
+                    client.delete(sandbox)
+                return
+
+        for path in sandbox_file_paths:
+            try:
+                content: bytes = sandbox.fs.download_file(path)
+                local = _local_path_for(thread_id, path)
+                local.parent.mkdir(parents=True, exist_ok=True)
+                local.write_bytes(content)
+                logger.info("Persisted sandbox file %s → %s", path, local)
+            except Exception:
+                logger.warning(
+                    "Failed to persist sandbox file %s for thread %s",
+                    path,
+                    thread_id,
+                    exc_info=True,
+                )
+
+        try:
+            client.delete(sandbox)
+            logger.info("Sandbox deleted after file persistence: %s", sandbox.id)
+        except Exception:
+            logger.warning(
+                "Failed to delete sandbox %s after persistence",
+                sandbox.id,
+                exc_info=True,
+            )
+
+    await asyncio.to_thread(_persist_and_delete)

surfsense_backend/app/agents/new_chat/system_prompt.py

@@ -645,6 +645,87 @@ However, from your video learning, it's important to note that asyncio is not su
 </citation_instructions>
 """
 
+# Sandbox / code execution instructions — appended when sandbox backend is enabled.
+# Inspired by Claude's computer-use prompt, scoped to code execution & data analytics.
+SANDBOX_EXECUTION_INSTRUCTIONS = """
+<code_execution>
+You have access to a secure, isolated Linux sandbox environment for running code and shell commands.
+This gives you the `execute` tool alongside the standard filesystem tools (`ls`, `read_file`, `write_file`, `edit_file`, `glob`, `grep`).
+
+## CRITICAL — CODE-FIRST RULE
+
+ALWAYS prefer executing code over giving a text-only response when the user's request involves ANY of the following:
+- **Creating a chart, plot, graph, or visualization** → Write Python code and generate the actual file. NEVER describe percentages or data in text and offer to "paste into Excel". Just produce the chart.
+- **Data analysis, statistics, or computation** → Write code to compute the answer. Do not do math by hand in text.
+- **Generating or transforming files** (CSV, PDF, images, etc.) → Write code to create the file.
+- **Running, testing, or debugging code** → Execute it in the sandbox.
+
+This applies even when you first retrieve data from the knowledge base. After `search_knowledge_base` returns relevant data, **immediately proceed to write and execute code** if the user's request matches any of the categories above. Do NOT stop at a text summary and wait for the user to ask you to "use Python" — that extra round-trip is a poor experience.
+
+Example (CORRECT):
+  User: "Create a pie chart of my benefits"
+  → 1. search_knowledge_base → retrieve benefits data
+  → 2. Immediately execute Python code (matplotlib) to generate the pie chart
+  → 3. Return the downloadable file + brief description
+
+Example (WRONG):
+  User: "Create a pie chart of my benefits"
+  → 1. search_knowledge_base → retrieve benefits data
+  → 2. Print a text table with percentages and ask the user if they want a chart ← NEVER do this
+
+## When to Use Code Execution
+
+Use the sandbox when the task benefits from actually running code rather than just describing it:
+- **Data analysis**: Load CSVs/JSON, compute statistics, filter/aggregate data, pivot tables
+- **Visualization**: Generate charts and plots (matplotlib, plotly, seaborn)
+- **Calculations**: Math, financial modeling, unit conversions, simulations
+- **Code validation**: Run and test code snippets the user provides or asks about
+- **File processing**: Parse, transform, or convert data files
+- **Quick prototyping**: Demonstrate working code for the user's problem
+- **Package exploration**: Install and test libraries the user is evaluating
+
+## When NOT to Use Code Execution
+
+Do not use the sandbox for:
+- Answering factual questions from your own knowledge
+- Summarizing or explaining concepts
+- Simple formatting or text generation tasks
+- Tasks that don't require running code to answer
+
+## Package Management
+
+- Use `pip install <package>` to install Python packages as needed
+- Common data/analytics packages (pandas, numpy, matplotlib, scipy, scikit-learn) may need to be installed on first use
+- Always verify a package installed successfully before using it
+
+## Working Guidelines
+
+- **Working directory**: The shell starts in the sandbox user's home directory (e.g. `/home/daytona`). Use **relative paths** or `/tmp/` for all files you create. NEVER write directly to `/home/` — that is the parent directory and is not writable. Use `pwd` if you need to discover the current working directory.
+- **Iterative approach**: For complex tasks, break work into steps — write code, run it, check output, refine
+- **Error handling**: If code fails, read the error, fix the issue, and retry. Don't just report the error without attempting a fix.
+- **Show results**: When generating plots or outputs, present the key findings directly in your response. For plots, save to a file and describe the results.
+- **Be efficient**: Install packages once per session. Combine related commands when possible.
+- **Large outputs**: If command output is very large, use `head`, `tail`, or save to a file and read selectively.
+
+## Sharing Generated Files
+
+When your code creates output files (images, CSVs, PDFs, etc.) in the sandbox:
+- **Print the absolute path** at the end of your script so the user can download the file. Example: `print("SANDBOX_FILE: /tmp/chart.png")`
+- **DO NOT call `display_image`** for files created inside the sandbox. Sandbox files are not accessible via public URLs, so `display_image` will always show "Image not available". The frontend automatically renders a download button from the `SANDBOX_FILE:` marker.
+- You can output multiple files, one per line: `print("SANDBOX_FILE: /tmp/report.csv")`, `print("SANDBOX_FILE: /tmp/chart.png")`
+- Always describe what the file contains in your response text so the user knows what they are downloading.
+- IMPORTANT: Every `execute` call that saves a file MUST print the `SANDBOX_FILE: <path>` marker. Without it the user cannot download the file.
+
+## Data Analytics Best Practices
+
+When the user asks you to analyze data:
+1. First, inspect the data structure (`head`, `shape`, `dtypes`, `describe()`)
+2. Clean and validate before computing (handle nulls, check types)
+3. Perform the analysis and present results clearly
+4. Offer follow-up insights or visualizations when appropriate
+</code_execution>
+"""
+
 # Anti-citation prompt - used when citations are disabled
 # This explicitly tells the model NOT to include citations
 SURFSENSE_NO_CITATION_INSTRUCTIONS = """
@@ -670,6 +751,7 @@ Your goal is to provide helpful, informative answers in a clean, readable format
 def build_surfsense_system_prompt(
     today: datetime | None = None,
     thread_visibility: ChatVisibility | None = None,
+    sandbox_enabled: bool = False,
 ) -> str:
     """
     Build the SurfSense system prompt with default settings.
@@ -678,10 +760,12 @@ def build_surfsense_system_prompt(
     - Default system instructions
     - Tools instructions (always included)
     - Citation instructions enabled
+    - Sandbox execution instructions (when sandbox_enabled=True)
 
     Args:
         today: Optional datetime for today's date (defaults to current UTC date)
         thread_visibility: Optional; when provided, used for conditional prompt (e.g. private vs shared memory wording). Defaults to private behavior when None.
+        sandbox_enabled: Whether the sandbox backend is active (adds code execution instructions).
 
     Returns:
         Complete system prompt string
@@ -691,7 +775,13 @@ def build_surfsense_system_prompt(
     system_instructions = _get_system_instructions(visibility, today)
     tools_instructions = _get_tools_instructions(visibility)
     citation_instructions = SURFSENSE_CITATION_INSTRUCTIONS
-    return system_instructions + tools_instructions + citation_instructions
+    sandbox_instructions = SANDBOX_EXECUTION_INSTRUCTIONS if sandbox_enabled else ""
+    return (
+        system_instructions
+        + tools_instructions
+        + citation_instructions
+        + sandbox_instructions
+    )
 
 
 def build_configurable_system_prompt(
@@ -700,14 +790,16 @@ def build_configurable_system_prompt(
     citations_enabled: bool = True,
     today: datetime | None = None,
     thread_visibility: ChatVisibility | None = None,
+    sandbox_enabled: bool = False,
 ) -> str:
     """
     Build a configurable SurfSense system prompt based on NewLLMConfig settings.
 
-    The prompt is composed of three parts:
+    The prompt is composed of up to four parts:
     1. System Instructions - either custom or default SURFSENSE_SYSTEM_INSTRUCTIONS
     2. Tools Instructions - always included (SURFSENSE_TOOLS_INSTRUCTIONS)
     3. Citation Instructions - either SURFSENSE_CITATION_INSTRUCTIONS or SURFSENSE_NO_CITATION_INSTRUCTIONS
+    4. Sandbox Execution Instructions - when sandbox_enabled=True
 
     Args:
         custom_system_instructions: Custom system instructions to use. If empty/None and
@@ -719,6 +811,7 @@ def build_configurable_system_prompt(
                           anti-citation instructions (False).
         today: Optional datetime for today's date (defaults to current UTC date)
         thread_visibility: Optional; when provided, used for conditional prompt (e.g. private vs shared memory wording). Defaults to private behavior when None.
+        sandbox_enabled: Whether the sandbox backend is active (adds code execution instructions).
 
     Returns:
         Complete system prompt string
@@ -727,7 +820,6 @@ def build_configurable_system_prompt(
 
     # Determine system instructions
     if custom_system_instructions and custom_system_instructions.strip():
-        # Use custom instructions, injecting the date placeholder if present
         system_instructions = custom_system_instructions.format(
             resolved_today=resolved_today
         )
@@ -735,7 +827,6 @@ def build_configurable_system_prompt(
         visibility = thread_visibility or ChatVisibility.PRIVATE
         system_instructions = _get_system_instructions(visibility, today)
     else:
-        # No system instructions (edge case)
         system_instructions = ""
 
     # Tools instructions: conditional on thread_visibility (private vs shared memory wording)
@@ -748,7 +839,14 @@ def build_configurable_system_prompt(
         else SURFSENSE_NO_CITATION_INSTRUCTIONS
     )
 
-    return system_instructions + tools_instructions + citation_instructions
+    sandbox_instructions = SANDBOX_EXECUTION_INSTRUCTIONS if sandbox_enabled else ""
+
+    return (
+        system_instructions
+        + tools_instructions
+        + citation_instructions
+        + sandbox_instructions
+    )
 
 
 def get_default_system_instructions() -> str:

surfsense_backend/app/agents/new_chat/tools/google_drive/__init__.py

@@ -0,0 +1,11 @@
+from app.agents.new_chat.tools.google_drive.create_file import (
+    create_create_google_drive_file_tool,
+)
+from app.agents.new_chat.tools.google_drive.trash_file import (
+    create_delete_google_drive_file_tool,
+)
+
+__all__ = [
+    "create_create_google_drive_file_tool",
+    "create_delete_google_drive_file_tool",
+]

surfsense_backend/app/agents/new_chat/tools/google_drive/create_file.py

@@ -0,0 +1,239 @@
+import logging
+from typing import Any, Literal
+
+from googleapiclient.errors import HttpError
+from langchain_core.tools import tool
+from langgraph.types import interrupt
+from sqlalchemy.ext.asyncio import AsyncSession
+
+from app.connectors.google_drive.client import GoogleDriveClient
+from app.connectors.google_drive.file_types import GOOGLE_DOC, GOOGLE_SHEET
+from app.services.google_drive import GoogleDriveToolMetadataService
+
+logger = logging.getLogger(__name__)
+
+_MIME_MAP: dict[str, str] = {
+    "google_doc": GOOGLE_DOC,
+    "google_sheet": GOOGLE_SHEET,
+}
+
+
+def create_create_google_drive_file_tool(
+    db_session: AsyncSession | None = None,
+    search_space_id: int | None = None,
+    user_id: str | None = None,
+):
+    @tool
+    async def create_google_drive_file(
+        name: str,
+        file_type: Literal["google_doc", "google_sheet"],
+        content: str | None = None,
+    ) -> dict[str, Any]:
+        """Create a new Google Doc or Google Sheet in Google Drive.
+
+        Use this tool when the user explicitly asks to create a new document
+        or spreadsheet in Google Drive.
+
+        Args:
+            name: The file name (without extension).
+            file_type: Either "google_doc" or "google_sheet".
+            content: Optional initial content. For google_doc, provide markdown text.
+                     For google_sheet, provide CSV-formatted text.
+
+        Returns:
+            Dictionary with:
+            - status: "success", "rejected", or "error"
+            - file_id: Google Drive file ID (if success)
+            - name: File name (if success)
+            - web_view_link: URL to open the file (if success)
+            - message: Result message
+
+            IMPORTANT:
+            - If status is "rejected", the user explicitly declined the action.
+              Respond with a brief acknowledgment and do NOT retry or suggest alternatives.
+            - If status is "insufficient_permissions", the connector lacks the required OAuth scope.
+              Inform the user they need to re-authenticate and do NOT retry the action.
+
+        Examples:
+            - "Create a Google Doc called 'Meeting Notes'"
+            - "Create a spreadsheet named 'Budget 2026' with some sample data"
+        """
+        logger.info(
+            f"create_google_drive_file called: name='{name}', type='{file_type}'"
+        )
+
+        if db_session is None or search_space_id is None or user_id is None:
+            return {
+                "status": "error",
+                "message": "Google Drive tool not properly configured. Please contact support.",
+            }
+
+        if file_type not in _MIME_MAP:
+            return {
+                "status": "error",
+                "message": f"Unsupported file type '{file_type}'. Use 'google_doc' or 'google_sheet'.",
+            }
+
+        try:
+            metadata_service = GoogleDriveToolMetadataService(db_session)
+            context = await metadata_service.get_creation_context(
+                search_space_id, user_id
+            )
+
+            if "error" in context:
+                logger.error(f"Failed to fetch creation context: {context['error']}")
+                return {"status": "error", "message": context["error"]}
+
+            logger.info(
+                f"Requesting approval for creating Google Drive file: name='{name}', type='{file_type}'"
+            )
+            approval = interrupt(
+                {
+                    "type": "google_drive_file_creation",
+                    "action": {
+                        "tool": "create_google_drive_file",
+                        "params": {
+                            "name": name,
+                            "file_type": file_type,
+                            "content": content,
+                            "connector_id": None,
+                            "parent_folder_id": None,
+                        },
+                    },
+                    "context": context,
+                }
+            )
+
+            decisions_raw = (
+                approval.get("decisions", []) if isinstance(approval, dict) else []
+            )
+            decisions = (
+                decisions_raw if isinstance(decisions_raw, list) else [decisions_raw]
+            )
+            decisions = [d for d in decisions if isinstance(d, dict)]
+            if not decisions:
+                logger.warning("No approval decision received")
+                return {"status": "error", "message": "No approval decision received"}
+
+            decision = decisions[0]
+            decision_type = decision.get("type") or decision.get("decision_type")
+            logger.info(f"User decision: {decision_type}")
+
+            if decision_type == "reject":
+                return {
+                    "status": "rejected",
+                    "message": "User declined. The file was not created. Do not ask again or suggest alternatives.",
+                }
+
+            final_params: dict[str, Any] = {}
+            edited_action = decision.get("edited_action")
+            if isinstance(edited_action, dict):
+                edited_args = edited_action.get("args")
+                if isinstance(edited_args, dict):
+                    final_params = edited_args
+            elif isinstance(decision.get("args"), dict):
+                final_params = decision["args"]
+
+            final_name = final_params.get("name", name)
+            final_file_type = final_params.get("file_type", file_type)
+            final_content = final_params.get("content", content)
+            final_connector_id = final_params.get("connector_id")
+            final_parent_folder_id = final_params.get("parent_folder_id")
+
+            if not final_name or not final_name.strip():
+                return {"status": "error", "message": "File name cannot be empty."}
+
+            mime_type = _MIME_MAP.get(final_file_type)
+            if not mime_type:
+                return {
+                    "status": "error",
+                    "message": f"Unsupported file type '{final_file_type}'.",
+                }
+
+            from sqlalchemy.future import select
+
+            from app.db import SearchSourceConnector, SearchSourceConnectorType
+
+            if final_connector_id is not None:
+                result = await db_session.execute(
+                    select(SearchSourceConnector).filter(
+                        SearchSourceConnector.id == final_connector_id,
+                        SearchSourceConnector.search_space_id == search_space_id,
+                        SearchSourceConnector.user_id == user_id,
+                        SearchSourceConnector.connector_type
+                        == SearchSourceConnectorType.GOOGLE_DRIVE_CONNECTOR,
+                    )
+                )
+                connector = result.scalars().first()
+                if not connector:
+                    return {
+                        "status": "error",
+                        "message": "Selected Google Drive connector is invalid or has been disconnected.",
+                    }
+                actual_connector_id = connector.id
+            else:
+                result = await db_session.execute(
+                    select(SearchSourceConnector).filter(
+                        SearchSourceConnector.search_space_id == search_space_id,
+                        SearchSourceConnector.user_id == user_id,
+                        SearchSourceConnector.connector_type
+                        == SearchSourceConnectorType.GOOGLE_DRIVE_CONNECTOR,
+                    )
+                )
+                connector = result.scalars().first()
+                if not connector:
+                    return {
+                        "status": "error",
+                        "message": "No Google Drive connector found. Please connect Google Drive in your workspace settings.",
+                    }
+                actual_connector_id = connector.id
+
+            logger.info(
+                f"Creating Google Drive file: name='{final_name}', type='{final_file_type}', connector={actual_connector_id}"
+            )
+            client = GoogleDriveClient(
+                session=db_session, connector_id=actual_connector_id
+            )
+            try:
+                created = await client.create_file(
+                    name=final_name,
+                    mime_type=mime_type,
+                    parent_folder_id=final_parent_folder_id,
+                    content=final_content,
+                )
+            except HttpError as http_err:
+                if http_err.resp.status == 403:
+                    logger.warning(
+                        f"Insufficient permissions for connector {actual_connector_id}: {http_err}"
+                    )
+                    return {
+                        "status": "insufficient_permissions",
+                        "connector_id": actual_connector_id,
+                        "message": "This Google Drive account needs additional permissions. Please re-authenticate.",
+                    }
+                raise
+
+            logger.info(
+                f"Google Drive file created: id={created.get('id')}, name={created.get('name')}"
+            )
+            return {
+                "status": "success",
+                "file_id": created.get("id"),
+                "name": created.get("name"),
+                "web_view_link": created.get("webViewLink"),
+                "message": f"Successfully created '{created.get('name')}' in Google Drive.",
+            }
+
+        except Exception as e:
+            from langgraph.errors import GraphInterrupt
+
+            if isinstance(e, GraphInterrupt):
+                raise
+
+            logger.error(f"Error creating Google Drive file: {e}", exc_info=True)
+            return {
+                "status": "error",
+                "message": "Something went wrong while creating the file. Please try again.",
+            }
+
+    return create_google_drive_file

surfsense_backend/app/agents/new_chat/tools/google_drive/trash_file.py

@@ -0,0 +1,243 @@
+import logging
+from typing import Any
+
+from googleapiclient.errors import HttpError
+from langchain_core.tools import tool
+from langgraph.types import interrupt
+from sqlalchemy.ext.asyncio import AsyncSession
+
+from app.connectors.google_drive.client import GoogleDriveClient
+from app.services.google_drive import GoogleDriveToolMetadataService
+
+logger = logging.getLogger(__name__)
+
+
+def create_delete_google_drive_file_tool(
+    db_session: AsyncSession | None = None,
+    search_space_id: int | None = None,
+    user_id: str | None = None,
+):
+    @tool
+    async def delete_google_drive_file(
+        file_name: str,
+        delete_from_kb: bool = False,
+    ) -> dict[str, Any]:
+        """Move a Google Drive file to trash.
+
+        Use this tool when the user explicitly asks to delete, remove, or trash
+        a file in Google Drive.
+
+        Args:
+            file_name: The exact name of the file to trash (as it appears in Drive).
+            delete_from_kb: Whether to also remove the file from the knowledge base.
+                          Default is False.
+                          Set to True to remove from both Google Drive and knowledge base.
+
+        Returns:
+            Dictionary with:
+            - status: "success", "rejected", "not_found", or "error"
+            - file_id: Google Drive file ID (if success)
+            - deleted_from_kb: whether the document was removed from the knowledge base
+            - message: Result message
+
+            IMPORTANT:
+            - If status is "rejected", the user explicitly declined. Respond with a brief
+              acknowledgment and do NOT retry or suggest alternatives.
+            - If status is "not_found", relay the exact message to the user and ask them
+              to verify the file name or check if it has been indexed.
+            - If status is "insufficient_permissions", the connector lacks the required OAuth scope.
+              Inform the user they need to re-authenticate and do NOT retry this tool.
+
+        Examples:
+            - "Delete the 'Meeting Notes' file from Google Drive"
+            - "Trash the 'Old Budget' spreadsheet"
+        """
+        logger.info(
+            f"delete_google_drive_file called: file_name='{file_name}', delete_from_kb={delete_from_kb}"
+        )
+
+        if db_session is None or search_space_id is None or user_id is None:
+            return {
+                "status": "error",
+                "message": "Google Drive tool not properly configured. Please contact support.",
+            }
+
+        try:
+            metadata_service = GoogleDriveToolMetadataService(db_session)
+            context = await metadata_service.get_trash_context(
+                search_space_id, user_id, file_name
+            )
+
+            if "error" in context:
+                error_msg = context["error"]
+                if "not found" in error_msg.lower():
+                    logger.warning(f"File not found: {error_msg}")
+                    return {"status": "not_found", "message": error_msg}
+                logger.error(f"Failed to fetch trash context: {error_msg}")
+                return {"status": "error", "message": error_msg}
+
+            file = context["file"]
+            file_id = file["file_id"]
+            document_id = file.get("document_id")
+            connector_id_from_context = context["account"]["id"]
+
+            if not file_id:
+                return {
+                    "status": "error",
+                    "message": "File ID is missing from the indexed document. Please re-index the file and try again.",
+                }
+
+            logger.info(
+                f"Requesting approval for deleting Google Drive file: '{file_name}' (file_id={file_id}, delete_from_kb={delete_from_kb})"
+            )
+            approval = interrupt(
+                {
+                    "type": "google_drive_file_trash",
+                    "action": {
+                        "tool": "delete_google_drive_file",
+                        "params": {
+                            "file_id": file_id,
+                            "connector_id": connector_id_from_context,
+                            "delete_from_kb": delete_from_kb,
+                        },
+                    },
+                    "context": context,
+                }
+            )
+
+            decisions_raw = (
+                approval.get("decisions", []) if isinstance(approval, dict) else []
+            )
+            decisions = (
+                decisions_raw if isinstance(decisions_raw, list) else [decisions_raw]
+            )
+            decisions = [d for d in decisions if isinstance(d, dict)]
+            if not decisions:
+                logger.warning("No approval decision received")
+                return {"status": "error", "message": "No approval decision received"}
+
+            decision = decisions[0]
+            decision_type = decision.get("type") or decision.get("decision_type")
+            logger.info(f"User decision: {decision_type}")
+
+            if decision_type == "reject":
+                return {
+                    "status": "rejected",
+                    "message": "User declined. The file was not trashed. Do not ask again or suggest alternatives.",
+                }
+
+            edited_action = decision.get("edited_action")
+            final_params: dict[str, Any] = {}
+            if isinstance(edited_action, dict):
+                edited_args = edited_action.get("args")
+                if isinstance(edited_args, dict):
+                    final_params = edited_args
+            elif isinstance(decision.get("args"), dict):
+                final_params = decision["args"]
+
+            final_file_id = final_params.get("file_id", file_id)
+            final_connector_id = final_params.get(
+                "connector_id", connector_id_from_context
+            )
+            final_delete_from_kb = final_params.get("delete_from_kb", delete_from_kb)
+
+            if not final_connector_id:
+                return {
+                    "status": "error",
+                    "message": "No connector found for this file.",
+                }
+
+            from sqlalchemy.future import select
+
+            from app.db import SearchSourceConnector, SearchSourceConnectorType
+
+            result = await db_session.execute(
+                select(SearchSourceConnector).filter(
+                    SearchSourceConnector.id == final_connector_id,
+                    SearchSourceConnector.search_space_id == search_space_id,
+                    SearchSourceConnector.user_id == user_id,
+                    SearchSourceConnector.connector_type
+                    == SearchSourceConnectorType.GOOGLE_DRIVE_CONNECTOR,
+                )
+            )
+            connector = result.scalars().first()
+            if not connector:
+                return {
+                    "status": "error",
+                    "message": "Selected Google Drive connector is invalid or has been disconnected.",
+                }
+
+            logger.info(
+                f"Deleting Google Drive file: file_id='{final_file_id}', connector={final_connector_id}"
+            )
+            client = GoogleDriveClient(session=db_session, connector_id=connector.id)
+            try:
+                await client.trash_file(file_id=final_file_id)
+            except HttpError as http_err:
+                if http_err.resp.status == 403:
+                    logger.warning(
+                        f"Insufficient permissions for connector {connector.id}: {http_err}"
+                    )
+                    return {
+                        "status": "insufficient_permissions",
+                        "connector_id": connector.id,
+                        "message": "This Google Drive account needs additional permissions. Please re-authenticate.",
+                    }
+                raise
+
+            logger.info(
+                f"Google Drive file deleted (moved to trash): file_id={final_file_id}"
+            )
+
+            trash_result: dict[str, Any] = {
+                "status": "success",
+                "file_id": final_file_id,
+                "message": f"Successfully moved '{file['name']}' to trash.",
+            }
+
+            deleted_from_kb = False
+            if final_delete_from_kb and document_id:
+                try:
+                    from app.db import Document
+
+                    doc_result = await db_session.execute(
+                        select(Document).filter(Document.id == document_id)
+                    )
+                    document = doc_result.scalars().first()
+                    if document:
+                        await db_session.delete(document)
+                        await db_session.commit()
+                        deleted_from_kb = True
+                        logger.info(
+                            f"Deleted document {document_id} from knowledge base"
+                        )
+                    else:
+                        logger.warning(f"Document {document_id} not found in KB")
+                except Exception as e:
+                    logger.error(f"Failed to delete document from KB: {e}")
+                    await db_session.rollback()
+                    trash_result["warning"] = (
+                        f"File moved to trash, but failed to remove from knowledge base: {e!s}"
+                    )
+
+            trash_result["deleted_from_kb"] = deleted_from_kb
+            if deleted_from_kb:
+                trash_result["message"] = (
+                    f"{trash_result.get('message', '')} (also removed from knowledge base)"
+                )
+
+            return trash_result
+
+        except Exception as e:
+            from langgraph.errors import GraphInterrupt
+
+            if isinstance(e, GraphInterrupt):
+                raise
+
+            logger.error(f"Error deleting Google Drive file: {e}", exc_info=True)
+            return {
+                "status": "error",
+                "message": "Something went wrong while trashing the file. Please try again.",
+            }
+
+    return delete_google_drive_file

surfsense_backend/app/agents/new_chat/tools/knowledge_base.py

@@ -172,12 +172,52 @@ def _normalize_connectors(
 # =============================================================================
 
 
-def format_documents_for_context(documents: list[dict[str, Any]]) -> str:
+# Fraction of the model's context window (in characters) that a single tool
+# result is allowed to occupy.  The remainder is reserved for system prompt,
+# conversation history, and model output.  With ~4 chars/token this gives a
+# tool result ≈ 25 % of the context budget in tokens.
+_TOOL_OUTPUT_CONTEXT_FRACTION = 0.25
+_CHARS_PER_TOKEN = 4
+
+# Hard-floor / ceiling so the budget is always sensible regardless of what
+# the model reports.
+_MIN_TOOL_OUTPUT_CHARS = 20_000  # ~5K tokens
+_MAX_TOOL_OUTPUT_CHARS = 400_000  # ~100K tokens
+_MAX_CHUNK_CHARS = 8_000
+
+
+def _compute_tool_output_budget(max_input_tokens: int | None) -> int:
+    """Derive a character budget from the model's context window.
+
+    Uses ``litellm.get_model_info`` via the value already resolved by
+    ``ChatLiteLLMRouter`` / ``ChatLiteLLM`` and passed through the dependency
+    chain as ``max_input_tokens``.  Falls back to a conservative default when
+    the value is unavailable.
+    """
+    if max_input_tokens is None or max_input_tokens <= 0:
+        return _MIN_TOOL_OUTPUT_CHARS  # conservative fallback
+
+    budget = int(max_input_tokens * _CHARS_PER_TOKEN * _TOOL_OUTPUT_CONTEXT_FRACTION)
+    return max(_MIN_TOOL_OUTPUT_CHARS, min(budget, _MAX_TOOL_OUTPUT_CHARS))
+
+
+def format_documents_for_context(
+    documents: list[dict[str, Any]],
+    *,
+    max_chars: int = _MAX_TOOL_OUTPUT_CHARS,
+    max_chunk_chars: int = _MAX_CHUNK_CHARS,
+) -> str:
     """
     Format retrieved documents into a readable context string for the LLM.
 
+    Documents are added in order (highest relevance first) until the character
+    budget is reached.  Individual chunks are capped at ``max_chunk_chars`` so
+    a single oversized chunk cannot monopolize the output.
+
     Args:
         documents: List of document dictionaries from connector search
+        max_chars: Approximate character budget for the entire output.
+        max_chunk_chars: Per-chunk character cap (content is tail-truncated).
 
     Returns:
         Formatted string with document contents and metadata
@@ -278,37 +318,57 @@ def format_documents_for_context(documents: list[dict[str, Any]]) -> str:
         "BAIDU_SEARCH_API",
     }
 
-    # Render XML expected by citation instructions
+    # Render XML expected by citation instructions, respecting the char budget.
     parts: list[str] = []
-    for g in grouped.values():
+    total_chars = 0
+    total_docs = len(grouped)
+
+    for doc_idx, g in enumerate(grouped.values()):
         metadata_json = json.dumps(g["metadata"], ensure_ascii=False)
         is_live_search = g["document_type"] in live_search_connectors
 
-        parts.append("<document>")
+        doc_lines: list[str] = [
-        parts.append("<document_metadata>")
+            "<document>",
-        parts.append(f"  <document_id>{g['document_id']}</document_id>")
+            "<document_metadata>",
-        parts.append(f"  <document_type>{g['document_type']}</document_type>")
+            f"  <document_id>{g['document_id']}</document_id>",
-        parts.append(f"  <title><![CDATA[{g['title']}]]></title>")
+            f"  <document_type>{g['document_type']}</document_type>",
-        parts.append(f"  <url><![CDATA[{g['url']}]]></url>")
+            f"  <title><![CDATA[{g['title']}]]></title>",
-        parts.append(f"  <metadata_json><![CDATA[{metadata_json}]]></metadata_json>")
+            f"  <url><![CDATA[{g['url']}]]></url>",
-        parts.append("</document_metadata>")
+            f"  <metadata_json><![CDATA[{metadata_json}]]></metadata_json>",
-        parts.append("")
+            "</document_metadata>",
-        parts.append("<document_content>")
+            "",
+            "<document_content>",
+        ]
 
         for ch in g["chunks"]:
             ch_content = ch["content"]
-            # For live search connectors, use the document URL as the chunk id
+            if max_chunk_chars and len(ch_content) > max_chunk_chars:
-            # so the LLM outputs [citation:https://...] which the frontend
+                ch_content = ch_content[:max_chunk_chars] + "\n...(truncated)"
-            # renders as a clickable link.
             ch_id = g["url"] if (is_live_search and g["url"]) else ch["chunk_id"]
             if ch_id is None:
-                parts.append(f"  <chunk><![CDATA[{ch_content}]]></chunk>")
+                doc_lines.append(f"  <chunk><![CDATA[{ch_content}]]></chunk>")
             else:
-                parts.append(f"  <chunk id='{ch_id}'><![CDATA[{ch_content}]]></chunk>")
+                doc_lines.append(
+                    f"  <chunk id='{ch_id}'><![CDATA[{ch_content}]]></chunk>"
+                )
 
-        parts.append("</document_content>")
+        doc_lines.extend(["</document_content>", "</document>", ""])
-        parts.append("</document>")
+
-        parts.append("")
+        doc_xml = "\n".join(doc_lines)
+        doc_len = len(doc_xml)
+
+        # Always include at least the first document; afterwards enforce budget.
+        if doc_idx > 0 and total_chars + doc_len > max_chars:
+            remaining = total_docs - doc_idx
+            parts.append(
+                f"<!-- Output truncated: {remaining} more document(s) omitted "
+                f"(budget {max_chars} chars). Refine your query or reduce top_k "
+                f"to retrieve different results. -->"
+            )
+            break
+
+        parts.append(doc_xml)
+        total_chars += doc_len
 
     return "\n".join(parts).strip()
 
@@ -328,6 +388,7 @@ async def search_knowledge_base_async(
     start_date: datetime | None = None,
     end_date: datetime | None = None,
     available_connectors: list[str] | None = None,
+    max_input_tokens: int | None = None,
 ) -> str:
     """
     Search the user's knowledge base for relevant documents.
@@ -345,6 +406,8 @@ async def search_knowledge_base_async(
         end_date: Optional end datetime (UTC) for filtering documents
         available_connectors: Optional list of connectors actually available in the search space.
                             If provided, only these connectors will be searched.
+        max_input_tokens: Model context window size (tokens).  Used to dynamically
+                         size the output so it fits within the model's limits.
 
     Returns:
         Formatted string with search results
@@ -488,7 +551,8 @@ async def search_knowledge_base_async(
 
         deduplicated.append(doc)
 
-    return format_documents_for_context(deduplicated)
+    output_budget = _compute_tool_output_budget(max_input_tokens)
+    return format_documents_for_context(deduplicated, max_chars=output_budget)
 
 
 def _build_connector_docstring(available_connectors: list[str] | None) -> str:
@@ -552,6 +616,7 @@ def create_search_knowledge_base_tool(
     connector_service: ConnectorService,
     available_connectors: list[str] | None = None,
     available_document_types: list[str] | None = None,
+    max_input_tokens: int | None = None,
 ) -> StructuredTool:
     """
     Factory function to create the search_knowledge_base tool with injected dependencies.
@@ -564,6 +629,8 @@ def create_search_knowledge_base_tool(
                             Used to dynamically generate the tool docstring.
         available_document_types: Optional list of document types that have data in the search space.
                                 Used to inform the LLM about what data exists.
+        max_input_tokens: Model context window (tokens) from litellm model info.
+                         Used to dynamically size tool output.
 
     Returns:
         A configured StructuredTool instance
@@ -634,6 +701,7 @@ NOTE: `WEBCRAWLER_CONNECTOR` is mapped internally to the canonical document type
             start_date=parsed_start,
             end_date=parsed_end,
             available_connectors=_available_connectors,
+            max_input_tokens=max_input_tokens,
         )
 
     # Create StructuredTool with dynamic description

surfsense_backend/app/agents/new_chat/tools/mcp_tool.py

@@ -11,6 +11,7 @@ This implements real MCP protocol support similar to Cursor's implementation.
 """
 
 import logging
+import time
 from typing import Any
 
 from langchain_core.tools import StructuredTool
@@ -25,6 +26,9 @@ from app.db import SearchSourceConnector, SearchSourceConnectorType
 
 logger = logging.getLogger(__name__)
 
+_MCP_CACHE_TTL_SECONDS = 300  # 5 minutes
+_mcp_tools_cache: dict[int, tuple[float, list[StructuredTool]]] = {}
+
 
 def _create_dynamic_input_model_from_schema(
     tool_name: str,
@@ -355,6 +359,19 @@ async def _load_http_mcp_tools(
     return tools
 
 
+def invalidate_mcp_tools_cache(search_space_id: int | None = None) -> None:
+    """Invalidate cached MCP tools.
+
+    Args:
+        search_space_id: If provided, only invalidate for this search space.
+                        If None, invalidate all cached MCP tools.
+    """
+    if search_space_id is not None:
+        _mcp_tools_cache.pop(search_space_id, None)
+    else:
+        _mcp_tools_cache.clear()
+
+
 async def load_mcp_tools(
     session: AsyncSession,
     search_space_id: int,
@@ -364,6 +381,9 @@ async def load_mcp_tools(
     This discovers tools dynamically from MCP servers using the protocol.
     Supports both stdio (local process) and HTTP (remote server) transports.
 
+    Results are cached per search space for up to 5 minutes to avoid
+    re-spawning MCP server processes on every chat message.
+
     Args:
         session: Database session
         search_space_id: User's search space ID
@@ -372,8 +392,20 @@ async def load_mcp_tools(
         List of LangChain StructuredTool instances
 
     """
+    now = time.monotonic()
+    cached = _mcp_tools_cache.get(search_space_id)
+    if cached is not None:
+        cached_at, cached_tools = cached
+        if now - cached_at < _MCP_CACHE_TTL_SECONDS:
+            logger.info(
+                "Using cached MCP tools for search space %s (%d tools, age=%.0fs)",
+                search_space_id,
+                len(cached_tools),
+                now - cached_at,
+            )
+            return list(cached_tools)
+
     try:
-        # Fetch all MCP connectors for this search space
         result = await session.execute(
             select(SearchSourceConnector).filter(
                 SearchSourceConnector.connector_type
@@ -385,27 +417,22 @@ async def load_mcp_tools(
         tools: list[StructuredTool] = []
         for connector in result.scalars():
             try:
-                # Early validation: Extract and validate connector config
                 config = connector.config or {}
                 server_config = config.get("server_config", {})
 
-                # Validate server_config exists and is a dict
                 if not server_config or not isinstance(server_config, dict):
                     logger.warning(
                         f"MCP connector {connector.id} (name: '{connector.name}') has invalid or missing server_config, skipping"
                     )
                     continue
 
-                # Determine transport type
                 transport = server_config.get("transport", "stdio")
 
                 if transport in ("streamable-http", "http", "sse"):
-                    # HTTP-based MCP server
                     connector_tools = await _load_http_mcp_tools(
                         connector.id, connector.name, server_config
                     )
                 else:
-                    # stdio-based MCP server (default)
                     connector_tools = await _load_stdio_mcp_tools(
                         connector.id, connector.name, server_config
                     )
@@ -417,6 +444,7 @@ async def load_mcp_tools(
                     f"Failed to load tools from MCP connector {connector.id}: {e!s}"
                 )
 
+        _mcp_tools_cache[search_space_id] = (now, tools)
         logger.info(f"Loaded {len(tools)} MCP tools for search space {search_space_id}")
         return tools
 

surfsense_backend/app/agents/new_chat/tools/registry.py

@@ -47,6 +47,10 @@ from app.db import ChatVisibility
 
 from .display_image import create_display_image_tool
 from .generate_image import create_generate_image_tool
+from .google_drive import (
+    create_create_google_drive_file_tool,
+    create_delete_google_drive_file_tool,
+)
 from .knowledge_base import create_search_knowledge_base_tool
 from .linear import (
     create_create_linear_issue_tool,
@@ -114,6 +118,7 @@ BUILTIN_TOOLS: list[ToolDefinition] = [
             # Optional: dynamically discovered connectors/document types
             available_connectors=deps.get("available_connectors"),
             available_document_types=deps.get("available_document_types"),
+            max_input_tokens=deps.get("max_input_tokens"),
         ),
         requires=["search_space_id", "db_session", "connector_service"],
         # Note: available_connectors and available_document_types are optional
@@ -292,6 +297,29 @@ BUILTIN_TOOLS: list[ToolDefinition] = [
         ),
         requires=["db_session", "search_space_id", "user_id"],
     ),
+    # =========================================================================
+    # GOOGLE DRIVE TOOLS - create files, delete files
+    # =========================================================================
+    ToolDefinition(
+        name="create_google_drive_file",
+        description="Create a new Google Doc or Google Sheet in Google Drive",
+        factory=lambda deps: create_create_google_drive_file_tool(
+            db_session=deps["db_session"],
+            search_space_id=deps["search_space_id"],
+            user_id=deps["user_id"],
+        ),
+        requires=["db_session", "search_space_id", "user_id"],
+    ),
+    ToolDefinition(
+        name="delete_google_drive_file",
+        description="Move an indexed Google Drive file to trash",
+        factory=lambda deps: create_delete_google_drive_file_tool(
+            db_session=deps["db_session"],
+            search_space_id=deps["search_space_id"],
+            user_id=deps["user_id"],
+        ),
+        requires=["db_session", "search_space_id", "user_id"],
+    ),
 ]
 
 
@@ -417,8 +445,18 @@ async def build_tools_async(
         List of configured tool instances ready for the agent, including MCP tools.
 
     """
-    # Build standard tools
+    import time
+
+    _perf_log = logging.getLogger("surfsense.perf")
+    _perf_log.setLevel(logging.DEBUG)
+
+    _t0 = time.perf_counter()
     tools = build_tools(dependencies, enabled_tools, disabled_tools, additional_tools)
+    _perf_log.info(
+        "[build_tools_async] Built-in tools in %.3fs (%d tools)",
+        time.perf_counter() - _t0,
+        len(tools),
+    )
 
     # Load MCP tools if requested and dependencies are available
     if (
@@ -427,10 +465,16 @@ async def build_tools_async(
         and "search_space_id" in dependencies
     ):
         try:
+            _t0 = time.perf_counter()
             mcp_tools = await load_mcp_tools(
                 dependencies["db_session"],
                 dependencies["search_space_id"],
             )
+            _perf_log.info(
+                "[build_tools_async] MCP tools loaded in %.3fs (%d tools)",
+                time.perf_counter() - _t0,
+                len(mcp_tools),
+            )
             tools.extend(mcp_tools)
             logging.info(
                 f"Registered {len(mcp_tools)} MCP tools: {[t.name for t in mcp_tools]}",

surfsense_backend/app/agents/new_chat/tools/search_surfsense_docs.py

@@ -14,8 +14,8 @@ from langchain_core.tools import tool
 from sqlalchemy import select
 from sqlalchemy.ext.asyncio import AsyncSession
 
-from app.config import config
 from app.db import SurfsenseDocsChunk, SurfsenseDocsDocument
+from app.utils.document_converters import embed_text
 
 
 def format_surfsense_docs_results(results: list[tuple]) -> str:
@@ -100,7 +100,7 @@ async def search_surfsense_docs_async(
         Formatted string with relevant documentation content
     """
     # Get embedding for the query
-    query_embedding = config.embedding_model_instance.embed(query)
+    query_embedding = embed_text(query)
 
     # Vector similarity search on chunks, joining with documents
     stmt = (

surfsense_backend/app/agents/new_chat/tools/shared_memory.py

@@ -8,8 +8,8 @@ from langchain_core.tools import tool
 from sqlalchemy import select
 from sqlalchemy.ext.asyncio import AsyncSession
 
-from app.config import config
 from app.db import MemoryCategory, SharedMemory, User
+from app.utils.document_converters import embed_text
 
 logger = logging.getLogger(__name__)
 
@@ -64,7 +64,7 @@ async def save_shared_memory(
         count = await get_shared_memory_count(db_session, search_space_id)
         if count >= MAX_MEMORIES_PER_SEARCH_SPACE:
             await delete_oldest_shared_memory(db_session, search_space_id)
-        embedding = config.embedding_model_instance.embed(content)
+        embedding = embed_text(content)
         row = SharedMemory(
             search_space_id=search_space_id,
             created_by_id=_to_uuid(created_by_id),
@@ -108,7 +108,7 @@ async def recall_shared_memory(
         if category and category in valid_categories:
             stmt = stmt.where(SharedMemory.category == MemoryCategory(category))
         if query:
-            query_embedding = config.embedding_model_instance.embed(query)
+            query_embedding = embed_text(query)
             stmt = stmt.order_by(
                 SharedMemory.embedding.op("<=>")(query_embedding)
             ).limit(top_k)

surfsense_backend/app/agents/new_chat/tools/user_memory.py

@@ -17,8 +17,8 @@ from langchain_core.tools import tool
 from sqlalchemy import select
 from sqlalchemy.ext.asyncio import AsyncSession
 
-from app.config import config
 from app.db import MemoryCategory, UserMemory
+from app.utils.document_converters import embed_text
 
 logger = logging.getLogger(__name__)
 
@@ -178,7 +178,7 @@ def create_save_memory_tool(
                 await delete_oldest_memory(db_session, user_id, search_space_id)
 
             # Generate embedding for the memory
-            embedding = config.embedding_model_instance.embed(content)
+            embedding = embed_text(content)
 
             # Create new memory using ORM
             # The pgvector Vector column type handles embedding conversion automatically
@@ -268,7 +268,7 @@ def create_recall_memory_tool(
 
             if query:
                 # Semantic search using embeddings
-                query_embedding = config.embedding_model_instance.embed(query)
+                query_embedding = embed_text(query)
 
                 # Build query with vector similarity
                 stmt = (

surfsense_backend/app/app.py

@@ -175,8 +175,39 @@ def rate_limit_password_reset(request: Request):
     )
 
 
+def _enable_slow_callback_logging(threshold_sec: float = 0.5) -> None:
+    """Monkey-patch the event loop to warn whenever a callback blocks longer than *threshold_sec*.
+
+    This helps pinpoint synchronous code that freezes the entire FastAPI server.
+    Only active when the PERF_DEBUG env var is set (to avoid overhead in production).
+    """
+    import os
+
+    if not os.environ.get("PERF_DEBUG"):
+        return
+
+    _slow_log = logging.getLogger("surfsense.perf.slow")
+    _slow_log.setLevel(logging.WARNING)
+    if not _slow_log.handlers:
+        _h = logging.StreamHandler()
+        _h.setFormatter(logging.Formatter("%(asctime)s [SLOW-CALLBACK] %(message)s"))
+        _slow_log.addHandler(_h)
+        _slow_log.propagate = False
+
+    loop = asyncio.get_running_loop()
+    loop.slow_callback_duration = threshold_sec  # type: ignore[attr-defined]
+    loop.set_debug(True)
+    _slow_log.warning(
+        "Event-loop slow-callback detector ENABLED (threshold=%.1fs). "
+        "Set PERF_DEBUG='' to disable.",
+        threshold_sec,
+    )
+
+
 @asynccontextmanager
 async def lifespan(app: FastAPI):
+    # Enable slow-callback detection (set PERF_DEBUG=1 env var to activate)
+    _enable_slow_callback_logging(threshold_sec=0.5)
     # Not needed if you setup a migration system like Alembic
     await create_db_and_tables()
     # Setup LangGraph checkpointer tables for conversation persistence

surfsense_backend/app/connectors/composio_gmail_connector.py

@@ -14,7 +14,6 @@ from sqlalchemy.ext.asyncio import AsyncSession
 from sqlalchemy.future import select
 from sqlalchemy.orm import selectinload
 
-from app.config import config
 from app.connectors.composio_connector import ComposioConnector
 from app.db import Document, DocumentStatus, DocumentType
 from app.services.composio_service import TOOLKIT_TO_DOCUMENT_TYPE
@@ -27,6 +26,7 @@ from app.tasks.connector_indexers.base import (
 )
 from app.utils.document_converters import (
     create_document_chunks,
+    embed_text,
     generate_content_hash,
     generate_document_summary,
     generate_unique_identifier_hash,
@@ -383,6 +383,7 @@ async def _process_gmail_messages_phase2(
     connector_id: int,
     search_space_id: int,
     user_id: str,
+    enable_summary: bool = False,
     on_heartbeat_callback: HeartbeatCallbackType | None = None,
 ) -> tuple[int, int]:
     """
@@ -415,7 +416,7 @@ async def _process_gmail_messages_phase2(
                 session, user_id, search_space_id
             )
 
-            if user_llm:
+            if user_llm and enable_summary:
                 document_metadata_for_summary = {
                     "message_id": item["message_id"],
                     "thread_id": item["thread_id"],
@@ -427,10 +428,8 @@ async def _process_gmail_messages_phase2(
                     item["markdown_content"], user_llm, document_metadata_for_summary
                 )
             else:
-                summary_content = f"Gmail: {item['subject']}\n\nFrom: {item['sender']}\nDate: {item['date_str']}"
+                summary_content = f"Gmail: {item['subject']}\n\nFrom: {item['sender']}\nDate: {item['date_str']}\n\n{item['markdown_content']}"
-                summary_embedding = config.embedding_model_instance.embed(
+                summary_embedding = embed_text(summary_content)
-                    summary_content
-                )
 
             chunks = await create_document_chunks(item["markdown_content"])
 
@@ -646,6 +645,7 @@ async def index_composio_gmail(
             connector_id=connector_id,
             search_space_id=search_space_id,
             user_id=user_id,
+            enable_summary=getattr(connector, "enable_summary", False),
             on_heartbeat_callback=on_heartbeat_callback,
         )
 

surfsense_backend/app/connectors/composio_google_calendar_connector.py

@@ -14,7 +14,6 @@ from sqlalchemy.ext.asyncio import AsyncSession
 from sqlalchemy.future import select
 from sqlalchemy.orm import selectinload
 
-from app.config import config
 from app.connectors.composio_connector import ComposioConnector
 from app.db import Document, DocumentStatus, DocumentType
 from app.services.composio_service import TOOLKIT_TO_DOCUMENT_TYPE
@@ -27,6 +26,7 @@ from app.tasks.connector_indexers.base import (
 )
 from app.utils.document_converters import (
     create_document_chunks,
+    embed_text,
     generate_content_hash,
     generate_document_summary,
     generate_unique_identifier_hash,
@@ -440,7 +440,7 @@ async def index_composio_google_calendar(
                     session, user_id, search_space_id
                 )
 
-                if user_llm:
+                if user_llm and connector.enable_summary:
                     document_metadata_for_summary = {
                         "event_id": item["event_id"],
                         "summary": item["summary"],
@@ -456,12 +456,10 @@ async def index_composio_google_calendar(
                         document_metadata_for_summary,
                     )
                 else:
-                    summary_content = f"Calendar: {item['summary']}\n\nStart: {item['start_time']}\nEnd: {item['end_time']}"
+                    summary_content = (
-                    if item["location"]:
+                        f"Calendar: {item['summary']}\n\n{item['markdown_content']}"
-                        summary_content += f"\nLocation: {item['location']}"
-                    summary_embedding = config.embedding_model_instance.embed(
-                        summary_content
                     )
+                    summary_embedding = embed_text(summary_content)
 
                 chunks = await create_document_chunks(item["markdown_content"])
 

surfsense_backend/app/connectors/composio_google_drive_connector.py

@@ -31,6 +31,7 @@ from app.tasks.connector_indexers.base import (
 )
 from app.utils.document_converters import (
     create_document_chunks,
+    embed_text,
     generate_content_hash,
     generate_document_summary,
     generate_unique_identifier_hash,
@@ -714,6 +715,7 @@ async def index_composio_google_drive(
                 max_items=max_items,
                 task_logger=task_logger,
                 log_entry=log_entry,
+                enable_summary=getattr(connector, "enable_summary", False),
                 on_heartbeat_callback=on_heartbeat_callback,
             )
         else:
@@ -747,6 +749,7 @@ async def index_composio_google_drive(
                 max_items=max_items,
                 task_logger=task_logger,
                 log_entry=log_entry,
+                enable_summary=getattr(connector, "enable_summary", False),
                 on_heartbeat_callback=on_heartbeat_callback,
             )
 
@@ -829,6 +832,7 @@ async def _index_composio_drive_delta_sync(
     max_items: int,
     task_logger: TaskLoggingService,
     log_entry,
+    enable_summary: bool = False,
     on_heartbeat_callback: HeartbeatCallbackType | None = None,
 ) -> tuple[int, int, list[str]]:
     """Index Google Drive files using delta sync with real-time document status updates.
@@ -1079,7 +1083,7 @@ async def _index_composio_drive_delta_sync(
                 session, user_id, search_space_id
             )
 
-            if user_llm:
+            if user_llm and enable_summary:
                 document_metadata_for_summary = {
                     "file_id": item["file_id"],
                     "file_name": item["file_name"],
@@ -1090,10 +1094,8 @@ async def _index_composio_drive_delta_sync(
                     markdown_content, user_llm, document_metadata_for_summary
                 )
             else:
-                summary_content = f"Google Drive File: {item['file_name']}\n\nType: {item['mime_type']}"
+                summary_content = f"Google Drive File: {item['file_name']}\n\nType: {item['mime_type']}\n\n{markdown_content}"
-                summary_embedding = config.embedding_model_instance.embed(
+                summary_embedding = embed_text(summary_content)
-                    summary_content
-                )
 
             chunks = await create_document_chunks(markdown_content)
 
@@ -1155,6 +1157,7 @@ async def _index_composio_drive_full_scan(
     max_items: int,
     task_logger: TaskLoggingService,
     log_entry,
+    enable_summary: bool = False,
     on_heartbeat_callback: HeartbeatCallbackType | None = None,
 ) -> tuple[int, int, list[str]]:
     """Index Google Drive files using full scan with real-time document status updates."""
@@ -1488,7 +1491,7 @@ async def _index_composio_drive_full_scan(
                 session, user_id, search_space_id
             )
 
-            if user_llm:
+            if user_llm and enable_summary:
                 document_metadata_for_summary = {
                     "file_id": item["file_id"],
                     "file_name": item["file_name"],
@@ -1499,10 +1502,8 @@ async def _index_composio_drive_full_scan(
                     markdown_content, user_llm, document_metadata_for_summary
                 )
             else:
-                summary_content = f"Google Drive File: {item['file_name']}\n\nType: {item['mime_type']}"
+                summary_content = f"Google Drive File: {item['file_name']}\n\nType: {item['mime_type']}\n\n{markdown_content}"
-                summary_embedding = config.embedding_model_instance.embed(
+                summary_embedding = embed_text(summary_content)
-                    summary_content
-                )
 
             chunks = await create_document_chunks(markdown_content)
 

surfsense_backend/app/connectors/google_drive/client.py

@@ -1,12 +1,15 @@
 """Google Drive API client."""
 
+import io
 from typing import Any
 
 from googleapiclient.discovery import build
 from googleapiclient.errors import HttpError
+from googleapiclient.http import MediaIoBaseUpload
 from sqlalchemy.ext.asyncio import AsyncSession
 
 from .credentials import get_valid_credentials
+from .file_types import GOOGLE_DOC, GOOGLE_SHEET
 
 
 class GoogleDriveClient:
@@ -179,3 +182,65 @@ class GoogleDriveClient:
             return None, f"HTTP error exporting file: {e.resp.status}"
         except Exception as e:
             return None, f"Error exporting file: {e!s}"
+
+    async def create_file(
+        self,
+        name: str,
+        mime_type: str,
+        parent_folder_id: str | None = None,
+        content: str | None = None,
+    ) -> dict[str, Any]:
+        service = await self.get_service()
+
+        body: dict[str, Any] = {"name": name, "mimeType": mime_type}
+        if parent_folder_id:
+            body["parents"] = [parent_folder_id]
+
+        media: MediaIoBaseUpload | None = None
+        if content:
+            if mime_type == GOOGLE_DOC:
+                import markdown as md_lib
+
+                html = md_lib.markdown(content)
+                media = MediaIoBaseUpload(
+                    io.BytesIO(html.encode("utf-8")),
+                    mimetype="text/html",
+                    resumable=False,
+                )
+            elif mime_type == GOOGLE_SHEET:
+                media = MediaIoBaseUpload(
+                    io.BytesIO(content.encode("utf-8")),
+                    mimetype="text/csv",
+                    resumable=False,
+                )
+
+        if media:
+            return (
+                service.files()
+                .create(
+                    body=body,
+                    media_body=media,
+                    fields="id,name,mimeType,webViewLink",
+                    supportsAllDrives=True,
+                )
+                .execute()
+            )
+
+        return (
+            service.files()
+            .create(
+                body=body,
+                fields="id,name,mimeType,webViewLink",
+                supportsAllDrives=True,
+            )
+            .execute()
+        )
+
+    async def trash_file(self, file_id: str) -> bool:
+        service = await self.get_service()
+        service.files().update(
+            fileId=file_id,
+            body={"trashed": True},
+            supportsAllDrives=True,
+        ).execute()
+        return True

surfsense_backend/app/db.py

@@ -1,6 +1,6 @@
 from collections.abc import AsyncGenerator
 from datetime import UTC, datetime
-from enum import Enum
+from enum import StrEnum
 
 from fastapi import Depends
 from fastapi_users.db import SQLAlchemyBaseUserTableUUID, SQLAlchemyUserDatabase
@@ -31,7 +31,7 @@ if config.AUTH_TYPE == "GOOGLE":
 DATABASE_URL = config.DATABASE_URL
 
 
-class DocumentType(str, Enum):
+class DocumentType(StrEnum):
     EXTENSION = "EXTENSION"
     CRAWLED_URL = "CRAWLED_URL"
     FILE = "FILE"
@@ -60,7 +60,7 @@ class DocumentType(str, Enum):
     COMPOSIO_GOOGLE_CALENDAR_CONNECTOR = "COMPOSIO_GOOGLE_CALENDAR_CONNECTOR"
 
 
-class SearchSourceConnectorType(str, Enum):
+class SearchSourceConnectorType(StrEnum):
     SERPER_API = "SERPER_API"  # NOT IMPLEMENTED YET : DON'T REMEMBER WHY : MOST PROBABLY BECAUSE WE NEED TO CRAWL THE RESULTS RETURNED BY IT
     TAVILY_API = "TAVILY_API"
     SEARXNG_API = "SEARXNG_API"
@@ -93,7 +93,7 @@ class SearchSourceConnectorType(str, Enum):
     COMPOSIO_GOOGLE_CALENDAR_CONNECTOR = "COMPOSIO_GOOGLE_CALENDAR_CONNECTOR"
 
 
-class PodcastStatus(str, Enum):
+class PodcastStatus(StrEnum):
     PENDING = "pending"
     GENERATING = "generating"
     READY = "ready"
@@ -177,7 +177,7 @@ class DocumentStatus:
         return None
 
 
-class LiteLLMProvider(str, Enum):
+class LiteLLMProvider(StrEnum):
     """
     Enum for LLM providers supported by LiteLLM.
     """
@@ -215,7 +215,7 @@ class LiteLLMProvider(str, Enum):
     CUSTOM = "CUSTOM"
 
 
-class ImageGenProvider(str, Enum):
+class ImageGenProvider(StrEnum):
     """
     Enum for image generation providers supported by LiteLLM.
     This is a subset of LLM providers — only those that support image generation.
@@ -233,7 +233,7 @@ class ImageGenProvider(str, Enum):
     NSCALE = "NSCALE"
 
 
-class LogLevel(str, Enum):
+class LogLevel(StrEnum):
     DEBUG = "DEBUG"
     INFO = "INFO"
     WARNING = "WARNING"
@@ -241,13 +241,13 @@ class LogLevel(str, Enum):
     CRITICAL = "CRITICAL"
 
 
-class LogStatus(str, Enum):
+class LogStatus(StrEnum):
     IN_PROGRESS = "IN_PROGRESS"
     SUCCESS = "SUCCESS"
     FAILED = "FAILED"
 
 
-class IncentiveTaskType(str, Enum):
+class IncentiveTaskType(StrEnum):
     """
     Enum for incentive task types that users can complete to earn free pages.
     Each task can only be completed once per user.
@@ -298,7 +298,7 @@ INCENTIVE_TASKS_CONFIG = {
 }
 
 
-class Permission(str, Enum):
+class Permission(StrEnum):
     """
     Granular permissions for search space resources.
     Use '*' (FULL_ACCESS) to grant all permissions.
@@ -471,7 +471,7 @@ class BaseModel(Base):
     id = Column(Integer, primary_key=True, index=True)
 
 
-class NewChatMessageRole(str, Enum):
+class NewChatMessageRole(StrEnum):
     """Role enum for new chat messages."""
 
     USER = "user"
@@ -479,7 +479,7 @@ class NewChatMessageRole(str, Enum):
     SYSTEM = "system"
 
 
-class ChatVisibility(str, Enum):
+class ChatVisibility(StrEnum):
     """
     Visibility/sharing level for chat threads.
 
@@ -788,7 +788,7 @@ class ChatSessionState(BaseModel):
     ai_responding_to_user = relationship("User")
 
 
-class MemoryCategory(str, Enum):
+class MemoryCategory(StrEnum):
     """Categories for user memories."""
 
     # Using lowercase keys to match PostgreSQL enum values
@@ -1317,6 +1317,12 @@ class SearchSourceConnector(BaseModel, TimestampMixin):
     last_indexed_at = Column(TIMESTAMP(timezone=True), nullable=True)
     config = Column(JSON, nullable=False)
 
+    # Summary generation (LLM-based) - disabled by default to save resources.
+    # When enabled, improves hybrid search quality at the cost of LLM calls.
+    enable_summary = Column(
+        Boolean, nullable=False, default=False, server_default="false"
+    )
+
     # Periodic indexing fields
     periodic_indexing_enabled = Column(Boolean, nullable=False, default=False)
     indexing_frequency_minutes = Column(Integer, nullable=True)

surfsense_backend/app/indexing_pipeline/adapters/file_upload_adapter.py

@@ -0,0 +1,47 @@
+from sqlalchemy.ext.asyncio import AsyncSession
+
+from app.db import DocumentStatus, DocumentType
+from app.indexing_pipeline.connector_document import ConnectorDocument
+from app.indexing_pipeline.indexing_pipeline_service import IndexingPipelineService
+
+
+async def index_uploaded_file(
+    markdown_content: str,
+    filename: str,
+    etl_service: str,
+    search_space_id: int,
+    user_id: str,
+    session: AsyncSession,
+    llm,
+    should_summarize: bool = False,
+) -> None:
+    connector_doc = ConnectorDocument(
+        title=filename,
+        source_markdown=markdown_content,
+        unique_id=filename,
+        document_type=DocumentType.FILE,
+        search_space_id=search_space_id,
+        created_by_id=user_id,
+        connector_id=None,
+        should_summarize=should_summarize,
+        should_use_code_chunker=False,
+        fallback_summary=markdown_content[:4000],
+        metadata={
+            "FILE_NAME": filename,
+            "ETL_SERVICE": etl_service,
+        },
+    )
+
+    service = IndexingPipelineService(session)
+    documents = await service.prepare_for_indexing([connector_doc])
+
+    if not documents:
+        raise RuntimeError("prepare_for_indexing returned no documents")
+
+    indexed = await service.index(documents[0], connector_doc, llm)
+
+    if not DocumentStatus.is_state(indexed.status, DocumentStatus.READY):
+        raise RuntimeError(indexed.status.get("reason", "Indexing failed"))
+
+    indexed.content_needs_reindexing = False
+    await session.commit()

surfsense_backend/app/indexing_pipeline/connector_document.py

@@ -0,0 +1,26 @@
+from pydantic import BaseModel, Field, field_validator
+
+from app.db import DocumentType
+
+
+class ConnectorDocument(BaseModel):
+    """Canonical data transfer object produced by connector adapters and consumed by the indexing pipeline."""
+
+    title: str
+    source_markdown: str
+    unique_id: str
+    document_type: DocumentType
+    search_space_id: int = Field(gt=0)
+    should_summarize: bool = True
+    should_use_code_chunker: bool = False
+    fallback_summary: str | None = None
+    metadata: dict = {}
+    connector_id: int | None = None
+    created_by_id: str
+
+    @field_validator("title", "source_markdown", "unique_id", "created_by_id")
+    @classmethod
+    def not_empty(cls, v: str, info) -> str:
+        if not v.strip():
+            raise ValueError(f"{info.field_name} must not be empty or whitespace")
+        return v

surfsense_backend/app/indexing_pipeline/document_chunker.py

@@ -0,0 +1,9 @@
+from app.config import config
+
+
+def chunk_text(text: str, use_code_chunker: bool = False) -> list[str]:
+    """Chunk a text string using the configured chunker and return the chunk texts."""
+    chunker = (
+        config.code_chunker_instance if use_code_chunker else config.chunker_instance
+    )
+    return [c.text for c in chunker.chunk(text)]

surfsense_backend/app/indexing_pipeline/document_embedder.py

@@ -0,0 +1,3 @@
+from app.utils.document_converters import embed_text
+
+__all__ = ["embed_text"]

surfsense_backend/app/indexing_pipeline/document_hashing.py

@@ -0,0 +1,15 @@
+import hashlib
+
+from app.indexing_pipeline.connector_document import ConnectorDocument
+
+
+def compute_unique_identifier_hash(doc: ConnectorDocument) -> str:
+    """Return a stable SHA-256 hash identifying a document by its source identity."""
+    combined = f"{doc.document_type.value}:{doc.unique_id}:{doc.search_space_id}"
+    return hashlib.sha256(combined.encode("utf-8")).hexdigest()
+
+
+def compute_content_hash(doc: ConnectorDocument) -> str:
+    """Return a SHA-256 hash of the document's content scoped to its search space."""
+    combined = f"{doc.search_space_id}:{doc.source_markdown}"
+    return hashlib.sha256(combined.encode("utf-8")).hexdigest()

surfsense_backend/app/indexing_pipeline/document_persistence.py

@@ -0,0 +1,39 @@
+from datetime import UTC, datetime
+
+from sqlalchemy.ext.asyncio import AsyncSession
+from sqlalchemy.orm import object_session
+from sqlalchemy.orm.attributes import set_committed_value
+
+from app.db import Document, DocumentStatus
+
+
+async def rollback_and_persist_failure(
+    session: AsyncSession, document: Document, message: str
+) -> None:
+    """Roll back the current transaction and best-effort persist a failed status.
+
+    Called exclusively from except blocks — must never raise, or the new exception
+    would chain with the original and mask it entirely.
+    """
+    try:
+        await session.rollback()
+    except Exception:
+        return  # Session is completely dead; nothing further we can do.
+    try:
+        await session.refresh(document)
+        document.updated_at = datetime.now(UTC)
+        document.status = DocumentStatus.failed(message)
+        await session.commit()
+    except Exception:
+        pass  # Best-effort; document will be retried on the next sync.
+
+
+def attach_chunks_to_document(document: Document, chunks: list) -> None:
+    """Assign chunks to a document without triggering SQLAlchemy async lazy loading."""
+    set_committed_value(document, "chunks", chunks)
+    session = object_session(document)
+    if session is not None:
+        if document.id is not None:
+            for chunk in chunks:
+                chunk.document_id = document.id
+        session.add_all(chunks)

surfsense_backend/app/indexing_pipeline/document_summarizer.py

@@ -0,0 +1,30 @@
+from app.prompts import SUMMARY_PROMPT_TEMPLATE
+from app.utils.document_converters import optimize_content_for_context_window
+
+
+async def summarize_document(
+    source_markdown: str, llm, metadata: dict | None = None
+) -> str:
+    """Generate a text summary of a document using an LLM, prefixed with metadata when provided."""
+    model_name = getattr(llm, "model", "gpt-3.5-turbo")
+    optimized_content = optimize_content_for_context_window(
+        source_markdown, metadata, model_name
+    )
+
+    summary_chain = SUMMARY_PROMPT_TEMPLATE | llm
+    content_with_metadata = (
+        f"<DOCUMENT><DOCUMENT_METADATA>\n\n{metadata}\n\n</DOCUMENT_METADATA>"
+        f"\n\n<DOCUMENT_CONTENT>\n\n{optimized_content}\n\n</DOCUMENT_CONTENT></DOCUMENT>"
+    )
+    summary_result = await summary_chain.ainvoke({"document": content_with_metadata})
+    summary_content = summary_result.content
+
+    if metadata:
+        metadata_parts = ["# DOCUMENT METADATA"]
+        for key, value in metadata.items():
+            if value:
+                metadata_parts.append(f"**{key.replace('_', ' ').title()}:** {value}")
+        metadata_section = "\n".join(metadata_parts)
+        return f"{metadata_section}\n\n# DOCUMENT SUMMARY\n\n{summary_content}"
+
+    return summary_content

surfsense_backend/app/indexing_pipeline/exceptions.py

@@ -0,0 +1,146 @@
+from litellm.exceptions import (
+    APIConnectionError,
+    APIResponseValidationError,
+    AuthenticationError,
+    BadGatewayError,
+    BadRequestError,
+    InternalServerError,
+    NotFoundError,
+    PermissionDeniedError,
+    RateLimitError,
+    ServiceUnavailableError,
+    Timeout,
+    UnprocessableEntityError,
+)
+from sqlalchemy.exc import IntegrityError as IntegrityError
+
+# Tuples for use directly in except clauses.
+RETRYABLE_LLM_ERRORS = (
+    RateLimitError,
+    Timeout,
+    ServiceUnavailableError,
+    BadGatewayError,
+    InternalServerError,
+    APIConnectionError,
+)
+
+PERMANENT_LLM_ERRORS = (
+    AuthenticationError,
+    PermissionDeniedError,
+    NotFoundError,
+    BadRequestError,
+    UnprocessableEntityError,
+    APIResponseValidationError,
+)
+
+# (LiteLLMEmbeddings, CohereEmbeddings, GeminiEmbeddings all normalize to RuntimeError).
+EMBEDDING_ERRORS = (
+    RuntimeError,  # local device failure or API backend normalization
+    OSError,  # model files missing or corrupted (local backends)
+    MemoryError,  # document too large for available RAM
+    OSError,  # model files missing or corrupted (local backends)
+    MemoryError,  # document too large for available RAM
+)
+
+
+class PipelineMessages:
+    RATE_LIMIT = "LLM rate limit exceeded. Will retry on next sync."
+    LLM_TIMEOUT = "LLM request timed out. Will retry on next sync."
+    LLM_UNAVAILABLE = "LLM service temporarily unavailable. Will retry on next sync."
+    LLM_BAD_GATEWAY = "LLM gateway error. Will retry on next sync."
+    LLM_SERVER_ERROR = "LLM internal server error. Will retry on next sync."
+    LLM_CONNECTION = "Could not reach the LLM service. Check network connectivity."
+    RATE_LIMIT = "LLM rate limit exceeded. Will retry on next sync."
+    LLM_TIMEOUT = "LLM request timed out. Will retry on next sync."
+    LLM_UNAVAILABLE = "LLM service temporarily unavailable. Will retry on next sync."
+    LLM_BAD_GATEWAY = "LLM gateway error. Will retry on next sync."
+    LLM_SERVER_ERROR = "LLM internal server error. Will retry on next sync."
+    LLM_CONNECTION = "Could not reach the LLM service. Check network connectivity."
+
+    LLM_AUTH = "LLM authentication failed. Check your API key."
+    LLM_PERMISSION = "LLM request denied. Check your account permissions."
+    LLM_NOT_FOUND = "LLM model not found. Check your model configuration."
+    LLM_BAD_REQUEST = "LLM rejected the request. Document content may be invalid."
+    LLM_UNPROCESSABLE = (
+        "Document exceeds the LLM context window even after optimization."
+    )
+    LLM_RESPONSE = "LLM returned an invalid response."
+    LLM_AUTH = "LLM authentication failed. Check your API key."
+    LLM_PERMISSION = "LLM request denied. Check your account permissions."
+    LLM_NOT_FOUND = "LLM model not found. Check your model configuration."
+    LLM_BAD_REQUEST = "LLM rejected the request. Document content may be invalid."
+    LLM_UNPROCESSABLE = (
+        "Document exceeds the LLM context window even after optimization."
+    )
+    LLM_RESPONSE = "LLM returned an invalid response."
+
+    EMBEDDING_FAILED = (
+        "Embedding failed. Check your embedding model configuration or service."
+    )
+    EMBEDDING_MODEL = "Embedding model files are missing or corrupted."
+    EMBEDDING_MEMORY = "Not enough memory to embed this document."
+    EMBEDDING_FAILED = (
+        "Embedding failed. Check your embedding model configuration or service."
+    )
+    EMBEDDING_MODEL = "Embedding model files are missing or corrupted."
+    EMBEDDING_MEMORY = "Not enough memory to embed this document."
+
+    CHUNKING_OVERFLOW = "Document structure is too deeply nested to chunk."
+
+
+def safe_exception_message(exc: Exception) -> str:
+    try:
+        return str(exc)
+    except Exception:
+        return "Something went wrong during indexing. Error details could not be retrieved."
+
+
+def llm_retryable_message(exc: Exception) -> str:
+    try:
+        if isinstance(exc, RateLimitError):
+            return PipelineMessages.RATE_LIMIT
+        if isinstance(exc, Timeout):
+            return PipelineMessages.LLM_TIMEOUT
+        if isinstance(exc, ServiceUnavailableError):
+            return PipelineMessages.LLM_UNAVAILABLE
+        if isinstance(exc, BadGatewayError):
+            return PipelineMessages.LLM_BAD_GATEWAY
+        if isinstance(exc, InternalServerError):
+            return PipelineMessages.LLM_SERVER_ERROR
+        if isinstance(exc, APIConnectionError):
+            return PipelineMessages.LLM_CONNECTION
+        return safe_exception_message(exc)
+    except Exception:
+        return "Something went wrong when calling the LLM."
+
+
+def llm_permanent_message(exc: Exception) -> str:
+    try:
+        if isinstance(exc, AuthenticationError):
+            return PipelineMessages.LLM_AUTH
+        if isinstance(exc, PermissionDeniedError):
+            return PipelineMessages.LLM_PERMISSION
+        if isinstance(exc, NotFoundError):
+            return PipelineMessages.LLM_NOT_FOUND
+        if isinstance(exc, BadRequestError):
+            return PipelineMessages.LLM_BAD_REQUEST
+        if isinstance(exc, UnprocessableEntityError):
+            return PipelineMessages.LLM_UNPROCESSABLE
+        if isinstance(exc, APIResponseValidationError):
+            return PipelineMessages.LLM_RESPONSE
+        return safe_exception_message(exc)
+    except Exception:
+        return "Something went wrong when calling the LLM."
+
+
+def embedding_message(exc: Exception) -> str:
+    try:
+        if isinstance(exc, RuntimeError):
+            return PipelineMessages.EMBEDDING_FAILED
+        if isinstance(exc, OSError):
+            return PipelineMessages.EMBEDDING_MODEL
+        if isinstance(exc, MemoryError):
+            return PipelineMessages.EMBEDDING_MEMORY
+        return safe_exception_message(exc)
+    except Exception:
+        return "Something went wrong when generating the embedding."

surfsense_backend/app/indexing_pipeline/indexing_pipeline_service.py

@@ -0,0 +1,237 @@
+import contextlib
+from datetime import UTC, datetime
+
+from sqlalchemy import delete, select
+from sqlalchemy.exc import IntegrityError
+from sqlalchemy.ext.asyncio import AsyncSession
+
+from app.db import Chunk, Document, DocumentStatus
+from app.indexing_pipeline.connector_document import ConnectorDocument
+from app.indexing_pipeline.document_chunker import chunk_text
+from app.indexing_pipeline.document_embedder import embed_text
+from app.indexing_pipeline.document_hashing import (
+    compute_content_hash,
+    compute_unique_identifier_hash,
+)
+from app.indexing_pipeline.document_persistence import (
+    attach_chunks_to_document,
+    rollback_and_persist_failure,
+)
+from app.indexing_pipeline.document_summarizer import summarize_document
+from app.indexing_pipeline.exceptions import (
+    EMBEDDING_ERRORS,
+    PERMANENT_LLM_ERRORS,
+    RETRYABLE_LLM_ERRORS,
+    PipelineMessages,
+    embedding_message,
+    llm_permanent_message,
+    llm_retryable_message,
+    safe_exception_message,
+)
+from app.indexing_pipeline.pipeline_logger import (
+    PipelineLogContext,
+    log_batch_aborted,
+    log_chunking_overflow,
+    log_doc_skipped_unknown,
+    log_document_queued,
+    log_document_requeued,
+    log_document_updated,
+    log_embedding_error,
+    log_index_started,
+    log_index_success,
+    log_permanent_llm_error,
+    log_race_condition,
+    log_retryable_llm_error,
+    log_unexpected_error,
+)
+
+
+class IndexingPipelineService:
+    """Single pipeline for indexing connector documents. All connectors use this service."""
+
+    def __init__(self, session: AsyncSession) -> None:
+        self.session = session
+
+    async def prepare_for_indexing(
+        self, connector_docs: list[ConnectorDocument]
+    ) -> list[Document]:
+        """
+        Persist new documents and detect changes, returning only those that need indexing.
+        """
+        documents = []
+        seen_hashes: set[str] = set()
+        batch_ctx = PipelineLogContext(
+            connector_id=connector_docs[0].connector_id if connector_docs else 0,
+            search_space_id=connector_docs[0].search_space_id if connector_docs else 0,
+            unique_id="batch",
+        )
+
+        for connector_doc in connector_docs:
+            ctx = PipelineLogContext(
+                connector_id=connector_doc.connector_id,
+                search_space_id=connector_doc.search_space_id,
+                unique_id=connector_doc.unique_id,
+            )
+            try:
+                unique_identifier_hash = compute_unique_identifier_hash(connector_doc)
+                content_hash = compute_content_hash(connector_doc)
+
+                if unique_identifier_hash in seen_hashes:
+                    continue
+                seen_hashes.add(unique_identifier_hash)
+
+                result = await self.session.execute(
+                    select(Document).filter(
+                        Document.unique_identifier_hash == unique_identifier_hash
+                    )
+                )
+                existing = result.scalars().first()
+
+                if existing is not None:
+                    if existing.content_hash == content_hash:
+                        if existing.title != connector_doc.title:
+                            existing.title = connector_doc.title
+                            existing.updated_at = datetime.now(UTC)
+                        if not DocumentStatus.is_state(
+                            existing.status, DocumentStatus.READY
+                        ):
+                            existing.status = DocumentStatus.pending()
+                            existing.updated_at = datetime.now(UTC)
+                            documents.append(existing)
+                            log_document_requeued(ctx)
+                        continue
+
+                    existing.title = connector_doc.title
+                    existing.content_hash = content_hash
+                    existing.source_markdown = connector_doc.source_markdown
+                    existing.document_metadata = connector_doc.metadata
+                    existing.updated_at = datetime.now(UTC)
+                    existing.status = DocumentStatus.pending()
+                    documents.append(existing)
+                    log_document_updated(ctx)
+                    continue
+
+                duplicate = await self.session.execute(
+                    select(Document).filter(Document.content_hash == content_hash)
+                )
+                if duplicate.scalars().first() is not None:
+                    continue
+
+                document = Document(
+                    title=connector_doc.title,
+                    document_type=connector_doc.document_type,
+                    content="Pending...",
+                    content_hash=content_hash,
+                    unique_identifier_hash=unique_identifier_hash,
+                    source_markdown=connector_doc.source_markdown,
+                    document_metadata=connector_doc.metadata,
+                    search_space_id=connector_doc.search_space_id,
+                    connector_id=connector_doc.connector_id,
+                    created_by_id=connector_doc.created_by_id,
+                    updated_at=datetime.now(UTC),
+                    status=DocumentStatus.pending(),
+                )
+                self.session.add(document)
+                documents.append(document)
+                log_document_queued(ctx)
+
+            except Exception as e:
+                log_doc_skipped_unknown(ctx, e)
+
+        try:
+            await self.session.commit()
+            return documents
+        except IntegrityError:
+            # A concurrent worker committed a document with the same content_hash
+            # or unique_identifier_hash between our check and our INSERT.
+            # The document already exists — roll back and let the next sync run handle it.
+            log_race_condition(batch_ctx)
+            await self.session.rollback()
+            return []
+        except Exception as e:
+            log_batch_aborted(batch_ctx, e)
+            await self.session.rollback()
+            return []
+
+    async def index(
+        self, document: Document, connector_doc: ConnectorDocument, llm
+    ) -> Document:
+        """
+        Run summarization, embedding, and chunking for a document and persist the results.
+        """
+        ctx = PipelineLogContext(
+            connector_id=connector_doc.connector_id,
+            search_space_id=connector_doc.search_space_id,
+            unique_id=connector_doc.unique_id,
+            doc_id=document.id,
+        )
+        try:
+            log_index_started(ctx)
+            document.status = DocumentStatus.processing()
+            await self.session.commit()
+
+            if connector_doc.should_summarize and llm is not None:
+                content = await summarize_document(
+                    connector_doc.source_markdown, llm, connector_doc.metadata
+                )
+            elif connector_doc.should_summarize and connector_doc.fallback_summary:
+                content = connector_doc.fallback_summary
+            else:
+                content = connector_doc.source_markdown
+
+            embedding = embed_text(content)
+
+            await self.session.execute(
+                delete(Chunk).where(Chunk.document_id == document.id)
+            )
+
+            chunks = [
+                Chunk(content=text, embedding=embed_text(text))
+                for text in chunk_text(
+                    connector_doc.source_markdown,
+                    use_code_chunker=connector_doc.should_use_code_chunker,
+                )
+            ]
+
+            document.content = content
+            document.embedding = embedding
+            attach_chunks_to_document(document, chunks)
+            document.updated_at = datetime.now(UTC)
+            document.status = DocumentStatus.ready()
+            await self.session.commit()
+            log_index_success(ctx, chunk_count=len(chunks))
+
+        except RETRYABLE_LLM_ERRORS as e:
+            log_retryable_llm_error(ctx, e)
+            await rollback_and_persist_failure(
+                self.session, document, llm_retryable_message(e)
+            )
+
+        except PERMANENT_LLM_ERRORS as e:
+            log_permanent_llm_error(ctx, e)
+            await rollback_and_persist_failure(
+                self.session, document, llm_permanent_message(e)
+            )
+
+        except RecursionError as e:
+            log_chunking_overflow(ctx, e)
+            await rollback_and_persist_failure(
+                self.session, document, PipelineMessages.CHUNKING_OVERFLOW
+            )
+
+        except EMBEDDING_ERRORS as e:
+            log_embedding_error(ctx, e)
+            await rollback_and_persist_failure(
+                self.session, document, embedding_message(e)
+            )
+
+        except Exception as e:
+            log_unexpected_error(ctx, e)
+            await rollback_and_persist_failure(
+                self.session, document, safe_exception_message(e)
+            )
+
+        with contextlib.suppress(Exception):
+            await self.session.refresh(document)
+
+        return document

surfsense_backend/app/indexing_pipeline/pipeline_logger.py

@@ -0,0 +1,126 @@
+import logging
+from dataclasses import dataclass
+
+logger = logging.getLogger(__name__)
+
+
+@dataclass
+class PipelineLogContext:
+    connector_id: int | None
+    search_space_id: int
+    unique_id: str  # always available from ConnectorDocument
+    doc_id: int | None = None  # set once the DB row exists (index phase only)
+
+
+class LogMessages:
+    # prepare_for_indexing
+    DOCUMENT_QUEUED = "New document queued for indexing."
+    DOCUMENT_UPDATED = "Document content changed, re-queued for indexing."
+    DOCUMENT_REQUEUED = "Stuck document re-queued for indexing."
+    DOC_SKIPPED_UNKNOWN = "Unexpected error — document skipped."
+    BATCH_ABORTED = "Fatal DB error — aborting prepare batch."
+    RACE_CONDITION = "Concurrent worker beat us to the commit — rolling back batch."
+
+    # index
+    INDEX_STARTED = "Document indexing started."
+    INDEX_SUCCESS = "Document indexed successfully."
+    LLM_RETRYABLE = (
+        "Retryable LLM error — document marked failed, will retry on next sync."
+    )
+    LLM_PERMANENT = "Permanent LLM error — document marked failed."
+    EMBEDDING_FAILED = "Embedding error — document marked failed."
+    CHUNKING_OVERFLOW = "Chunking overflow — document marked failed."
+    UNEXPECTED = "Unexpected error — document marked failed."
+
+
+def _format_context(ctx: PipelineLogContext) -> str:
+    parts = [
+        f"connector_id={ctx.connector_id}",
+        f"search_space_id={ctx.search_space_id}",
+        f"unique_id={ctx.unique_id}",
+    ]
+    if ctx.doc_id is not None:
+        parts.append(f"doc_id={ctx.doc_id}")
+    return " ".join(parts)
+
+
+def _build_message(msg: str, ctx: PipelineLogContext, **extra) -> str:
+    try:
+        parts = [msg, _format_context(ctx)]
+        for key, val in extra.items():
+            parts.append(f"{key}={val}")
+        return " ".join(parts)
+    except Exception:
+        return msg
+
+
+def _safe_log(
+    level_fn, msg: str, ctx: PipelineLogContext, exc_info=None, **extra
+) -> None:
+    # Logging must never raise — a broken log call inside an except block would
+    # chain with the original exception and mask it entirely.
+    try:
+        message = _build_message(msg, ctx, **extra)
+        level_fn(message, exc_info=exc_info)
+    except Exception:
+        pass
+
+
+# ── prepare_for_indexing ──────────────────────────────────────────────────────
+
+
+def log_document_queued(ctx: PipelineLogContext) -> None:
+    _safe_log(logger.info, LogMessages.DOCUMENT_QUEUED, ctx)
+
+
+def log_document_updated(ctx: PipelineLogContext) -> None:
+    _safe_log(logger.info, LogMessages.DOCUMENT_UPDATED, ctx)
+
+
+def log_document_requeued(ctx: PipelineLogContext) -> None:
+    _safe_log(logger.info, LogMessages.DOCUMENT_REQUEUED, ctx)
+
+
+def log_doc_skipped_unknown(ctx: PipelineLogContext, exc: Exception) -> None:
+    _safe_log(
+        logger.warning, LogMessages.DOC_SKIPPED_UNKNOWN, ctx, exc_info=exc, error=exc
+    )
+
+
+def log_race_condition(ctx: PipelineLogContext) -> None:
+    _safe_log(logger.warning, LogMessages.RACE_CONDITION, ctx)
+
+
+def log_batch_aborted(ctx: PipelineLogContext, exc: Exception) -> None:
+    _safe_log(logger.error, LogMessages.BATCH_ABORTED, ctx, exc_info=exc, error=exc)
+
+
+# ── index ─────────────────────────────────────────────────────────────────────
+
+
+def log_index_started(ctx: PipelineLogContext) -> None:
+    _safe_log(logger.info, LogMessages.INDEX_STARTED, ctx)
+
+
+def log_index_success(ctx: PipelineLogContext, chunk_count: int) -> None:
+    _safe_log(logger.info, LogMessages.INDEX_SUCCESS, ctx, chunk_count=chunk_count)
+
+
+def log_retryable_llm_error(ctx: PipelineLogContext, exc: Exception) -> None:
+    _safe_log(logger.warning, LogMessages.LLM_RETRYABLE, ctx, exc_info=exc, error=exc)
+
+
+def log_permanent_llm_error(ctx: PipelineLogContext, exc: Exception) -> None:
+    _safe_log(logger.error, LogMessages.LLM_PERMANENT, ctx, exc_info=exc, error=exc)
+
+
+def log_embedding_error(ctx: PipelineLogContext, exc: Exception) -> None:
+    _safe_log(logger.error, LogMessages.EMBEDDING_FAILED, ctx, exc_info=exc, error=exc)
+
+
+def log_chunking_overflow(ctx: PipelineLogContext, exc: Exception) -> None:
+    _safe_log(logger.error, LogMessages.CHUNKING_OVERFLOW, ctx, exc_info=exc, error=exc)
+
+
+def log_unexpected_error(ctx: PipelineLogContext, exc: Exception) -> None:
+    _safe_log(logger.error, LogMessages.UNEXPECTED, ctx, exc_info=exc, error=exc)

surfsense_backend/app/routes/__init__.py

@@ -36,6 +36,7 @@ from .podcasts_routes import router as podcasts_router
 from .public_chat_routes import router as public_chat_router
 from .rbac_routes import router as rbac_router
 from .reports_routes import router as reports_router
+from .sandbox_routes import router as sandbox_router
 from .search_source_connectors_routes import router as search_source_connectors_router
 from .search_spaces_routes import router as search_spaces_router
 from .slack_add_connector_route import router as slack_add_connector_router
@@ -50,6 +51,7 @@ router.include_router(editor_router)
 router.include_router(documents_router)
 router.include_router(notes_router)
 router.include_router(new_chat_router)  # Chat with assistant-ui persistence
+router.include_router(sandbox_router)  # Sandbox file downloads (Daytona)
 router.include_router(chat_comments_router)
 router.include_router(podcasts_router)  # Podcast task status and audio
 router.include_router(reports_router)  # Report CRUD and export (PDF/DOCX)

surfsense_backend/app/routes/documents_routes.py

@@ -28,6 +28,7 @@ from app.schemas import (
     DocumentWithChunksRead,
     PaginatedResponse,
 )
+from app.services.task_dispatcher import TaskDispatcher, get_task_dispatcher
 from app.users import current_active_user
 from app.utils.rbac import check_permission
 
@@ -44,6 +45,10 @@ os.environ["UNSTRUCTURED_HAS_PATCHED_LOOP"] = "1"
 
 router = APIRouter()
 
+MAX_FILES_PER_UPLOAD = 10
+MAX_FILE_SIZE_BYTES = 50 * 1024 * 1024  # 50 MB per file
+MAX_TOTAL_SIZE_BYTES = 200 * 1024 * 1024  # 200 MB total
+
 
 @router.post("/documents")
 async def create_documents(
@@ -114,8 +119,10 @@ async def create_documents(
 async def create_documents_file_upload(
     files: list[UploadFile],
     search_space_id: int = Form(...),
+    should_summarize: bool = Form(False),
     session: AsyncSession = Depends(get_async_session),
     user: User = Depends(current_active_user),
+    dispatcher: TaskDispatcher = Depends(get_task_dispatcher),
 ):
     """
     Upload files as documents with real-time status tracking.
@@ -148,12 +155,37 @@ async def create_documents_file_upload(
         if not files:
             raise HTTPException(status_code=400, detail="No files provided")
 
+        if len(files) > MAX_FILES_PER_UPLOAD:
+            raise HTTPException(
+                status_code=413,
+                detail=f"Too many files. Maximum {MAX_FILES_PER_UPLOAD} files per upload.",
+            )
+
+        total_size = 0
+        for file in files:
+            file_size = file.size or 0
+            if file_size > MAX_FILE_SIZE_BYTES:
+                raise HTTPException(
+                    status_code=413,
+                    detail=f"File '{file.filename}' ({file_size / (1024 * 1024):.1f} MB) "
+                    f"exceeds the {MAX_FILE_SIZE_BYTES // (1024 * 1024)} MB per-file limit.",
+                )
+            total_size += file_size
+
+        if total_size > MAX_TOTAL_SIZE_BYTES:
+            raise HTTPException(
+                status_code=413,
+                detail=f"Total upload size ({total_size / (1024 * 1024):.1f} MB) "
+                f"exceeds the {MAX_TOTAL_SIZE_BYTES // (1024 * 1024)} MB limit.",
+            )
+
         created_documents: list[Document] = []
         files_to_process: list[
             tuple[Document, str, str]
         ] = []  # (document, temp_path, filename)
         skipped_duplicates = 0
         duplicate_document_ids: list[int] = []
+        actual_total_size = 0
 
         # ===== PHASE 1: Create pending documents for all files =====
         # This makes ALL documents visible in the UI immediately with pending status
@@ -169,11 +201,28 @@ async def create_documents_file_upload(
                     temp_path = temp_file.name
 
                 content = await file.read()
+                file_size = len(content)
+
+                if file_size > MAX_FILE_SIZE_BYTES:
+                    os.unlink(temp_path)
+                    raise HTTPException(
+                        status_code=413,
+                        detail=f"File '{file.filename}' ({file_size / (1024 * 1024):.1f} MB) "
+                        f"exceeds the {MAX_FILE_SIZE_BYTES // (1024 * 1024)} MB per-file limit.",
+                    )
+
+                actual_total_size += file_size
+                if actual_total_size > MAX_TOTAL_SIZE_BYTES:
+                    os.unlink(temp_path)
+                    raise HTTPException(
+                        status_code=413,
+                        detail=f"Total upload size ({actual_total_size / (1024 * 1024):.1f} MB) "
+                        f"exceeds the {MAX_TOTAL_SIZE_BYTES // (1024 * 1024)} MB limit.",
+                    )
+
                 with open(temp_path, "wb") as f:
                     f.write(content)
 
-                file_size = len(content)
-
                 # Generate unique identifier for deduplication check
                 unique_identifier_hash = generate_unique_identifier_hash(
                     DocumentType.FILE, file.filename or "unknown", search_space_id
@@ -244,19 +293,16 @@ async def create_documents_file_upload(
             for doc in created_documents:
                 await session.refresh(doc)
 
-        # ===== PHASE 2: Dispatch Celery tasks for each file =====
+        # ===== PHASE 2: Dispatch tasks for each file =====
         # Each task will update document status: pending → processing → ready/failed
-        from app.tasks.celery_tasks.document_tasks import (
-            process_file_upload_with_document_task,
-        )
-
         for document, temp_path, filename in files_to_process:
-            process_file_upload_with_document_task.delay(
+            await dispatcher.dispatch_file_processing(
                 document_id=document.id,
                 temp_path=temp_path,
                 filename=filename,
                 search_space_id=search_space_id,
                 user_id=str(user.id),
+                should_summarize=should_summarize,
             )
 
         return {
@@ -373,10 +419,11 @@ async def read_documents(
         # Convert database objects to API-friendly format
         api_documents = []
         for doc in db_documents:
-            # Get user name (display_name or email fallback)
             created_by_name = None
+            created_by_email = None
             if doc.created_by:
-                created_by_name = doc.created_by.display_name or doc.created_by.email
+                created_by_name = doc.created_by.display_name
+                created_by_email = doc.created_by.email
 
             # Parse status from JSONB
             status_data = None
@@ -400,6 +447,7 @@ async def read_documents(
                     search_space_id=doc.search_space_id,
                     created_by_id=doc.created_by_id,
                     created_by_name=created_by_name,
+                    created_by_email=created_by_email,
                     status=status_data,
                 )
             )
@@ -528,10 +576,11 @@ async def search_documents(
         # Convert database objects to API-friendly format
         api_documents = []
         for doc in db_documents:
-            # Get user name (display_name or email fallback)
             created_by_name = None
+            created_by_email = None
             if doc.created_by:
-                created_by_name = doc.created_by.display_name or doc.created_by.email
+                created_by_name = doc.created_by.display_name
+                created_by_email = doc.created_by.email
 
             # Parse status from JSONB
             status_data = None
@@ -555,6 +604,7 @@ async def search_documents(
                     search_space_id=doc.search_space_id,
                     created_by_id=doc.created_by_id,
                     created_by_name=created_by_name,
+                    created_by_email=created_by_email,
                     status=status_data,
                 )
             )

surfsense_backend/app/routes/google_drive_add_connector_route.py

@@ -76,9 +76,9 @@ def get_token_encryption() -> TokenEncryption:
 
 # Google Drive OAuth scopes
 SCOPES = [
-    "https://www.googleapis.com/auth/drive.readonly",  # Read-only access to Drive
+    "https://www.googleapis.com/auth/drive",
-    "https://www.googleapis.com/auth/userinfo.email",  # User email
+    "https://www.googleapis.com/auth/userinfo.email",
-    "https://www.googleapis.com/auth/userinfo.profile",  # User profile
+    "https://www.googleapis.com/auth/userinfo.profile",
     "openid",
 ]
 
@@ -151,6 +151,75 @@ async def connect_drive(space_id: int, user: User = Depends(current_active_user)
         ) from e
 
 
+@router.get("/auth/google/drive/connector/reauth")
+async def reauth_drive(
+    space_id: int,
+    connector_id: int,
+    return_url: str | None = None,
+    user: User = Depends(current_active_user),
+    session: AsyncSession = Depends(get_async_session),
+):
+    """
+    Initiate Google Drive re-authentication to upgrade OAuth scopes.
+
+    Query params:
+        space_id: Search space ID the connector belongs to
+        connector_id: ID of the existing connector to re-authenticate
+
+    Returns:
+        JSON with auth_url to redirect user to Google authorization
+    """
+    try:
+        result = await session.execute(
+            select(SearchSourceConnector).filter(
+                SearchSourceConnector.id == connector_id,
+                SearchSourceConnector.user_id == user.id,
+                SearchSourceConnector.search_space_id == space_id,
+                SearchSourceConnector.connector_type
+                == SearchSourceConnectorType.GOOGLE_DRIVE_CONNECTOR,
+            )
+        )
+        connector = result.scalars().first()
+        if not connector:
+            raise HTTPException(
+                status_code=404,
+                detail="Google Drive connector not found or access denied",
+            )
+
+        if not config.SECRET_KEY:
+            raise HTTPException(
+                status_code=500, detail="SECRET_KEY not configured for OAuth security."
+            )
+
+        flow = get_google_flow()
+
+        state_manager = get_state_manager()
+        extra: dict = {"connector_id": connector_id}
+        if return_url and return_url.startswith("/"):
+            extra["return_url"] = return_url
+        state_encoded = state_manager.generate_secure_state(space_id, user.id, **extra)
+
+        auth_url, _ = flow.authorization_url(
+            access_type="offline",
+            prompt="consent",
+            include_granted_scopes="true",
+            state=state_encoded,
+        )
+
+        logger.info(
+            f"Initiating Google Drive re-auth for user {user.id}, connector {connector_id}"
+        )
+        return {"auth_url": auth_url}
+
+    except HTTPException:
+        raise
+    except Exception as e:
+        logger.error(f"Failed to initiate Google Drive re-auth: {e!s}", exc_info=True)
+        raise HTTPException(
+            status_code=500, detail=f"Failed to initiate Google re-auth: {e!s}"
+        ) from e
+
+
 @router.get("/auth/google/drive/connector/callback")
 async def drive_callback(
     request: Request,
@@ -214,6 +283,8 @@ async def drive_callback(
 
         user_id = UUID(data["user_id"])
         space_id = data["space_id"]
+        reauth_connector_id = data.get("connector_id")
+        reauth_return_url = data.get("return_url")
 
         logger.info(
             f"Processing Google Drive callback for user {user_id}, space {space_id}"
@@ -253,7 +324,45 @@ async def drive_callback(
         # Mark that credentials are encrypted for backward compatibility
         creds_dict["_token_encrypted"] = True
 
-        # Check for duplicate connector (same account already connected)
+        if reauth_connector_id:
+            result = await session.execute(
+                select(SearchSourceConnector).filter(
+                    SearchSourceConnector.id == reauth_connector_id,
+                    SearchSourceConnector.user_id == user_id,
+                    SearchSourceConnector.search_space_id == space_id,
+                    SearchSourceConnector.connector_type
+                    == SearchSourceConnectorType.GOOGLE_DRIVE_CONNECTOR,
+                )
+            )
+            db_connector = result.scalars().first()
+            if not db_connector:
+                raise HTTPException(
+                    status_code=404,
+                    detail="Connector not found or access denied during re-auth",
+                )
+
+            existing_start_page_token = db_connector.config.get("start_page_token")
+            db_connector.config = {
+                **creds_dict,
+                "start_page_token": existing_start_page_token,
+            }
+            from sqlalchemy.orm.attributes import flag_modified
+
+            flag_modified(db_connector, "config")
+            await session.commit()
+            await session.refresh(db_connector)
+
+            logger.info(
+                f"Re-authenticated Google Drive connector {db_connector.id} for user {user_id}"
+            )
+            if reauth_return_url and reauth_return_url.startswith("/"):
+                return RedirectResponse(
+                    url=f"{config.NEXT_FRONTEND_URL}{reauth_return_url}"
+                )
+            return RedirectResponse(
+                url=f"{config.NEXT_FRONTEND_URL}/dashboard/{space_id}/new-chat?modal=connectors&tab=all&success=true&connector=google-drive-connector&connectorId={db_connector.id}"
+            )
+
         is_duplicate = await check_duplicate_connector(
             session,
             SearchSourceConnectorType.GOOGLE_DRIVE_CONNECTOR,

surfsense_backend/app/routes/new_chat_routes.py

@@ -10,6 +10,8 @@ These endpoints support the ThreadHistoryAdapter pattern from assistant-ui:
 - POST /threads/{thread_id}/messages - Append message
 """
 
+import asyncio
+import logging
 from datetime import UTC, datetime
 
 from fastapi import APIRouter, Depends, HTTPException, Request
@@ -52,9 +54,50 @@ from app.tasks.chat.stream_new_chat import stream_new_chat, stream_resume_chat
 from app.users import current_active_user
 from app.utils.rbac import check_permission
 
+_logger = logging.getLogger(__name__)
+_background_tasks: set[asyncio.Task] = set()
+
 router = APIRouter()
 
 
+def _try_delete_sandbox(thread_id: int) -> None:
+    """Fire-and-forget sandbox + local file deletion so the HTTP response isn't blocked."""
+    from app.agents.new_chat.sandbox import (
+        delete_local_sandbox_files,
+        delete_sandbox,
+        is_sandbox_enabled,
+    )
+
+    if not is_sandbox_enabled():
+        return
+
+    async def _bg() -> None:
+        try:
+            await delete_sandbox(thread_id)
+        except Exception:
+            _logger.warning(
+                "Background sandbox delete failed for thread %s",
+                thread_id,
+                exc_info=True,
+            )
+        try:
+            delete_local_sandbox_files(thread_id)
+        except Exception:
+            _logger.warning(
+                "Local sandbox file cleanup failed for thread %s",
+                thread_id,
+                exc_info=True,
+            )
+
+    try:
+        loop = asyncio.get_running_loop()
+        task = loop.create_task(_bg())
+        _background_tasks.add(task)
+        task.add_done_callback(_background_tasks.discard)
+    except RuntimeError:
+        pass
+
+
 async def check_thread_access(
     session: AsyncSession,
     thread: NewChatThread,
@@ -648,6 +691,9 @@ async def delete_thread(
 
         await session.delete(db_thread)
         await session.commit()
+
+        _try_delete_sandbox(thread_id)
+
         return {"message": "Thread deleted successfully"}
 
     except HTTPException:

surfsense_backend/app/routes/reports_routes.py

@@ -17,7 +17,7 @@ import logging
 import os
 import re
 import tempfile
-from enum import Enum
+from enum import StrEnum
 
 import pypandoc
 import typst
@@ -46,7 +46,7 @@ router = APIRouter()
 MAX_REPORT_LIST_LIMIT = 500
 
 
-class ExportFormat(str, Enum):
+class ExportFormat(StrEnum):
     PDF = "pdf"
     DOCX = "docx"
 

surfsense_backend/app/routes/sandbox_routes.py

@@ -0,0 +1,105 @@
+"""Routes for downloading files from Daytona sandbox environments."""
+
+from __future__ import annotations
+
+import asyncio
+import logging
+
+from fastapi import APIRouter, Depends, HTTPException, Query
+from fastapi.responses import Response
+from sqlalchemy.ext.asyncio import AsyncSession
+from sqlalchemy.future import select
+
+from app.db import NewChatThread, Permission, User, get_async_session
+from app.users import current_active_user
+from app.utils.rbac import check_permission
+
+logger = logging.getLogger(__name__)
+
+router = APIRouter()
+
+MIME_TYPES: dict[str, str] = {
+    ".png": "image/png",
+    ".jpg": "image/jpeg",
+    ".jpeg": "image/jpeg",
+    ".gif": "image/gif",
+    ".webp": "image/webp",
+    ".svg": "image/svg+xml",
+    ".pdf": "application/pdf",
+    ".csv": "text/csv",
+    ".json": "application/json",
+    ".txt": "text/plain",
+    ".html": "text/html",
+    ".md": "text/markdown",
+    ".py": "text/x-python",
+    ".xlsx": "application/vnd.openxmlformats-officedocument.spreadsheetml.sheet",
+    ".zip": "application/zip",
+}
+
+
+def _guess_media_type(filename: str) -> str:
+    ext = ("." + filename.rsplit(".", 1)[-1].lower()) if "." in filename else ""
+    return MIME_TYPES.get(ext, "application/octet-stream")
+
+
+@router.get("/threads/{thread_id}/sandbox/download")
+async def download_sandbox_file(
+    thread_id: int,
+    path: str = Query(..., description="Absolute path of the file inside the sandbox"),
+    session: AsyncSession = Depends(get_async_session),
+    user: User = Depends(current_active_user),
+):
+    """Download a file from the Daytona sandbox associated with a chat thread."""
+
+    from app.agents.new_chat.sandbox import get_or_create_sandbox, is_sandbox_enabled
+
+    if not is_sandbox_enabled():
+        raise HTTPException(status_code=404, detail="Sandbox is not enabled")
+
+    result = await session.execute(
+        select(NewChatThread).filter(NewChatThread.id == thread_id)
+    )
+    thread = result.scalars().first()
+    if not thread:
+        raise HTTPException(status_code=404, detail="Thread not found")
+
+    await check_permission(
+        session,
+        user,
+        thread.search_space_id,
+        Permission.CHATS_READ.value,
+        "You don't have permission to access files in this thread",
+    )
+
+    from app.agents.new_chat.sandbox import get_local_sandbox_file
+
+    # Prefer locally-persisted copy (sandbox may already be deleted)
+    local_content = get_local_sandbox_file(thread_id, path)
+    if local_content is not None:
+        filename = path.rsplit("/", 1)[-1] if "/" in path else path
+        media_type = _guess_media_type(filename)
+        return Response(
+            content=local_content,
+            media_type=media_type,
+            headers={"Content-Disposition": f'attachment; filename="{filename}"'},
+        )
+
+    # Fall back to live sandbox download
+    try:
+        sandbox = await get_or_create_sandbox(thread_id)
+        raw_sandbox = sandbox._sandbox
+        content: bytes = await asyncio.to_thread(raw_sandbox.fs.download_file, path)
+    except Exception as exc:
+        logger.warning("Sandbox file download failed for %s: %s", path, exc)
+        raise HTTPException(
+            status_code=404, detail=f"Could not download file: {exc}"
+        ) from exc
+
+    filename = path.rsplit("/", 1)[-1] if "/" in path else path
+    media_type = _guess_media_type(filename)
+
+    return Response(
+        content=content,
+        media_type=media_type,
+        headers={"Content-Disposition": f'attachment; filename="{filename}"'},
+    )

surfsense_backend/app/routes/search_source_connectors_routes.py

@@ -2735,7 +2735,10 @@ async def create_mcp_connector(
             f"for user {user.id} in search space {search_space_id}"
         )
 
-        # Convert to read schema
+        from app.agents.new_chat.tools.mcp_tool import invalidate_mcp_tools_cache
+
+        invalidate_mcp_tools_cache(search_space_id)
+
         connector_read = SearchSourceConnectorRead.model_validate(db_connector)
         return MCPConnectorRead.from_connector(connector_read)
 
@@ -2910,6 +2913,10 @@ async def update_mcp_connector(
 
         logger.info(f"Updated MCP connector {connector_id}")
 
+        from app.agents.new_chat.tools.mcp_tool import invalidate_mcp_tools_cache
+
+        invalidate_mcp_tools_cache(connector.search_space_id)
+
         connector_read = SearchSourceConnectorRead.model_validate(connector)
         return MCPConnectorRead.from_connector(connector_read)
 
@@ -2960,9 +2967,14 @@ async def delete_mcp_connector(
             "You don't have permission to delete this connector",
         )
 
+        search_space_id = connector.search_space_id
         await session.delete(connector)
         await session.commit()
 
+        from app.agents.new_chat.tools.mcp_tool import invalidate_mcp_tools_cache
+
+        invalidate_mcp_tools_cache(search_space_id)
+
         logger.info(f"Deleted MCP connector {connector_id}")
 
     except HTTPException:

surfsense_backend/app/schemas/documents.py

@@ -60,9 +60,8 @@ class DocumentRead(BaseModel):
     updated_at: datetime | None
     search_space_id: int
     created_by_id: UUID | None = None  # User who created/uploaded this document
-    created_by_name: str | None = (
+    created_by_name: str | None = None
-        None  # Display name or email of the user who created this document
+    created_by_email: str | None = None
-    )
     status: DocumentStatusSchema | None = (
         None  # Processing status (ready, processing, failed)
     )

surfsense_backend/app/schemas/podcasts.py

@@ -1,13 +1,13 @@
 """Podcast schemas for API responses."""
 
 from datetime import datetime
-from enum import Enum
+from enum import StrEnum
 from typing import Any
 
 from pydantic import BaseModel
 
 
-class PodcastStatusEnum(str, Enum):
+class PodcastStatusEnum(StrEnum):
     PENDING = "pending"
     GENERATING = "generating"
     READY = "ready"

surfsense_backend/app/schemas/search_source_connector.py

@@ -16,6 +16,7 @@ class SearchSourceConnectorBase(BaseModel):
     is_indexable: bool
     last_indexed_at: datetime | None = None
     config: dict[str, Any]
+    enable_summary: bool = False
     periodic_indexing_enabled: bool = False
     indexing_frequency_minutes: int | None = None
     next_scheduled_at: datetime | None = None
@@ -65,6 +66,7 @@ class SearchSourceConnectorUpdate(BaseModel):
     is_indexable: bool | None = None
     last_indexed_at: datetime | None = None
     config: dict[str, Any] | None = None
+    enable_summary: bool | None = None
     periodic_indexing_enabled: bool | None = None
     indexing_frequency_minutes: int | None = None
     next_scheduled_at: datetime | None = None

surfsense_backend/app/services/connector_service.py

@@ -1303,10 +1303,9 @@ class ConnectorService:
 
         sources_list = self._build_chunk_sources_from_documents(
             github_docs,
-            description_fn=lambda chunk, _doc_info, metadata: metadata.get(
+            description_fn=lambda chunk, _doc_info, metadata: (
-                "description"
+                metadata.get("description") or chunk.get("content", "")
-            )
+            ),
-            or chunk.get("content", ""),
             url_fn=lambda _doc_info, metadata: metadata.get("url", "") or "",
         )
 

surfsense_backend/app/services/google_drive/__init__.py

@@ -0,0 +1,11 @@
+from app.services.google_drive.tool_metadata_service import (
+    GoogleDriveAccount,
+    GoogleDriveFile,
+    GoogleDriveToolMetadataService,
+)
+
+__all__ = [
+    "GoogleDriveAccount",
+    "GoogleDriveFile",
+    "GoogleDriveToolMetadataService",
+]

surfsense_backend/app/services/google_drive/tool_metadata_service.py

@@ -0,0 +1,149 @@
+from dataclasses import dataclass
+
+from sqlalchemy import and_, func
+from sqlalchemy.ext.asyncio import AsyncSession
+from sqlalchemy.future import select
+
+from app.db import (
+    Document,
+    DocumentType,
+    SearchSourceConnector,
+    SearchSourceConnectorType,
+)
+
+
+@dataclass
+class GoogleDriveAccount:
+    id: int
+    name: str
+
+    @classmethod
+    def from_connector(cls, connector: SearchSourceConnector) -> "GoogleDriveAccount":
+        return cls(id=connector.id, name=connector.name)
+
+    def to_dict(self) -> dict:
+        return {"id": self.id, "name": self.name}
+
+
+@dataclass
+class GoogleDriveFile:
+    file_id: str
+    name: str
+    mime_type: str
+    web_view_link: str
+    connector_id: int
+    document_id: int
+
+    @classmethod
+    def from_document(cls, document: Document) -> "GoogleDriveFile":
+        meta = document.document_metadata or {}
+        return cls(
+            file_id=meta.get("google_drive_file_id", ""),
+            name=meta.get("google_drive_file_name", document.title),
+            mime_type=meta.get("google_drive_mime_type", ""),
+            web_view_link=meta.get("web_view_link", ""),
+            connector_id=document.connector_id,
+            document_id=document.id,
+        )
+
+    def to_dict(self) -> dict:
+        return {
+            "file_id": self.file_id,
+            "name": self.name,
+            "mime_type": self.mime_type,
+            "web_view_link": self.web_view_link,
+            "connector_id": self.connector_id,
+            "document_id": self.document_id,
+        }
+
+
+class GoogleDriveToolMetadataService:
+    def __init__(self, db_session: AsyncSession):
+        self._db_session = db_session
+
+    async def get_creation_context(self, search_space_id: int, user_id: str) -> dict:
+        accounts = await self._get_google_drive_accounts(search_space_id, user_id)
+
+        if not accounts:
+            return {
+                "accounts": [],
+                "supported_types": [],
+                "error": "No Google Drive account connected",
+            }
+
+        return {
+            "accounts": [acc.to_dict() for acc in accounts],
+            "supported_types": ["google_doc", "google_sheet"],
+        }
+
+    async def get_trash_context(
+        self, search_space_id: int, user_id: str, file_name: str
+    ) -> dict:
+        result = await self._db_session.execute(
+            select(Document)
+            .join(
+                SearchSourceConnector, Document.connector_id == SearchSourceConnector.id
+            )
+            .filter(
+                and_(
+                    Document.search_space_id == search_space_id,
+                    Document.document_type == DocumentType.GOOGLE_DRIVE_FILE,
+                    func.lower(Document.title) == func.lower(file_name),
+                    SearchSourceConnector.user_id == user_id,
+                )
+            )
+        )
+        document = result.scalars().first()
+
+        if not document:
+            return {
+                "error": (
+                    f"File '{file_name}' not found in your indexed Google Drive files. "
+                    "This could mean: (1) the file doesn't exist, (2) it hasn't been indexed yet, "
+                    "or (3) the file name is different."
+                )
+            }
+
+        if not document.connector_id:
+            return {"error": "Document has no associated connector"}
+
+        result = await self._db_session.execute(
+            select(SearchSourceConnector).filter(
+                and_(
+                    SearchSourceConnector.id == document.connector_id,
+                    SearchSourceConnector.user_id == user_id,
+                    SearchSourceConnector.connector_type
+                    == SearchSourceConnectorType.GOOGLE_DRIVE_CONNECTOR,
+                )
+            )
+        )
+        connector = result.scalars().first()
+
+        if not connector:
+            return {"error": "Connector not found or access denied"}
+
+        account = GoogleDriveAccount.from_connector(connector)
+        file = GoogleDriveFile.from_document(document)
+
+        return {
+            "account": account.to_dict(),
+            "file": file.to_dict(),
+        }
+
+    async def _get_google_drive_accounts(
+        self, search_space_id: int, user_id: str
+    ) -> list[GoogleDriveAccount]:
+        result = await self._db_session.execute(
+            select(SearchSourceConnector)
+            .filter(
+                and_(
+                    SearchSourceConnector.search_space_id == search_space_id,
+                    SearchSourceConnector.user_id == user_id,
+                    SearchSourceConnector.connector_type
+                    == SearchSourceConnectorType.GOOGLE_DRIVE_CONNECTOR,
+                )
+            )
+            .order_by(SearchSourceConnector.last_indexed_at.desc())
+        )
+        connectors = result.scalars().all()
+        return [GoogleDriveAccount.from_connector(c) for c in connectors]

surfsense_backend/app/services/linear/kb_sync_service.py

@@ -4,12 +4,12 @@ from datetime import datetime
 from sqlalchemy import delete
 from sqlalchemy.ext.asyncio import AsyncSession
 
-from app.config import config
 from app.connectors.linear_connector import LinearConnector
 from app.db import Chunk, Document
 from app.services.llm_service import get_user_long_context_llm
 from app.utils.document_converters import (
     create_document_chunks,
+    embed_text,
     generate_content_hash,
     generate_document_summary,
 )
@@ -80,7 +80,7 @@ class LinearKBSyncService:
             state = formatted_issue.get("state", "Unknown")
             priority = issue_raw.get("priorityLabel", "Unknown")
             comment_count = len(formatted_issue.get("comments", []))
-            description = formatted_issue.get("description", "")
+            formatted_issue.get("description", "")
 
             user_llm = await get_user_long_context_llm(
                 self.db_session, user_id, search_space_id, disable_streaming=True
@@ -100,18 +100,10 @@ class LinearKBSyncService:
                     issue_content, user_llm, document_metadata_for_summary
                 )
             else:
-                if description and len(description) > 1000:
-                    description = description[:997] + "..."
                 summary_content = (
-                    f"Linear Issue {issue_identifier}: {issue_title}\n\n"
+                    f"Linear Issue {issue_identifier}: {issue_title}\n\n{issue_content}"
-                    f"Status: {state}\n\n"
-                )
-                if description:
-                    summary_content += f"Description: {description}\n\n"
-                summary_content += f"Comments: {comment_count}"
-                summary_embedding = config.embedding_model_instance.embed(
-                    summary_content
                 )
+                summary_embedding = embed_text(summary_content)
 
             await self.db_session.execute(
                 delete(Chunk).where(Chunk.document_id == document.id)

surfsense_backend/app/services/llm_router_service.py

@@ -12,16 +12,35 @@ synchronous ChatLiteLLM-like interface and async methods.
 """
 
 import logging
+import re
 from typing import Any
 
 from langchain_core.callbacks import CallbackManagerForLLMRun
+from langchain_core.exceptions import ContextOverflowError
 from langchain_core.language_models import BaseChatModel
 from langchain_core.messages import AIMessage, AIMessageChunk, BaseMessage
 from langchain_core.outputs import ChatGeneration, ChatGenerationChunk, ChatResult
 from litellm import Router
+from litellm.exceptions import (
+    BadRequestError as LiteLLMBadRequestError,
+    ContextWindowExceededError,
+)
 
 logger = logging.getLogger(__name__)
 
+_CONTEXT_OVERFLOW_PATTERNS = re.compile(
+    r"(input tokens exceed|context.{0,20}(length|window|limit)|"
+    r"maximum context length|token.{0,20}(limit|exceed)|"
+    r"too many tokens|reduce the length)",
+    re.IGNORECASE,
+)
+
+
+def _is_context_overflow_error(exc: LiteLLMBadRequestError) -> bool:
+    """Check if a BadRequestError is actually a context window overflow."""
+    return bool(_CONTEXT_OVERFLOW_PATTERNS.search(str(exc)))
+
+
 # Special ID for Auto mode - uses router for load balancing
 AUTO_MODE_ID = 0
 
@@ -234,6 +253,10 @@ class ChatLiteLLMRouter(BaseChatModel):
 
     This wraps the LiteLLM Router to provide the same interface as ChatLiteLLM,
     making it a drop-in replacement for auto-mode routing.
+
+    Exposes a ``profile`` with ``max_input_tokens`` set to the smallest context
+    window across all router deployments so that deepagents
+    SummarizationMiddleware can use fraction-based triggers.
     """
 
     # Use model_config for Pydantic v2 compatibility
@@ -265,7 +288,6 @@ class ChatLiteLLMRouter(BaseChatModel):
         """
         try:
             super().__init__(**kwargs)
-            # Store router and tools as private attributes
             resolved_router = router or LLMRouterService.get_router()
             object.__setattr__(self, "_router", resolved_router)
             object.__setattr__(self, "_bound_tools", bound_tools)
@@ -274,6 +296,12 @@ class ChatLiteLLMRouter(BaseChatModel):
                 raise ValueError(
                     "LLM Router not initialized. Call LLMRouterService.initialize() first."
                 )
+
+            # Set profile so deepagents SummarizationMiddleware gets fraction-based triggers
+            computed_profile = self._compute_min_context_profile()
+            if computed_profile is not None:
+                object.__setattr__(self, "profile", computed_profile)
+
             logger.info(
                 f"ChatLiteLLMRouter initialized with {LLMRouterService.get_model_count()} models"
             )
@@ -281,6 +309,39 @@ class ChatLiteLLMRouter(BaseChatModel):
             logger.error(f"Failed to initialize ChatLiteLLMRouter: {e}")
             raise
 
+    def _compute_min_context_profile(self) -> dict | None:
+        """Derive a profile dict with max_input_tokens from router deployments.
+
+        Uses litellm.get_model_info to look up each deployment's context window
+        and picks the *minimum* so that summarization triggers before ANY model
+        in the pool overflows.
+        """
+        from litellm import get_model_info
+
+        if not self._router:
+            return None
+
+        min_ctx: int | None = None
+        for deployment in self._router.model_list:
+            params = deployment.get("litellm_params", {})
+            base_model = params.get("base_model") or params.get("model", "")
+            try:
+                info = get_model_info(base_model)
+                ctx = info.get("max_input_tokens")
+                if (
+                    isinstance(ctx, int)
+                    and ctx > 0
+                    and (min_ctx is None or ctx < min_ctx)
+                ):
+                    min_ctx = ctx
+            except Exception:
+                continue
+
+        if min_ctx is not None:
+            logger.info(f"ChatLiteLLMRouter profile: max_input_tokens={min_ctx}")
+            return {"max_input_tokens": min_ctx}
+        return None
+
     @property
     def _llm_type(self) -> str:
         return "litellm-router"
@@ -359,13 +420,19 @@ class ChatLiteLLMRouter(BaseChatModel):
         if self._tool_choice is not None:
             call_kwargs["tool_choice"] = self._tool_choice
 
-        # Call router completion
+        try:
-        response = self._router.completion(
+            response = self._router.completion(
-            model=self.model,
+                model=self.model,
-            messages=formatted_messages,
+                messages=formatted_messages,
-            stop=stop,
+                stop=stop,
-            **call_kwargs,
+                **call_kwargs,
-        )
+            )
+        except ContextWindowExceededError as e:
+            raise ContextOverflowError(str(e)) from e
+        except LiteLLMBadRequestError as e:
+            if _is_context_overflow_error(e):
+                raise ContextOverflowError(str(e)) from e
+            raise
 
         # Convert response to ChatResult with potential tool calls
         message = self._convert_response_to_message(response.choices[0].message)
@@ -396,13 +463,19 @@ class ChatLiteLLMRouter(BaseChatModel):
         if self._tool_choice is not None:
             call_kwargs["tool_choice"] = self._tool_choice
 
-        # Call router async completion
+        try:
-        response = await self._router.acompletion(
+            response = await self._router.acompletion(
-            model=self.model,
+                model=self.model,
-            messages=formatted_messages,
+                messages=formatted_messages,
-            stop=stop,
+                stop=stop,
-            **call_kwargs,
+                **call_kwargs,
-        )
+            )
+        except ContextWindowExceededError as e:
+            raise ContextOverflowError(str(e)) from e
+        except LiteLLMBadRequestError as e:
+            if _is_context_overflow_error(e):
+                raise ContextOverflowError(str(e)) from e
+            raise
 
         # Convert response to ChatResult with potential tool calls
         message = self._convert_response_to_message(response.choices[0].message)
@@ -432,14 +505,20 @@ class ChatLiteLLMRouter(BaseChatModel):
         if self._tool_choice is not None:
             call_kwargs["tool_choice"] = self._tool_choice
 
-        # Call router completion with streaming
+        try:
-        response = self._router.completion(
+            response = self._router.completion(
-            model=self.model,
+                model=self.model,
-            messages=formatted_messages,
+                messages=formatted_messages,
-            stop=stop,
+                stop=stop,
-            stream=True,
+                stream=True,
-            **call_kwargs,
+                **call_kwargs,
-        )
+            )
+        except ContextWindowExceededError as e:
+            raise ContextOverflowError(str(e)) from e
+        except LiteLLMBadRequestError as e:
+            if _is_context_overflow_error(e):
+                raise ContextOverflowError(str(e)) from e
+            raise
 
         # Yield chunks
         for chunk in response:
@@ -471,14 +550,20 @@ class ChatLiteLLMRouter(BaseChatModel):
         if self._tool_choice is not None:
             call_kwargs["tool_choice"] = self._tool_choice
 
-        # Call router async completion with streaming
+        try:
-        response = await self._router.acompletion(
+            response = await self._router.acompletion(
-            model=self.model,
+                model=self.model,
-            messages=formatted_messages,
+                messages=formatted_messages,
-            stop=stop,
+                stop=stop,
-            stream=True,
+                stream=True,
-            **call_kwargs,
+                **call_kwargs,
-        )
+            )
+        except ContextWindowExceededError as e:
+            raise ContextOverflowError(str(e)) from e
+        except LiteLLMBadRequestError as e:
+            if _is_context_overflow_error(e):
+                raise ContextOverflowError(str(e)) from e
+            raise
 
         # Yield chunks asynchronously
         async for chunk in response:

surfsense_backend/app/services/notion/kb_sync_service.py

@@ -4,11 +4,11 @@ from datetime import datetime
 from sqlalchemy import delete
 from sqlalchemy.ext.asyncio import AsyncSession
 
-from app.config import config
 from app.db import Chunk, Document
 from app.services.llm_service import get_user_long_context_llm
 from app.utils.document_converters import (
     create_document_chunks,
+    embed_text,
     generate_content_hash,
     generate_document_summary,
 )
@@ -127,10 +127,8 @@ class NotionKBSyncService:
                 logger.debug(f"Generated summary length: {len(summary_content)} chars")
             else:
                 logger.warning("No LLM configured - using fallback summary")
-                summary_content = f"Notion Page: {document.document_metadata.get('page_title')}\n\n{full_content[:500]}..."
+                summary_content = f"Notion Page: {document.document_metadata.get('page_title')}\n\n{full_content}"
-                summary_embedding = config.embedding_model_instance.embed(
+                summary_embedding = embed_text(summary_content)
-                    summary_content
-                )
 
             logger.debug(f"Deleting old chunks for document {document_id}")
             await self.db_session.execute(

surfsense_backend/app/services/task_dispatcher.py

@@ -0,0 +1,53 @@
+"""Task dispatcher abstraction for background document processing.
+
+Decouples the upload endpoint from Celery so tests can swap in a
+synchronous (inline) implementation that needs only PostgreSQL.
+"""
+
+from __future__ import annotations
+
+from typing import Protocol
+
+
+class TaskDispatcher(Protocol):
+    async def dispatch_file_processing(
+        self,
+        *,
+        document_id: int,
+        temp_path: str,
+        filename: str,
+        search_space_id: int,
+        user_id: str,
+        should_summarize: bool = False,
+    ) -> None: ...
+
+
+class CeleryTaskDispatcher:
+    """Production dispatcher — fires Celery tasks via Redis broker."""
+
+    async def dispatch_file_processing(
+        self,
+        *,
+        document_id: int,
+        temp_path: str,
+        filename: str,
+        search_space_id: int,
+        user_id: str,
+        should_summarize: bool = False,
+    ) -> None:
+        from app.tasks.celery_tasks.document_tasks import (
+            process_file_upload_with_document_task,
+        )
+
+        process_file_upload_with_document_task.delay(
+            document_id=document_id,
+            temp_path=temp_path,
+            filename=filename,
+            search_space_id=search_space_id,
+            user_id=user_id,
+            should_summarize=should_summarize,
+        )
+
+
+async def get_task_dispatcher() -> TaskDispatcher:
+    return CeleryTaskDispatcher()

surfsense_backend/app/tasks/celery_tasks/document_tasks.py

@@ -626,6 +626,7 @@ def process_file_upload_with_document_task(
     filename: str,
     search_space_id: int,
     user_id: str,
+    should_summarize: bool = False,
 ):
     """
     Celery task to process uploaded file with existing pending document.
@@ -640,6 +641,7 @@ def process_file_upload_with_document_task(
         filename: Original filename
         search_space_id: ID of the search space
         user_id: ID of the user
+        should_summarize: Whether to generate an LLM summary
     """
     import traceback
 
@@ -674,7 +676,12 @@ def process_file_upload_with_document_task(
     try:
         loop.run_until_complete(
             _process_file_with_document(
-                document_id, temp_path, filename, search_space_id, user_id
+                document_id,
+                temp_path,
+                filename,
+                search_space_id,
+                user_id,
+                should_summarize=should_summarize,
             )
         )
         logger.info(
@@ -710,6 +717,7 @@ async def _process_file_with_document(
     filename: str,
     search_space_id: int,
     user_id: str,
+    should_summarize: bool = False,
 ):
     """
     Process file and update existing pending document status.
@@ -811,6 +819,7 @@ async def _process_file_with_document(
                 task_logger=task_logger,
                 log_entry=log_entry,
                 notification=notification,
+                should_summarize=should_summarize,
             )
 
             # Update notification on success

surfsense_backend/app/tasks/chat/stream_new_chat.py

@@ -9,17 +9,21 @@ Supports loading LLM configurations from:
 - NewLLMConfig database table (positive IDs for user-created configs with prompt settings)
 """
 
+import asyncio
 import json
+import logging
+import re
+import time
 from collections.abc import AsyncGenerator
-from dataclasses import dataclass
+from dataclasses import dataclass, field
 from typing import Any
 from uuid import UUID
 
-import logging
-
 from langchain_core.messages import HumanMessage
+from sqlalchemy import func
 from sqlalchemy.ext.asyncio import AsyncSession
 from sqlalchemy.future import select
+from sqlalchemy.orm import selectinload
 
 from app.agents.new_chat.chat_deepagent import create_surfsense_deep_agent
 from app.agents.new_chat.checkpointer import get_checkpointer
@@ -30,7 +34,20 @@ from app.agents.new_chat.llm_config import (
     load_agent_config,
     load_llm_config_from_yaml,
 )
-from app.db import ChatVisibility, Document, Report, SurfsenseDocsDocument, async_session_maker
+from app.agents.new_chat.sandbox import (
+    get_or_create_sandbox,
+    is_sandbox_enabled,
+)
+from app.db import (
+    ChatVisibility,
+    Document,
+    NewChatMessage,
+    NewChatThread,
+    Report,
+    SearchSourceConnectorType,
+    SurfsenseDocsDocument,
+    async_session_maker,
+)
 from app.prompts import TITLE_GENERATION_PROMPT_TEMPLATE
 from app.services.chat_session_state_service import (
     clear_ai_responding,
@@ -40,6 +57,16 @@ from app.services.connector_service import ConnectorService
 from app.services.new_streaming_service import VercelStreamingService
 from app.utils.content_utils import bootstrap_history_from_db
 
+_perf_log = logging.getLogger("surfsense.perf")
+_perf_log.setLevel(logging.DEBUG)
+if not _perf_log.handlers:
+    _h = logging.StreamHandler()
+    _h.setFormatter(logging.Formatter("%(asctime)s [PERF] %(message)s"))
+    _perf_log.addHandler(_h)
+    _perf_log.propagate = False
+
+_background_tasks: set[asyncio.Task] = set()
+
 
 def format_mentioned_documents_as_context(documents: list[Document]) -> str:
     """
@@ -187,6 +214,7 @@ class StreamResult:
     accumulated_text: str = ""
     is_interrupted: bool = False
     interrupt_value: dict[str, Any] | None = None
+    sandbox_files: list[str] = field(default_factory=list)
 
 
 async def _stream_agent_events(
@@ -404,6 +432,21 @@ async def _stream_agent_events(
                     status="in_progress",
                     items=last_active_step_items,
                 )
+            elif tool_name == "execute":
+                cmd = (
+                    tool_input.get("command", "")
+                    if isinstance(tool_input, dict)
+                    else str(tool_input)
+                )
+                display_cmd = cmd[:80] + ("…" if len(cmd) > 80 else "")
+                last_active_step_title = "Running command"
+                last_active_step_items = [f"$ {display_cmd}"]
+                yield streaming_service.format_thinking_step(
+                    step_id=tool_step_id,
+                    title="Running command",
+                    status="in_progress",
+                    items=last_active_step_items,
+                )
             else:
                 last_active_step_title = f"Using {tool_name.replace('_', ' ')}"
                 last_active_step_items = []
@@ -620,6 +663,32 @@ async def _stream_agent_events(
                     status="completed",
                     items=completed_items,
                 )
+            elif tool_name == "execute":
+                raw_text = (
+                    tool_output.get("result", "")
+                    if isinstance(tool_output, dict)
+                    else str(tool_output)
+                )
+                m = re.match(r"^Exit code:\s*(\d+)", raw_text)
+                exit_code_val = int(m.group(1)) if m else None
+                if exit_code_val is not None and exit_code_val == 0:
+                    completed_items = [
+                        *last_active_step_items,
+                        "Completed successfully",
+                    ]
+                elif exit_code_val is not None:
+                    completed_items = [
+                        *last_active_step_items,
+                        f"Exit code: {exit_code_val}",
+                    ]
+                else:
+                    completed_items = [*last_active_step_items, "Finished"]
+                yield streaming_service.format_thinking_step(
+                    step_id=original_step_id,
+                    title="Running command",
+                    status="completed",
+                    items=completed_items,
+                )
             elif tool_name == "ls":
                 if isinstance(tool_output, dict):
                     ls_output = tool_output.get("result", "")
@@ -804,6 +873,8 @@ async def _stream_agent_events(
                 "create_linear_issue",
                 "update_linear_issue",
                 "delete_linear_issue",
+                "create_google_drive_file",
+                "delete_google_drive_file",
             ):
                 yield streaming_service.format_tool_output_available(
                     tool_call_id,
@@ -811,6 +882,36 @@ async def _stream_agent_events(
                     if isinstance(tool_output, dict)
                     else {"result": tool_output},
                 )
+            elif tool_name == "execute":
+                raw_text = (
+                    tool_output.get("result", "")
+                    if isinstance(tool_output, dict)
+                    else str(tool_output)
+                )
+                exit_code: int | None = None
+                output_text = raw_text
+                m = re.match(r"^Exit code:\s*(\d+)", raw_text)
+                if m:
+                    exit_code = int(m.group(1))
+                    om = re.search(r"\nOutput:\n([\s\S]*)", raw_text)
+                    output_text = om.group(1) if om else ""
+                thread_id_str = config.get("configurable", {}).get("thread_id", "")
+
+                for sf_match in re.finditer(
+                    r"^SANDBOX_FILE:\s*(.+)$", output_text, re.MULTILINE
+                ):
+                    fpath = sf_match.group(1).strip()
+                    if fpath and fpath not in result.sandbox_files:
+                        result.sandbox_files.append(fpath)
+
+                yield streaming_service.format_tool_output_available(
+                    tool_call_id,
+                    {
+                        "exit_code": exit_code,
+                        "output": output_text,
+                        "thread_id": thread_id_str,
+                    },
+                )
             else:
                 yield streaming_service.format_tool_output_available(
                     tool_call_id,
@@ -879,6 +980,38 @@ async def _stream_agent_events(
         yield streaming_service.format_interrupt_request(result.interrupt_value)
 
 
+def _try_persist_and_delete_sandbox(
+    thread_id: int,
+    sandbox_files: list[str],
+) -> None:
+    """Fire-and-forget: persist sandbox files locally then delete the sandbox."""
+    from app.agents.new_chat.sandbox import (
+        is_sandbox_enabled,
+        persist_and_delete_sandbox,
+    )
+
+    if not is_sandbox_enabled():
+        return
+
+    async def _run() -> None:
+        try:
+            await persist_and_delete_sandbox(thread_id, sandbox_files)
+        except Exception:
+            logging.getLogger(__name__).warning(
+                "persist_and_delete_sandbox failed for thread %s",
+                thread_id,
+                exc_info=True,
+            )
+
+    try:
+        loop = asyncio.get_running_loop()
+        task = loop.create_task(_run())
+        _background_tasks.add(task)
+        task.add_done_callback(_background_tasks.discard)
+    except RuntimeError:
+        pass
+
+
 async def stream_new_chat(
     user_query: str,
     search_space_id: int,
@@ -915,6 +1048,8 @@ async def stream_new_chat(
         str: SSE formatted response strings
     """
     streaming_service = VercelStreamingService()
+    stream_result = StreamResult()
+    _t_total = time.perf_counter()
 
     try:
         # Mark AI as responding to this user for live collaboration
@@ -923,6 +1058,7 @@ async def stream_new_chat(
         # Load LLM config - supports both YAML (negative IDs) and database (positive IDs)
         agent_config: AgentConfig | None = None
 
+        _t0 = time.perf_counter()
         if llm_config_id >= 0:
             # Positive ID: Load from NewLLMConfig database table
             agent_config = await load_agent_config(
@@ -953,6 +1089,11 @@ async def stream_new_chat(
             llm = create_chat_litellm_from_config(llm_config)
             # Create AgentConfig from YAML for consistency (uses defaults for prompt settings)
             agent_config = AgentConfig.from_yaml_config(llm_config)
+        _perf_log.info(
+            "[stream_new_chat] LLM config loaded in %.3fs (config_id=%s)",
+            time.perf_counter() - _t0,
+            llm_config_id,
+        )
 
         if not llm:
             yield streaming_service.format_error("Failed to create LLM instance")
@@ -960,22 +1101,45 @@ async def stream_new_chat(
             return
 
         # Create connector service
+        _t0 = time.perf_counter()
         connector_service = ConnectorService(session, search_space_id=search_space_id)
 
-        # Get Firecrawl API key from webcrawler connector if configured
-        from app.db import SearchSourceConnectorType
-
         firecrawl_api_key = None
         webcrawler_connector = await connector_service.get_connector_by_type(
             SearchSourceConnectorType.WEBCRAWLER_CONNECTOR, search_space_id
         )
         if webcrawler_connector and webcrawler_connector.config:
             firecrawl_api_key = webcrawler_connector.config.get("FIRECRAWL_API_KEY")
+        _perf_log.info(
+            "[stream_new_chat] Connector service + firecrawl key in %.3fs",
+            time.perf_counter() - _t0,
+        )
 
         # Get the PostgreSQL checkpointer for persistent conversation memory
+        _t0 = time.perf_counter()
         checkpointer = await get_checkpointer()
+        _perf_log.info(
+            "[stream_new_chat] Checkpointer ready in %.3fs", time.perf_counter() - _t0
+        )
+
+        sandbox_backend = None
+        _t0 = time.perf_counter()
+        if is_sandbox_enabled():
+            try:
+                sandbox_backend = await get_or_create_sandbox(chat_id)
+            except Exception as sandbox_err:
+                logging.getLogger(__name__).warning(
+                    "Sandbox creation failed, continuing without execute tool: %s",
+                    sandbox_err,
+                )
+        _perf_log.info(
+            "[stream_new_chat] Sandbox provisioning in %.3fs (enabled=%s)",
+            time.perf_counter() - _t0,
+            sandbox_backend is not None,
+        )
 
         visibility = thread_visibility or ChatVisibility.PRIVATE
+        _t0 = time.perf_counter()
         agent = await create_surfsense_deep_agent(
             llm=llm,
             search_space_id=search_space_id,
@@ -987,20 +1151,22 @@ async def stream_new_chat(
             agent_config=agent_config,
             firecrawl_api_key=firecrawl_api_key,
             thread_visibility=visibility,
+            sandbox_backend=sandbox_backend,
+        )
+        _perf_log.info(
+            "[stream_new_chat] Agent created in %.3fs", time.perf_counter() - _t0
         )
 
         # Build input with message history
         langchain_messages = []
 
+        _t0 = time.perf_counter()
         # Bootstrap history for cloned chats (no LangGraph checkpoint exists yet)
         if needs_history_bootstrap:
             langchain_messages = await bootstrap_history_from_db(
                 session, chat_id, thread_visibility=visibility
             )
 
-            # Clear the flag so we don't bootstrap again on next message
-            from app.db import NewChatThread
-
             thread_result = await session.execute(
                 select(NewChatThread).filter(NewChatThread.id == chat_id)
             )
@@ -1012,11 +1178,9 @@ async def stream_new_chat(
         # Fetch mentioned documents if any (with chunks for proper citations)
         mentioned_documents: list[Document] = []
         if mentioned_document_ids:
-            from sqlalchemy.orm import selectinload as doc_selectinload
-
             result = await session.execute(
                 select(Document)
-                .options(doc_selectinload(Document.chunks))
+                .options(selectinload(Document.chunks))
                 .filter(
                     Document.id.in_(mentioned_document_ids),
                     Document.search_space_id == search_space_id,
@@ -1027,8 +1191,6 @@ async def stream_new_chat(
         # Fetch mentioned SurfSense docs if any
         mentioned_surfsense_docs: list[SurfsenseDocsDocument] = []
         if mentioned_surfsense_doc_ids:
-            from sqlalchemy.orm import selectinload
-
             result = await session.execute(
                 select(SurfsenseDocsDocument)
                 .options(selectinload(SurfsenseDocsDocument.chunks))
@@ -1112,6 +1274,11 @@ async def stream_new_chat(
             "search_space_id": search_space_id,
         }
 
+        _perf_log.info(
+            "[stream_new_chat] History bootstrap + doc/report queries in %.3fs",
+            time.perf_counter() - _t0,
+        )
+
         # All pre-streaming DB reads are done.  Commit to release the
         # transaction and its ACCESS SHARE locks so we don't block DDL
         # (e.g. migrations) for the entire duration of LLM streaming.
@@ -1119,6 +1286,12 @@ async def stream_new_chat(
         # short-lived transactions (or use isolated sessions).
         await session.commit()
 
+        _perf_log.info(
+            "[stream_new_chat] Total pre-stream setup in %.3fs (chat_id=%s)",
+            time.perf_counter() - _t_total,
+            chat_id,
+        )
+
         # Configure LangGraph with thread_id for memory
         # If checkpoint_id is provided, fork from that checkpoint (for edit/reload)
         configurable = {"thread_id": str(chat_id)}
@@ -1180,7 +1353,8 @@ async def stream_new_chat(
             items=initial_items,
         )
 
-        stream_result = StreamResult()
+        _t_stream_start = time.perf_counter()
+        _first_event_logged = False
         async for sse in _stream_agent_events(
             agent=agent,
             config=config,
@@ -1192,8 +1366,23 @@ async def stream_new_chat(
             initial_step_title=initial_title,
             initial_step_items=initial_items,
         ):
+            if not _first_event_logged:
+                _perf_log.info(
+                    "[stream_new_chat] First agent event in %.3fs (time since stream start), "
+                    "%.3fs (total since request start) (chat_id=%s)",
+                    time.perf_counter() - _t_stream_start,
+                    time.perf_counter() - _t_total,
+                    chat_id,
+                )
+                _first_event_logged = True
             yield sse
 
+        _perf_log.info(
+            "[stream_new_chat] Agent stream completed in %.3fs (chat_id=%s)",
+            time.perf_counter() - _t_stream_start,
+            chat_id,
+        )
+
         if stream_result.is_interrupted:
             yield streaming_service.format_finish_step()
             yield streaming_service.format_finish()
@@ -1202,12 +1391,6 @@ async def stream_new_chat(
 
         accumulated_text = stream_result.accumulated_text
 
-        # Generate LLM title for new chats after first response
-        # Check if this is the first assistant response by counting existing assistant messages
-        from sqlalchemy import func
-
-        from app.db import NewChatMessage, NewChatThread
-
         assistant_count_result = await session.execute(
             select(func.count(NewChatMessage.id)).filter(
                 NewChatMessage.thread_id == chat_id,
@@ -1294,6 +1477,8 @@ async def stream_new_chat(
                     "Failed to clear AI responding state for thread %s", chat_id
                 )
 
+        _try_persist_and_delete_sandbox(chat_id, stream_result.sandbox_files)
+
 
 async def stream_resume_chat(
     chat_id: int,
@@ -1305,12 +1490,15 @@ async def stream_resume_chat(
     thread_visibility: ChatVisibility | None = None,
 ) -> AsyncGenerator[str, None]:
     streaming_service = VercelStreamingService()
+    stream_result = StreamResult()
+    _t_total = time.perf_counter()
 
     try:
         if user_id:
             await set_ai_responding(session, chat_id, UUID(user_id))
 
         agent_config: AgentConfig | None = None
+        _t0 = time.perf_counter()
         if llm_config_id >= 0:
             agent_config = await load_agent_config(
                 session=session,
@@ -1334,26 +1522,54 @@ async def stream_resume_chat(
                 return
             llm = create_chat_litellm_from_config(llm_config)
             agent_config = AgentConfig.from_yaml_config(llm_config)
+        _perf_log.info(
+            "[stream_resume] LLM config loaded in %.3fs", time.perf_counter() - _t0
+        )
 
         if not llm:
             yield streaming_service.format_error("Failed to create LLM instance")
             yield streaming_service.format_done()
             return
 
+        _t0 = time.perf_counter()
         connector_service = ConnectorService(session, search_space_id=search_space_id)
 
-        from app.db import SearchSourceConnectorType
-
         firecrawl_api_key = None
         webcrawler_connector = await connector_service.get_connector_by_type(
             SearchSourceConnectorType.WEBCRAWLER_CONNECTOR, search_space_id
         )
         if webcrawler_connector and webcrawler_connector.config:
             firecrawl_api_key = webcrawler_connector.config.get("FIRECRAWL_API_KEY")
+        _perf_log.info(
+            "[stream_resume] Connector service + firecrawl key in %.3fs",
+            time.perf_counter() - _t0,
+        )
 
+        _t0 = time.perf_counter()
         checkpointer = await get_checkpointer()
+        _perf_log.info(
+            "[stream_resume] Checkpointer ready in %.3fs", time.perf_counter() - _t0
+        )
+
+        sandbox_backend = None
+        _t0 = time.perf_counter()
+        if is_sandbox_enabled():
+            try:
+                sandbox_backend = await get_or_create_sandbox(chat_id)
+            except Exception as sandbox_err:
+                logging.getLogger(__name__).warning(
+                    "Sandbox creation failed, continuing without execute tool: %s",
+                    sandbox_err,
+                )
+        _perf_log.info(
+            "[stream_resume] Sandbox provisioning in %.3fs (enabled=%s)",
+            time.perf_counter() - _t0,
+            sandbox_backend is not None,
+        )
+
         visibility = thread_visibility or ChatVisibility.PRIVATE
 
+        _t0 = time.perf_counter()
         agent = await create_surfsense_deep_agent(
             llm=llm,
             search_space_id=search_space_id,
@@ -1365,11 +1581,21 @@ async def stream_resume_chat(
             agent_config=agent_config,
             firecrawl_api_key=firecrawl_api_key,
             thread_visibility=visibility,
+            sandbox_backend=sandbox_backend,
+        )
+        _perf_log.info(
+            "[stream_resume] Agent created in %.3fs", time.perf_counter() - _t0
         )
 
         # Release the transaction before streaming (same rationale as stream_new_chat).
         await session.commit()
 
+        _perf_log.info(
+            "[stream_resume] Total pre-stream setup in %.3fs (chat_id=%s)",
+            time.perf_counter() - _t_total,
+            chat_id,
+        )
+
         from langgraph.types import Command
 
         config = {
@@ -1380,7 +1606,8 @@ async def stream_resume_chat(
         yield streaming_service.format_message_start()
         yield streaming_service.format_start_step()
 
-        stream_result = StreamResult()
+        _t_stream_start = time.perf_counter()
+        _first_event_logged = False
         async for sse in _stream_agent_events(
             agent=agent,
             config=config,
@@ -1389,7 +1616,20 @@ async def stream_resume_chat(
             result=stream_result,
             step_prefix="thinking-resume",
         ):
+            if not _first_event_logged:
+                _perf_log.info(
+                    "[stream_resume] First agent event in %.3fs (stream), %.3fs (total) (chat_id=%s)",
+                    time.perf_counter() - _t_stream_start,
+                    time.perf_counter() - _t_total,
+                    chat_id,
+                )
+                _first_event_logged = True
             yield sse
+        _perf_log.info(
+            "[stream_resume] Agent stream completed in %.3fs (chat_id=%s)",
+            time.perf_counter() - _t_stream_start,
+            chat_id,
+        )
         if stream_result.is_interrupted:
             yield streaming_service.format_finish_step()
             yield streaming_service.format_finish()
@@ -1423,3 +1663,5 @@ async def stream_resume_chat(
                 logging.getLogger(__name__).warning(
                     "Failed to clear AI responding state for thread %s", chat_id
                 )
+
+        _try_persist_and_delete_sandbox(chat_id, stream_result.sandbox_files)

surfsense_backend/app/tasks/connector_indexers/airtable_indexer.py

@@ -12,13 +12,13 @@ from collections.abc import Awaitable, Callable
 from sqlalchemy.exc import SQLAlchemyError
 from sqlalchemy.ext.asyncio import AsyncSession
 
-from app.config import config
 from app.connectors.airtable_history import AirtableHistoryConnector
 from app.db import Document, DocumentStatus, DocumentType, SearchSourceConnectorType
 from app.services.llm_service import get_user_long_context_llm
 from app.services.task_logging_service import TaskLoggingService
 from app.utils.document_converters import (
     create_document_chunks,
+    embed_text,
     generate_content_hash,
     generate_document_summary,
     generate_unique_identifier_hash,
@@ -399,7 +399,7 @@ async def index_airtable_records(
                         session, user_id, search_space_id
                     )
 
-                    if user_llm:
+                    if user_llm and connector.enable_summary:
                         document_metadata_for_summary = {
                             "record_id": item["record_id"],
                             "created_time": item["record"].get("CREATED_TIME()", ""),
@@ -415,11 +415,8 @@ async def index_airtable_records(
                             document_metadata_for_summary,
                         )
                     else:
-                        # Fallback to simple summary if no LLM configured
+                        summary_content = f"Airtable Record: {item['record_id']}\n\n{item['markdown_content']}"
-                        summary_content = f"Airtable Record: {item['record_id']}\n\n"
+                        summary_embedding = embed_text(summary_content)
-                        summary_embedding = config.embedding_model_instance.embed(
-                            summary_content
-                        )
 
                     chunks = await create_document_chunks(item["markdown_content"])
 

surfsense_backend/app/tasks/connector_indexers/bookstack_indexer.py

@@ -13,13 +13,13 @@ from datetime import datetime
 from sqlalchemy.exc import SQLAlchemyError
 from sqlalchemy.ext.asyncio import AsyncSession
 
-from app.config import config
 from app.connectors.bookstack_connector import BookStackConnector
 from app.db import Document, DocumentStatus, DocumentType, SearchSourceConnectorType
 from app.services.llm_service import get_user_long_context_llm
 from app.services.task_logging_service import TaskLoggingService
 from app.utils.document_converters import (
     create_document_chunks,
+    embed_text,
     generate_content_hash,
     generate_document_summary,
     generate_unique_identifier_hash,
@@ -403,7 +403,7 @@ async def index_bookstack_pages(
                     "connector_id": connector_id,
                 }
 
-                if user_llm:
+                if user_llm and connector.enable_summary:
                     summary_metadata = {
                         "page_name": item["page_name"],
                         "page_id": item["page_id"],
@@ -418,17 +418,8 @@ async def index_bookstack_pages(
                         item["full_content"], user_llm, summary_metadata
                     )
                 else:
-                    # Fallback to simple summary if no LLM configured
+                    summary_content = f"BookStack Page: {item['page_name']}\n\nBook ID: {item['book_id']}\n\n{item['full_content']}"
-                    summary_content = f"BookStack Page: {item['page_name']}\n\nBook ID: {item['book_id']}\n\n"
+                    summary_embedding = embed_text(summary_content)
-                    if item["page_content"]:
-                        # Take first 1000 characters of content for summary
-                        content_preview = item["page_content"][:1000]
-                        if len(item["page_content"]) > 1000:
-                            content_preview += "..."
-                        summary_content += f"Content Preview: {content_preview}\n\n"
-                    summary_embedding = config.embedding_model_instance.embed(
-                        summary_content
-                    )
 
                 # Process chunks - using the full page content
                 chunks = await create_document_chunks(item["full_content"])

surfsense_backend/app/tasks/connector_indexers/clickup_indexer.py

@@ -14,13 +14,13 @@ from datetime import datetime
 from sqlalchemy.exc import SQLAlchemyError
 from sqlalchemy.ext.asyncio import AsyncSession
 
-from app.config import config
 from app.connectors.clickup_history import ClickUpHistoryConnector
 from app.db import Document, DocumentStatus, DocumentType, SearchSourceConnectorType
 from app.services.llm_service import get_user_long_context_llm
 from app.services.task_logging_service import TaskLoggingService
 from app.utils.document_converters import (
     create_document_chunks,
+    embed_text,
     generate_content_hash,
     generate_document_summary,
     generate_unique_identifier_hash,
@@ -398,7 +398,7 @@ async def index_clickup_tasks(
                     session, user_id, search_space_id
                 )
 
-                if user_llm:
+                if user_llm and connector.enable_summary:
                     document_metadata_for_summary = {
                         "task_id": item["task_id"],
                         "task_name": item["task_name"],
@@ -418,9 +418,7 @@ async def index_clickup_tasks(
                     )
                 else:
                     summary_content = item["task_content"]
-                    summary_embedding = config.embedding_model_instance.embed(
+                    summary_embedding = embed_text(item["task_content"])
-                        item["task_content"]
-                    )
 
                 chunks = await create_document_chunks(item["task_content"])
 

surfsense_backend/app/tasks/connector_indexers/confluence_indexer.py

@@ -14,13 +14,13 @@ from datetime import datetime
 from sqlalchemy.exc import SQLAlchemyError
 from sqlalchemy.ext.asyncio import AsyncSession
 
-from app.config import config
 from app.connectors.confluence_history import ConfluenceHistoryConnector
 from app.db import Document, DocumentStatus, DocumentType, SearchSourceConnectorType
 from app.services.llm_service import get_user_long_context_llm
 from app.services.task_logging_service import TaskLoggingService
 from app.utils.document_converters import (
     create_document_chunks,
+    embed_text,
     generate_content_hash,
     generate_document_summary,
     generate_unique_identifier_hash,
@@ -378,7 +378,7 @@ async def index_confluence_pages(
                     session, user_id, search_space_id
                 )
 
-                if user_llm:
+                if user_llm and connector.enable_summary:
                     document_metadata = {
                         "page_title": item["page_title"],
                         "page_id": item["page_id"],
@@ -394,18 +394,8 @@ async def index_confluence_pages(
                         item["full_content"], user_llm, document_metadata
                     )
                 else:
-                    # Fallback to simple summary if no LLM configured
+                    summary_content = f"Confluence Page: {item['page_title']}\n\nSpace ID: {item['space_id']}\n\n{item['full_content']}"
-                    summary_content = f"Confluence Page: {item['page_title']}\n\nSpace ID: {item['space_id']}\n\n"
+                    summary_embedding = embed_text(summary_content)
-                    if item["page_content"]:
-                        # Take first 1000 characters of content for summary
-                        content_preview = item["page_content"][:1000]
-                        if len(item["page_content"]) > 1000:
-                            content_preview += "..."
-                        summary_content += f"Content Preview: {content_preview}\n\n"
-                    summary_content += f"Comments: {item['comment_count']}"
-                    summary_embedding = config.embedding_model_instance.embed(
-                        summary_content
-                    )
 
                 # Process chunks - using the full page content with comments
                 chunks = await create_document_chunks(item["full_content"])

surfsense_backend/app/tasks/connector_indexers/discord_indexer.py

@@ -23,6 +23,7 @@ from app.db import Document, DocumentStatus, DocumentType, SearchSourceConnector
 from app.services.task_logging_service import TaskLoggingService
 from app.utils.document_converters import (
     create_document_chunks,
+    embed_text,
     generate_content_hash,
     generate_unique_identifier_hash,
 )
@@ -669,9 +670,7 @@ async def index_discord_messages(
 
                 # Heavy processing (embeddings, chunks)
                 chunks = await create_document_chunks(item["combined_document_string"])
-                doc_embedding = config.embedding_model_instance.embed(
+                doc_embedding = embed_text(item["combined_document_string"])
-                    item["combined_document_string"]
-                )
 
                 # Update document to READY with actual content
                 document.title = f"{item['guild_name']}#{item['channel_name']}"

surfsense_backend/app/tasks/connector_indexers/github_indexer.py

@@ -16,13 +16,13 @@ from datetime import UTC, datetime
 from sqlalchemy.exc import SQLAlchemyError
 from sqlalchemy.ext.asyncio import AsyncSession
 
-from app.config import config
 from app.connectors.github_connector import GitHubConnector
 from app.db import Document, DocumentStatus, DocumentType, SearchSourceConnectorType
 from app.services.llm_service import get_user_long_context_llm
 from app.services.task_logging_service import TaskLoggingService
 from app.utils.document_converters import (
     create_document_chunks,
+    embed_text,
     generate_content_hash,
     generate_document_summary,
     generate_unique_identifier_hash,
@@ -367,7 +367,7 @@ async def index_github_repos(
                     "estimated_tokens": digest.estimated_tokens,
                 }
 
-                if user_llm:
+                if user_llm and connector.enable_summary:
                     # Prepare content for summarization
                     summary_content = digest.full_digest
                     if len(summary_content) > MAX_DIGEST_CHARS:
@@ -381,15 +381,12 @@ async def index_github_repos(
                         summary_content, user_llm, document_metadata_for_summary
                     )
                 else:
-                    # Fallback to simple summary if no LLM configured
                     summary_text = (
                         f"# GitHub Repository: {repo_full_name}\n\n"
                         f"## Summary\n{digest.summary}\n\n"
-                        f"## File Structure\n{digest.tree[:3000]}"
+                        f"## File Structure\n{digest.tree}"
-                    )
-                    summary_embedding = config.embedding_model_instance.embed(
-                        summary_text
                     )
+                    summary_embedding = embed_text(summary_text)
 
                 # Chunk the full digest content for granular search
                 try:
@@ -551,7 +548,7 @@ async def _simple_chunk_content(content: str, chunk_size: int = 4000) -> list:
             chunks.append(
                 Chunk(
                     content=chunk_text,
-                    embedding=config.embedding_model_instance.embed(chunk_text),
+                    embedding=embed_text(chunk_text),
                 )
             )
 

surfsense_backend/app/tasks/connector_indexers/google_calendar_indexer.py

@@ -20,6 +20,7 @@ from app.services.llm_service import get_user_long_context_llm
 from app.services.task_logging_service import TaskLoggingService
 from app.utils.document_converters import (
     create_document_chunks,
+    embed_text,
     generate_content_hash,
     generate_document_summary,
     generate_unique_identifier_hash,
@@ -489,7 +490,7 @@ async def index_google_calendar_events(
                     session, user_id, search_space_id
                 )
 
-                if user_llm:
+                if user_llm and connector.enable_summary:
                     document_metadata_for_summary = {
                         "event_id": item["event_id"],
                         "event_summary": item["event_summary"],
@@ -507,22 +508,8 @@ async def index_google_calendar_events(
                         item["event_markdown"], user_llm, document_metadata_for_summary
                     )
                 else:
-                    summary_content = (
+                    summary_content = f"Google Calendar Event: {item['event_summary']}\n\n{item['event_markdown']}"
-                        f"Google Calendar Event: {item['event_summary']}\n\n"
+                    summary_embedding = embed_text(summary_content)
-                    )
-                    summary_content += f"Calendar: {item['calendar_id']}\n"
-                    summary_content += f"Start: {item['start_time']}\n"
-                    summary_content += f"End: {item['end_time']}\n"
-                    if item["location"]:
-                        summary_content += f"Location: {item['location']}\n"
-                    if item["description"]:
-                        desc_preview = item["description"][:1000]
-                        if len(item["description"]) > 1000:
-                            desc_preview += "..."
-                        summary_content += f"Description: {desc_preview}\n"
-                    summary_embedding = config.embedding_model_instance.embed(
-                        summary_content
-                    )
 
                 chunks = await create_document_chunks(item["event_markdown"])
 

surfsense_backend/app/tasks/connector_indexers/google_drive_indexer.py

@@ -352,7 +352,7 @@ async def index_google_drive_single_file(
             await session.commit()
 
         # Process the file
-        indexed, skipped, failed = await _process_single_file(
+        indexed, _skipped, failed = await _process_single_file(
             drive_client=drive_client,
             session=session,
             file=file,
@@ -608,7 +608,7 @@ async def _index_with_delta_sync(
         {"stage": "delta_sync", "start_token": start_page_token},
     )
 
-    changes, final_token, error = await fetch_all_changes(
+    changes, _final_token, error = await fetch_all_changes(
         drive_client, start_page_token, folder_id
     )
 
@@ -1011,7 +1011,7 @@ async def _process_single_file(
             pending_document.status = DocumentStatus.processing()
             await session.commit()
 
-        _, error, metadata = await download_and_process_file(
+        _, error, _metadata = await download_and_process_file(
             client=drive_client,
             file=file,
             search_space_id=search_space_id,

surfsense_backend/app/tasks/connector_indexers/google_gmail_indexer.py

@@ -25,6 +25,7 @@ from app.services.llm_service import get_user_long_context_llm
 from app.services.task_logging_service import TaskLoggingService
 from app.utils.document_converters import (
     create_document_chunks,
+    embed_text,
     generate_content_hash,
     generate_document_summary,
     generate_unique_identifier_hash,
@@ -413,7 +414,7 @@ async def index_google_gmail_messages(
                     session, user_id, search_space_id
                 )
 
-                if user_llm:
+                if user_llm and connector.enable_summary:
                     document_metadata_for_summary = {
                         "message_id": item["message_id"],
                         "thread_id": item["thread_id"],
@@ -432,12 +433,8 @@ async def index_google_gmail_messages(
                         document_metadata_for_summary,
                     )
                 else:
-                    summary_content = f"Google Gmail Message: {item['subject']}\n\n"
+                    summary_content = f"Google Gmail Message: {item['subject']}\n\nFrom: {item['sender']}\nDate: {item['date_str']}\n\n{item['markdown_content']}"
-                    summary_content += f"Sender: {item['sender']}\n"
+                    summary_embedding = embed_text(summary_content)
-                    summary_content += f"Date: {item['date_str']}\n"
-                    summary_embedding = config.embedding_model_instance.embed(
-                        summary_content
-                    )
 
                 chunks = await create_document_chunks(item["markdown_content"])
 

surfsense_backend/app/tasks/connector_indexers/jira_indexer.py

@@ -14,13 +14,13 @@ from datetime import datetime
 from sqlalchemy.exc import SQLAlchemyError
 from sqlalchemy.ext.asyncio import AsyncSession
 
-from app.config import config
 from app.connectors.jira_history import JiraHistoryConnector
 from app.db import Document, DocumentStatus, DocumentType, SearchSourceConnectorType
 from app.services.llm_service import get_user_long_context_llm
 from app.services.task_logging_service import TaskLoggingService
 from app.utils.document_converters import (
     create_document_chunks,
+    embed_text,
     generate_content_hash,
     generate_document_summary,
     generate_unique_identifier_hash,
@@ -356,7 +356,7 @@ async def index_jira_issues(
                     session, user_id, search_space_id
                 )
 
-                if user_llm:
+                if user_llm and connector.enable_summary:
                     document_metadata = {
                         "issue_key": item["issue_identifier"],
                         "issue_title": item["issue_title"],
@@ -373,14 +373,8 @@ async def index_jira_issues(
                         item["issue_content"], user_llm, document_metadata
                     )
                 else:
-                    # Fallback to simple summary if no LLM configured
+                    summary_content = f"Jira Issue {item['issue_identifier']}: {item['issue_title']}\n\n{item['issue_content']}"
-                    summary_content = f"Jira Issue {item['issue_identifier']}: {item['issue_title']}\n\nStatus: {item['formatted_issue'].get('status', 'Unknown')}\n\n"
+                    summary_embedding = embed_text(summary_content)
-                    if item["formatted_issue"].get("description"):
-                        summary_content += f"Description: {item['formatted_issue'].get('description')}\n\n"
-                    summary_content += f"Comments: {item['comment_count']}"
-                    summary_embedding = config.embedding_model_instance.embed(
-                        summary_content
-                    )
 
                 # Process chunks - using the full issue content with comments
                 chunks = await create_document_chunks(item["issue_content"])

surfsense_backend/app/tasks/connector_indexers/linear_indexer.py

@@ -13,13 +13,13 @@ from datetime import datetime
 from sqlalchemy.exc import SQLAlchemyError
 from sqlalchemy.ext.asyncio import AsyncSession
 
-from app.config import config
 from app.connectors.linear_connector import LinearConnector
 from app.db import Document, DocumentStatus, DocumentType, SearchSourceConnectorType
 from app.services.llm_service import get_user_long_context_llm
 from app.services.task_logging_service import TaskLoggingService
 from app.utils.document_converters import (
     create_document_chunks,
+    embed_text,
     generate_content_hash,
     generate_document_summary,
     generate_unique_identifier_hash,
@@ -395,7 +395,7 @@ async def index_linear_issues(
                     session, user_id, search_space_id
                 )
 
-                if user_llm:
+                if user_llm and connector.enable_summary:
                     document_metadata_for_summary = {
                         "issue_id": item["issue_identifier"],
                         "issue_title": item["issue_title"],
@@ -412,17 +412,8 @@ async def index_linear_issues(
                         item["issue_content"], user_llm, document_metadata_for_summary
                     )
                 else:
-                    # Fallback to simple summary if no LLM configured
+                    summary_content = f"Linear Issue {item['issue_identifier']}: {item['issue_title']}\n\nStatus: {item['state']}\n\n{item['issue_content']}"
-                    description = item["description"]
+                    summary_embedding = embed_text(summary_content)
-                    if description and len(description) > 1000:
-                        description = description[:997] + "..."
-                    summary_content = f"Linear Issue {item['issue_identifier']}: {item['issue_title']}\n\nStatus: {item['state']}\n\n"
-                    if description:
-                        summary_content += f"Description: {description}\n\n"
-                    summary_content += f"Comments: {item['comment_count']}"
-                    summary_embedding = config.embedding_model_instance.embed(
-                        summary_content
-                    )
 
                 chunks = await create_document_chunks(item["issue_content"])
 

surfsense_backend/app/tasks/connector_indexers/luma_indexer.py

@@ -13,13 +13,13 @@ from datetime import datetime, timedelta
 from sqlalchemy.exc import SQLAlchemyError
 from sqlalchemy.ext.asyncio import AsyncSession
 
-from app.config import config
 from app.connectors.luma_connector import LumaConnector
 from app.db import Document, DocumentStatus, DocumentType, SearchSourceConnectorType
 from app.services.llm_service import get_user_long_context_llm
 from app.services.task_logging_service import TaskLoggingService
 from app.utils.document_converters import (
     create_document_chunks,
+    embed_text,
     generate_content_hash,
     generate_document_summary,
     generate_unique_identifier_hash,
@@ -441,7 +441,7 @@ async def index_luma_events(
                     session, user_id, search_space_id
                 )
 
-                if user_llm:
+                if user_llm and connector.enable_summary:
                     document_metadata_for_summary = {
                         "event_id": item["event_id"],
                         "event_name": item["event_name"],
@@ -462,29 +462,10 @@ async def index_luma_events(
                         item["event_markdown"], user_llm, document_metadata_for_summary
                     )
                 else:
-                    # Fallback to simple summary if no LLM configured
+                    summary_content = (
-                    summary_content = f"Luma Event: {item['event_name']}\n\n"
+                        f"Luma Event: {item['event_name']}\n\n{item['event_markdown']}"
-                    if item["event_url"]:
-                        summary_content += f"URL: {item['event_url']}\n"
-                    summary_content += f"Start: {item['start_at']}\n"
-                    summary_content += f"End: {item['end_at']}\n"
-                    if item["timezone"]:
-                        summary_content += f"Timezone: {item['timezone']}\n"
-                    if item["location"]:
-                        summary_content += f"Location: {item['location']}\n"
-                    if item["city"]:
-                        summary_content += f"City: {item['city']}\n"
-                    if item["host_names"]:
-                        summary_content += f"Hosts: {item['host_names']}\n"
-                    if item["description"]:
-                        desc_preview = item["description"][:1000]
-                        if len(item["description"]) > 1000:
-                            desc_preview += "..."
-                        summary_content += f"Description: {desc_preview}\n"
-
-                    summary_embedding = config.embedding_model_instance.embed(
-                        summary_content
                     )
+                    summary_embedding = embed_text(summary_content)
 
                 chunks = await create_document_chunks(item["event_markdown"])
 

surfsense_backend/app/tasks/connector_indexers/notion_indexer.py

@@ -13,13 +13,13 @@ from datetime import datetime
 from sqlalchemy.exc import SQLAlchemyError
 from sqlalchemy.ext.asyncio import AsyncSession
 
-from app.config import config
 from app.connectors.notion_history import NotionHistoryConnector
 from app.db import Document, DocumentStatus, DocumentType, SearchSourceConnectorType
 from app.services.llm_service import get_user_long_context_llm
 from app.services.task_logging_service import TaskLoggingService
 from app.utils.document_converters import (
     create_document_chunks,
+    embed_text,
     generate_content_hash,
     generate_document_summary,
     generate_unique_identifier_hash,
@@ -447,7 +447,7 @@ async def index_notion_pages(
                     session, user_id, search_space_id
                 )
 
-                if user_llm:
+                if user_llm and connector.enable_summary:
                     document_metadata_for_summary = {
                         "page_title": item["page_title"],
                         "page_id": item["page_id"],
@@ -463,11 +463,8 @@ async def index_notion_pages(
                         document_metadata_for_summary,
                     )
                 else:
-                    # Fallback to simple summary if no LLM configured
+                    summary_content = f"Notion Page: {item['page_title']}\n\n{item['markdown_content']}"
-                    summary_content = f"Notion Page: {item['page_title']}\n\n{item['markdown_content'][:500]}..."
+                    summary_embedding = embed_text(summary_content)
-                    summary_embedding = config.embedding_model_instance.embed(
-                        summary_content
-                    )
 
                 chunks = await create_document_chunks(item["markdown_content"])
 

surfsense_backend/app/tasks/connector_indexers/obsidian_indexer.py

@@ -26,6 +26,7 @@ from app.services.llm_service import get_user_long_context_llm
 from app.services.task_logging_service import TaskLoggingService
 from app.utils.document_converters import (
     create_document_chunks,
+    embed_text,
     generate_content_hash,
     generate_document_summary,
     generate_unique_identifier_hash,
@@ -546,7 +547,7 @@ async def index_obsidian_vault(
 
                 # Generate summary
                 summary_content = ""
-                if long_context_llm:
+                if long_context_llm and connector.enable_summary:
                     summary_content, _ = await generate_document_summary(
                         document_string,
                         long_context_llm,
@@ -554,7 +555,7 @@ async def index_obsidian_vault(
                     )
 
                 # Generate embedding
-                embedding = config.embedding_model_instance.embed(document_string)
+                embedding = embed_text(document_string)
 
                 # Add URL and summary to metadata
                 document_metadata["url"] = f"obsidian://{vault_name}/{relative_path}"

surfsense_backend/app/tasks/connector_indexers/slack_indexer.py

@@ -17,12 +17,12 @@ from slack_sdk.errors import SlackApiError
 from sqlalchemy.exc import SQLAlchemyError
 from sqlalchemy.ext.asyncio import AsyncSession
 
-from app.config import config
 from app.connectors.slack_history import SlackHistory
 from app.db import Document, DocumentStatus, DocumentType, SearchSourceConnectorType
 from app.services.task_logging_service import TaskLoggingService
 from app.utils.document_converters import (
     create_document_chunks,
+    embed_text,
     generate_content_hash,
     generate_unique_identifier_hash,
 )
@@ -542,9 +542,7 @@ async def index_slack_messages(
 
                 # Heavy processing (embeddings, chunks)
                 chunks = await create_document_chunks(item["combined_document_string"])
-                doc_embedding = config.embedding_model_instance.embed(
+                doc_embedding = embed_text(item["combined_document_string"])
-                    item["combined_document_string"]
-                )
 
                 # Update document to READY with actual content
                 document.title = f"{item['team_name']}#{item['channel_name']}"

surfsense_backend/app/tasks/connector_indexers/teams_indexer.py

@@ -16,12 +16,12 @@ from datetime import UTC, datetime
 from sqlalchemy.exc import SQLAlchemyError
 from sqlalchemy.ext.asyncio import AsyncSession
 
-from app.config import config
 from app.connectors.teams_history import TeamsHistory
 from app.db import Document, DocumentStatus, DocumentType, SearchSourceConnectorType
 from app.services.task_logging_service import TaskLoggingService
 from app.utils.document_converters import (
     create_document_chunks,
+    embed_text,
     generate_content_hash,
     generate_unique_identifier_hash,
 )
@@ -581,9 +581,7 @@ async def index_teams_messages(
 
                 # Heavy processing (embeddings, chunks)
                 chunks = await create_document_chunks(item["combined_document_string"])
-                doc_embedding = config.embedding_model_instance.embed(
+                doc_embedding = embed_text(item["combined_document_string"])
-                    item["combined_document_string"]
-                )
 
                 # Update document to READY with actual content
                 document.title = f"{item['team_name']} - {item['channel_name']}"

surfsense_backend/app/tasks/connector_indexers/webcrawler_indexer.py

@@ -13,13 +13,13 @@ from datetime import datetime
 from sqlalchemy.exc import SQLAlchemyError
 from sqlalchemy.ext.asyncio import AsyncSession
 
-from app.config import config
 from app.connectors.webcrawler_connector import WebCrawlerConnector
 from app.db import Document, DocumentStatus, DocumentType, SearchSourceConnectorType
 from app.services.llm_service import get_user_long_context_llm
 from app.services.task_logging_service import TaskLoggingService
 from app.utils.document_converters import (
     create_document_chunks,
+    embed_text,
     generate_content_hash,
     generate_document_summary,
     generate_unique_identifier_hash,
@@ -377,7 +377,7 @@ async def index_crawled_urls(
                     session, user_id, search_space_id
                 )
 
-                if user_llm:
+                if user_llm and connector.enable_summary:
                     document_metadata_for_summary = {
                         "url": url,
                         "title": title,
@@ -393,24 +393,8 @@ async def index_crawled_urls(
                         structured_document, user_llm, document_metadata_for_summary
                     )
                 else:
-                    # Fallback to simple summary if no LLM configured
+                    summary_content = f"Crawled URL: {title}\n\nURL: {url}\n\n{content}"
-                    summary_content = f"Crawled URL: {title}\n\n"
+                    summary_embedding = embed_text(summary_content)
-                    summary_content += f"URL: {url}\n"
-                    if description:
-                        summary_content += f"Description: {description}\n"
-                    if language:
-                        summary_content += f"Language: {language}\n"
-                    summary_content += f"Crawler: {crawler_type}\n\n"
-
-                    # Add content preview
-                    content_preview = content[:1000]
-                    if len(content) > 1000:
-                        content_preview += "..."
-                    summary_content += f"Content Preview:\n{content_preview}\n"
-
-                    summary_embedding = config.embedding_model_instance.embed(
-                        summary_content
-                    )
 
                 # Process chunks
                 chunks = await create_document_chunks(content)

surfsense_backend/app/tasks/document_processors/file_processors.py

@@ -18,12 +18,14 @@ from sqlalchemy.ext.asyncio import AsyncSession
 
 from app.config import config as app_config
 from app.db import Document, DocumentStatus, DocumentType, Log, Notification
+from app.indexing_pipeline.adapters.file_upload_adapter import index_uploaded_file
 from app.services.llm_service import get_user_long_context_llm
 from app.services.notification_service import NotificationService
 from app.services.task_logging_service import TaskLoggingService
 from app.utils.document_converters import (
     convert_document_to_markdown,
     create_document_chunks,
+    embed_text,
     generate_content_hash,
     generate_document_summary,
     generate_unique_identifier_hash,
@@ -33,7 +35,6 @@ from .base import (
     check_document_by_unique_identifier,
     check_duplicate_document,
     get_current_timestamp,
-    safe_set_chunks,
 )
 from .markdown_processor import add_received_markdown_file_document
 
@@ -760,11 +761,7 @@ async def add_received_file_document_using_docling(
             f"{metadata_section}\n\n# DOCUMENT SUMMARY\n\n{summary_content}"
         )
 
-        from app.config import config
+        summary_embedding = embed_text(enhanced_summary_content)
-
-        summary_embedding = config.embedding_model_instance.embed(
-            enhanced_summary_content
-        )
 
         # Process chunks
         chunks = await create_document_chunks(file_in_markdown)
@@ -1599,6 +1596,7 @@ async def process_file_in_background_with_document(
     log_entry: Log,
     connector: dict | None = None,
     notification: Notification | None = None,
+    should_summarize: bool = False,
 ) -> Document | None:
     """
     Process file and update existing pending document (2-phase pattern).
@@ -1632,6 +1630,8 @@ async def process_file_in_background_with_document(
     from app.config import config as app_config
     from app.services.llm_service import get_user_long_context_llm
 
+    doc_id = document.id
+
     try:
         markdown_content = None
         etl_service = None
@@ -1855,7 +1855,7 @@ async def process_file_in_background_with_document(
         content_hash = generate_content_hash(markdown_content, search_space_id)
 
         existing_by_content = await check_duplicate_document(session, content_hash)
-        if existing_by_content and existing_by_content.id != document.id:
+        if existing_by_content and existing_by_content.id != doc_id:
             # Duplicate content found - mark this document as failed
             logging.info(
                 f"Duplicate content detected for {filename}, "
@@ -1863,7 +1863,7 @@ async def process_file_in_background_with_document(
             )
             return None
 
-        # ===== STEP 3: Generate embeddings and chunks =====
+        # ===== STEP 3+4: Index via pipeline =====
         if notification:
             await NotificationService.document_processing.notify_processing_progress(
                 session, notification, stage="chunking"
@@ -1871,57 +1871,24 @@ async def process_file_in_background_with_document(
 
         user_llm = await get_user_long_context_llm(session, user_id, search_space_id)
 
-        if user_llm:
+        await index_uploaded_file(
-            document_metadata = {
+            markdown_content=markdown_content,
-                "file_name": filename,
+            filename=filename,
-                "etl_service": etl_service,
+            etl_service=etl_service,
-                "document_type": "File Document",
+            search_space_id=search_space_id,
-            }
+            user_id=user_id,
-            summary_content, summary_embedding = await generate_document_summary(
+            session=session,
-                markdown_content, user_llm, document_metadata
+            llm=user_llm,
-            )
+            should_summarize=should_summarize,
-        else:
+        )
-            # Fallback: use truncated content as summary
-            summary_content = markdown_content[:4000]
-            from app.config import config
-
-            summary_embedding = config.embedding_model_instance.embed(summary_content)
-
-        chunks = await create_document_chunks(markdown_content)
-
-        # ===== STEP 4: Update document to READY =====
-        from sqlalchemy.orm.attributes import flag_modified
-
-        document.title = filename
-        document.content = summary_content
-        document.content_hash = content_hash
-        document.embedding = summary_embedding
-        document.document_metadata = {
-            "FILE_NAME": filename,
-            "ETL_SERVICE": etl_service or "UNKNOWN",
-            **(document.document_metadata or {}),
-        }
-        flag_modified(document, "document_metadata")
-
-        # Use safe_set_chunks to avoid async issues
-        safe_set_chunks(document, chunks)
-
-        document.source_markdown = markdown_content
-        document.content_needs_reindexing = False
-        document.updated_at = get_current_timestamp()
-        document.status = DocumentStatus.ready()  # Shows checkmark in UI
-
-        await session.commit()
-        await session.refresh(document)
 
         await task_logger.log_task_success(
             log_entry,
             f"Successfully processed file: {filename}",
             {
-                "document_id": document.id,
+                "document_id": doc_id,
                 "content_hash": content_hash,
                 "file_type": etl_service,
-                "chunks_count": len(chunks),
             },
         )
 
@@ -1946,7 +1913,7 @@ async def process_file_in_background_with_document(
             {
                 "error_type": type(e).__name__,
                 "filename": filename,
-                "document_id": document.id,
+                "document_id": doc_id,
             },
         )
         logging.error(f"Error processing file with document: {error_message}")

surfsense_backend/app/tasks/surfsense_docs_indexer.py

@@ -15,6 +15,7 @@ from sqlalchemy.orm import selectinload
 
 from app.config import config
 from app.db import SurfsenseDocsChunk, SurfsenseDocsDocument, async_session_maker
+from app.utils.document_converters import embed_text
 
 logger = logging.getLogger(__name__)
 
@@ -89,7 +90,7 @@ def create_surfsense_docs_chunks(content: str) -> list[SurfsenseDocsChunk]:
     return [
         SurfsenseDocsChunk(
             content=chunk.text,
-            embedding=config.embedding_model_instance.embed(chunk.text),
+            embedding=embed_text(chunk.text),
         )
         for chunk in config.chunker_instance.chunk(content)
     ]
@@ -154,7 +155,7 @@ async def index_surfsense_docs(session: AsyncSession) -> tuple[int, int, int, in
                 existing_doc.title = title
                 existing_doc.content = content
                 existing_doc.content_hash = content_hash
-                existing_doc.embedding = config.embedding_model_instance.embed(content)
+                existing_doc.embedding = embed_text(content)
                 existing_doc.chunks = chunks
                 existing_doc.updated_at = datetime.now(UTC)
 
@@ -170,7 +171,7 @@ async def index_surfsense_docs(session: AsyncSession) -> tuple[int, int, int, in
                     title=title,
                     content=content,
                     content_hash=content_hash,
-                    embedding=config.embedding_model_instance.embed(content),
+                    embedding=embed_text(content),
                     chunks=chunks,
                     updated_at=datetime.now(UTC),
                 )

surfsense_backend/app/utils/document_converters.py

@@ -1,11 +1,59 @@
 import hashlib
+import logging
+import warnings
 
+import numpy as np
 from litellm import get_model_info, token_counter
 
 from app.config import config
 from app.db import Chunk, DocumentType
 from app.prompts import SUMMARY_PROMPT_TEMPLATE
 
+logger = logging.getLogger(__name__)
+
+
+def _get_embedding_max_tokens() -> int:
+    """Get the max token limit for the configured embedding model.
+
+    Checks model properties in order: max_seq_length, _max_tokens.
+    Falls back to 8192 (OpenAI embedding default).
+    """
+    model = config.embedding_model_instance
+    for attr in ("max_seq_length", "_max_tokens"):
+        val = getattr(model, attr, None)
+        if isinstance(val, int) and val > 0:
+            return val
+    return 8192
+
+
+def truncate_for_embedding(text: str) -> str:
+    """Truncate text to fit within the embedding model's context window.
+
+    Uses the embedding model's own tokenizer for accurate token counting,
+    so the result is model-agnostic regardless of the underlying provider.
+    """
+    max_tokens = _get_embedding_max_tokens()
+    if len(text) // 3 <= max_tokens:
+        return text
+
+    tokenizer = config.embedding_model_instance.get_tokenizer()
+    tokens = tokenizer.encode(text)
+    if len(tokens) <= max_tokens:
+        return text
+
+    warnings.warn(
+        f"Truncating text from {len(tokens)} to {max_tokens} tokens for embedding.",
+        stacklevel=2,
+    )
+    return tokenizer.decode(tokens[:max_tokens])
+
+
+def embed_text(text: str) -> np.ndarray:
+    """Truncate text to fit and embed it. Drop-in replacement for
+    ``config.embedding_model_instance.embed(text)`` that never exceeds the
+    model's context window."""
+    return config.embedding_model_instance.embed(truncate_for_embedding(text))
+
 
 def get_model_context_window(model_name: str) -> int:
     """Get the total context window size for a model (input + output tokens)."""
@@ -146,7 +194,7 @@ async def generate_document_summary(
     else:
         enhanced_summary_content = summary_content
 
-    summary_embedding = config.embedding_model_instance.embed(enhanced_summary_content)
+    summary_embedding = embed_text(enhanced_summary_content)
 
     return enhanced_summary_content, summary_embedding
 
@@ -164,7 +212,7 @@ async def create_document_chunks(content: str) -> list[Chunk]:
     return [
         Chunk(
             content=chunk.text,
-            embedding=config.embedding_model_instance.embed(chunk.text),
+            embedding=embed_text(chunk.text),
         )
         for chunk in config.chunker_instance.chunk(content)
     ]

surfsense_backend/main.py

@@ -29,4 +29,7 @@ if __name__ == "__main__":
     config = uvicorn.Config(**config_kwargs)
     server = uvicorn.Server(config)
 
-    server.run()
+    if sys.platform == "win32":
+        asyncio.run(server.serve(), loop_factory=asyncio.SelectorEventLoop)
+    else:
+        server.run()

surfsense_backend/pyproject.toml

@@ -17,6 +17,7 @@ dependencies = [
     "kokoro>=0.9.4",
     "linkup-sdk>=0.2.4",
     "llama-cloud-services>=0.6.25",
+    "Markdown>=3.7",
     "markdownify>=0.14.1",
     "notion-client>=2.3.0",
     "numpy>=1.24.0",
@@ -65,11 +66,16 @@ dependencies = [
     "pypandoc_binary>=1.16.2",
     "typst>=0.14.0",
     "deepagents>=0.4.3",
+    "langchain-daytona>=0.0.2",
 ]
 
 [dependency-groups]
 dev = [
     "ruff>=0.12.5",
+    "pytest>=9.0.2",
+    "pytest-asyncio>=1.3.0",
+    "pytest-mock>=3.14",
+    "httpx>=0.28.1",
 ]
 
 [tool.ruff]
@@ -157,10 +163,27 @@ line-ending = "auto"
 
 [tool.ruff.lint.isort]
 # Group imports by type
-known-first-party = ["app"]
+known-first-party = ["app", "tests"]
 force-single-line = false
 combine-as-imports = true
 
+[tool.pytest.ini_options]
+asyncio_mode = "auto"
+asyncio_default_fixture_loop_scope = "session"
+asyncio_default_test_loop_scope = "session"
+testpaths = ["tests"]
+python_files = ["test_*.py"]
+python_classes = ["Test*"]
+python_functions = ["test_*"]
+addopts = "-v --tb=short -x --strict-markers -ra --durations=5"
+markers = [
+    "unit: pure logic tests, no DB or external services",
+    "integration: tests that require a real PostgreSQL database"
+]
+filterwarnings = [
+    "ignore::UserWarning:chonkie",
+]
+
 [tool.setuptools.packages.find]
 where = ["."]
 include = ["app*", "alembic*"]

surfsense_backend/tests/conftest.py

@@ -0,0 +1,61 @@
+"""Root conftest — shared fixtures available to all test modules."""
+
+from __future__ import annotations
+
+import os
+
+_DEFAULT_TEST_DB = (
+    "postgresql+asyncpg://postgres:postgres@localhost:5432/surfsense_test"
+)
+TEST_DATABASE_URL = os.environ.get("TEST_DATABASE_URL", _DEFAULT_TEST_DB)
+
+# Force the app to use the test database regardless of any pre-existing
+# DATABASE_URL in the environment (e.g. from .env or shell profile).
+os.environ["DATABASE_URL"] = TEST_DATABASE_URL
+
+import pytest  # noqa: E402
+
+from app.db import DocumentType  # noqa: E402
+from app.indexing_pipeline.connector_document import ConnectorDocument  # noqa: E402
+
+# ---------------------------------------------------------------------------
+# Unit test fixtures
+# ---------------------------------------------------------------------------
+
+
+@pytest.fixture
+def sample_user_id() -> str:
+    return "00000000-0000-0000-0000-000000000001"
+
+
+@pytest.fixture
+def sample_search_space_id() -> int:
+    return 1
+
+
+@pytest.fixture
+def sample_connector_id() -> int:
+    return 42
+
+
+@pytest.fixture
+def make_connector_document():
+    """
+    Generic factory for unit tests. Overridden in tests/integration/conftest.py
+    with real DB-backed IDs for integration tests.
+    """
+
+    def _make(**overrides):
+        defaults = {
+            "title": "Test Document",
+            "source_markdown": "## Heading\n\nSome content.",
+            "unique_id": "test-id-001",
+            "document_type": DocumentType.CLICKUP_CONNECTOR,
+            "search_space_id": 1,
+            "connector_id": 1,
+            "created_by_id": "00000000-0000-0000-0000-000000000001",
+        }
+        defaults.update(overrides)
+        return ConnectorDocument(**defaults)
+
+    return _make

surfsense_backend/tests/fixtures/sample.md

@@ -0,0 +1,51 @@
+# SurfSense Test Document
+
+## Overview
+
+This is a **sample markdown document** used for end-to-end testing of the manual
+document upload pipeline. It includes various markdown formatting elements.
+
+## Key Features
+
+- Document upload and processing
+- Automatic chunking of content
+- Embedding generation for semantic search
+- Real-time status tracking via ElectricSQL
+
+## Technical Architecture
+
+### Backend Stack
+
+The SurfSense backend is built with:
+
+1. **FastAPI** for the REST API
+2. **PostgreSQL** with pgvector for vector storage
+3. **Celery** with Redis for background task processing
+4. **Docling/Unstructured** for document parsing (ETL)
+
+### Processing Pipeline
+
+Documents go through a multi-stage pipeline:
+
+| Stage | Description |
+|-------|-------------|
+| Upload | File received via API endpoint |
+| Parsing | Content extracted using ETL service |
+| Chunking | Text split into semantic chunks |
+| Embedding | Vector representations generated |
+| Storage | Chunks stored with embeddings in pgvector |
+
+## Code Example
+
+```python
+async def process_document(file_path: str) -> Document:
+    content = extract_content(file_path)
+    chunks = create_chunks(content)
+    embeddings = generate_embeddings(chunks)
+    return store_document(chunks, embeddings)
+```
+
+## Conclusion
+
+This document serves as a test fixture to validate the complete document processing
+pipeline from upload through to chunk creation and embedding storage.

surfsense_backend/tests/fixtures/sample.txt

@@ -0,0 +1,34 @@
+SurfSense Document Upload Test
+
+This is a sample text document used for end-to-end testing of the manual document
+upload pipeline in SurfSense. The document contains multiple paragraphs to ensure
+that the chunking system has enough content to work with.
+
+Artificial Intelligence and Machine Learning
+
+Artificial intelligence (AI) is a broad field of computer science concerned with
+building smart machines capable of performing tasks that typically require human
+intelligence. Machine learning is a subset of AI that enables systems to learn and
+improve from experience without being explicitly programmed.
+
+Natural Language Processing
+
+Natural language processing (NLP) is a subfield of linguistics, computer science,
+and artificial intelligence concerned with the interactions between computers and
+human language. Key applications include machine translation, sentiment analysis,
+text summarization, and question answering systems.
+
+Vector Databases and Semantic Search
+
+Vector databases store data as high-dimensional vectors, enabling efficient
+similarity search operations. When combined with embedding models, they power
+semantic search systems that understand the meaning behind queries rather than
+relying on exact keyword matches. This technology is fundamental to modern
+retrieval-augmented generation (RAG) systems.
+
+Document Processing Pipelines
+
+Modern document processing pipelines involve several stages: extraction, transformation,
+chunking, embedding generation, and storage. Each stage plays a critical role in
+converting raw documents into searchable, structured knowledge that can be retrieved
+and used by AI systems for accurate information retrieval and generation.

surfsense_backend/tests/integration/conftest.py

@@ -0,0 +1,168 @@
+import uuid
+from unittest.mock import AsyncMock, MagicMock
+
+import pytest
+import pytest_asyncio
+from sqlalchemy import text
+from sqlalchemy.ext.asyncio import AsyncSession, create_async_engine
+from sqlalchemy.pool import NullPool
+
+from app.config import config as app_config
+from app.db import (
+    Base,
+    DocumentType,
+    SearchSourceConnector,
+    SearchSourceConnectorType,
+    SearchSpace,
+    User,
+)
+from app.indexing_pipeline.connector_document import ConnectorDocument
+from tests.conftest import TEST_DATABASE_URL
+
+_EMBEDDING_DIM = app_config.embedding_model_instance.dimension
+
+
+@pytest_asyncio.fixture(scope="session")
+async def async_engine():
+    engine = create_async_engine(
+        TEST_DATABASE_URL,
+        poolclass=NullPool,
+        echo=False,
+        # Required for asyncpg + savepoints: disables prepared statement cache
+        # to prevent "another operation is in progress" errors during savepoint rollbacks.
+        connect_args={"prepared_statement_cache_size": 0},
+    )
+
+    async with engine.begin() as conn:
+        await conn.execute(text("CREATE EXTENSION IF NOT EXISTS vector"))
+        await conn.run_sync(Base.metadata.create_all)
+
+    yield engine
+
+    # drop_all fails on circular FKs (new_chat_threads ↔ public_chat_snapshots).
+    # DROP SCHEMA CASCADE handles this without needing topological sort.
+    async with engine.begin() as conn:
+        await conn.execute(text("DROP SCHEMA public CASCADE"))
+        await conn.execute(text("CREATE SCHEMA public"))
+
+    await engine.dispose()
+
+
+@pytest_asyncio.fixture
+async def db_session(async_engine) -> AsyncSession:
+    # Bind the session to a connection that holds an outer transaction.
+    # join_transaction_mode="create_savepoint" makes session.commit() release
+    # a SAVEPOINT instead of committing the outer transaction, so the final
+    # transaction.rollback() undoes everything — including commits made by the
+    # service under test — leaving the DB clean for the next test.
+    async with async_engine.connect() as conn:
+        transaction = await conn.begin()
+        async with AsyncSession(
+            bind=conn,
+            expire_on_commit=False,
+            join_transaction_mode="create_savepoint",
+        ) as session:
+            yield session
+        await transaction.rollback()
+
+
+@pytest_asyncio.fixture
+async def db_user(db_session: AsyncSession) -> User:
+    user = User(
+        id=uuid.uuid4(),
+        email="test@surfsense.net",
+        hashed_password="hashed",
+        is_active=True,
+        is_superuser=False,
+        is_verified=True,
+    )
+    db_session.add(user)
+    await db_session.flush()
+    return user
+
+
+@pytest_asyncio.fixture
+async def db_connector(
+    db_session: AsyncSession, db_user: User, db_search_space: "SearchSpace"
+) -> SearchSourceConnector:
+    connector = SearchSourceConnector(
+        name="Test Connector",
+        connector_type=SearchSourceConnectorType.CLICKUP_CONNECTOR,
+        config={},
+        search_space_id=db_search_space.id,
+        user_id=db_user.id,
+    )
+    db_session.add(connector)
+    await db_session.flush()
+    return connector
+
+
+@pytest_asyncio.fixture
+async def db_search_space(db_session: AsyncSession, db_user: User) -> SearchSpace:
+    space = SearchSpace(
+        name="Test Space",
+        user_id=db_user.id,
+    )
+    db_session.add(space)
+    await db_session.flush()
+    return space
+
+
+@pytest.fixture
+def patched_summarize(monkeypatch) -> AsyncMock:
+    mock = AsyncMock(return_value="Mocked summary.")
+    monkeypatch.setattr(
+        "app.indexing_pipeline.indexing_pipeline_service.summarize_document",
+        mock,
+    )
+    return mock
+
+
+@pytest.fixture
+def patched_summarize_raises(monkeypatch) -> AsyncMock:
+    mock = AsyncMock(side_effect=RuntimeError("LLM unavailable"))
+    monkeypatch.setattr(
+        "app.indexing_pipeline.indexing_pipeline_service.summarize_document",
+        mock,
+    )
+    return mock
+
+
+@pytest.fixture
+def patched_embed_text(monkeypatch) -> MagicMock:
+    mock = MagicMock(return_value=[0.1] * _EMBEDDING_DIM)
+    monkeypatch.setattr(
+        "app.indexing_pipeline.indexing_pipeline_service.embed_text",
+        mock,
+    )
+    return mock
+
+
+@pytest.fixture
+def patched_chunk_text(monkeypatch) -> MagicMock:
+    mock = MagicMock(return_value=["Test chunk content."])
+    monkeypatch.setattr(
+        "app.indexing_pipeline.indexing_pipeline_service.chunk_text",
+        mock,
+    )
+    return mock
+
+
+@pytest.fixture
+def make_connector_document(db_connector, db_user):
+    """Integration-scoped override: uses real DB connector and user IDs."""
+
+    def _make(**overrides):
+        defaults = {
+            "title": "Test Document",
+            "source_markdown": "## Heading\n\nSome content.",
+            "unique_id": "test-id-001",
+            "document_type": DocumentType.CLICKUP_CONNECTOR,
+            "search_space_id": db_connector.search_space_id,
+            "connector_id": db_connector.id,
+            "created_by_id": str(db_user.id),
+        }
+        defaults.update(overrides)
+        return ConnectorDocument(**defaults)
+
+    return _make

surfsense_backend/tests/integration/document_upload/conftest.py

@@ -0,0 +1,289 @@
+"""Integration conftest — runs the FastAPI app in-process via ASGITransport.
+
+Prerequisites: PostgreSQL + pgvector only.
+
+External system boundaries are mocked:
+  - LLM summarization, text embedding, text chunking (external APIs)
+  - Redis heartbeat (external infrastructure)
+  - Task dispatch is swapped via DI (InlineTaskDispatcher)
+"""
+
+from __future__ import annotations
+
+import contextlib
+from collections.abc import AsyncGenerator
+from unittest.mock import AsyncMock, MagicMock
+
+import asyncpg
+import httpx
+import pytest
+from httpx import ASGITransport
+from sqlalchemy import text
+from sqlalchemy.ext.asyncio import create_async_engine
+from sqlalchemy.pool import NullPool
+
+from app.app import app
+from app.config import config as app_config
+from app.db import Base
+from app.services.task_dispatcher import get_task_dispatcher
+from tests.integration.conftest import TEST_DATABASE_URL
+from tests.utils.helpers import (
+    TEST_EMAIL,
+    auth_headers,
+    delete_document,
+    get_auth_token,
+    get_search_space_id,
+)
+
+_EMBEDDING_DIM = app_config.embedding_model_instance.dimension
+_ASYNCPG_URL = TEST_DATABASE_URL.replace("postgresql+asyncpg://", "postgresql://")
+
+pytestmark = pytest.mark.integration
+
+
+# ---------------------------------------------------------------------------
+# Inline task dispatcher (replaces Celery via DI — not a mock)
+# ---------------------------------------------------------------------------
+
+
+class InlineTaskDispatcher:
+    """Processes files synchronously in the calling coroutine.
+
+    Swapped in via FastAPI dependency_overrides so the upload endpoint
+    processes documents inline instead of dispatching to Celery.
+
+    Exceptions are caught to match Celery's fire-and-forget semantics —
+    the processing function already marks documents as failed internally.
+    """
+
+    async def dispatch_file_processing(
+        self,
+        *,
+        document_id: int,
+        temp_path: str,
+        filename: str,
+        search_space_id: int,
+        user_id: str,
+        should_summarize: bool = False,
+    ) -> None:
+        from app.tasks.celery_tasks.document_tasks import (
+            _process_file_with_document,
+        )
+
+        with contextlib.suppress(Exception):
+            await _process_file_with_document(
+                document_id,
+                temp_path,
+                filename,
+                search_space_id,
+                user_id,
+                should_summarize=should_summarize,
+            )
+
+
+app.dependency_overrides[get_task_dispatcher] = lambda: InlineTaskDispatcher()
+
+
+# ---------------------------------------------------------------------------
+# Database setup (ASGITransport skips the app lifespan)
+# ---------------------------------------------------------------------------
+
+
+@pytest.fixture(scope="session")
+async def _ensure_tables():
+    """Create DB tables and extensions once per session."""
+    engine = create_async_engine(TEST_DATABASE_URL, poolclass=NullPool)
+    async with engine.begin() as conn:
+        await conn.execute(text("CREATE EXTENSION IF NOT EXISTS vector"))
+        await conn.execute(text("CREATE EXTENSION IF NOT EXISTS pg_trgm"))
+        await conn.run_sync(Base.metadata.create_all)
+    await engine.dispose()
+
+
+# ---------------------------------------------------------------------------
+# Auth & search space (session-scoped, via the in-process app)
+# ---------------------------------------------------------------------------
+
+
+@pytest.fixture(scope="session")
+async def auth_token(_ensure_tables) -> str:
+    """Authenticate once per session, registering the user if needed."""
+    async with httpx.AsyncClient(
+        transport=ASGITransport(app=app), base_url="http://test", timeout=30.0
+    ) as c:
+        return await get_auth_token(c)
+
+
+@pytest.fixture(scope="session")
+async def search_space_id(auth_token: str) -> int:
+    """Discover the first search space belonging to the test user."""
+    async with httpx.AsyncClient(
+        transport=ASGITransport(app=app), base_url="http://test", timeout=30.0
+    ) as c:
+        return await get_search_space_id(c, auth_token)
+
+
+@pytest.fixture(scope="session")
+def headers(auth_token: str) -> dict[str, str]:
+    return auth_headers(auth_token)
+
+
+# ---------------------------------------------------------------------------
+# Per-test HTTP client & cleanup
+# ---------------------------------------------------------------------------
+
+
+@pytest.fixture
+async def client() -> AsyncGenerator[httpx.AsyncClient]:
+    """Per-test async HTTP client using ASGITransport (no running server)."""
+    async with httpx.AsyncClient(
+        transport=ASGITransport(app=app), base_url="http://test", timeout=180.0
+    ) as c:
+        yield c
+
+
+@pytest.fixture
+def cleanup_doc_ids() -> list[int]:
+    """Accumulator for document IDs that should be deleted after the test."""
+    return []
+
+
+@pytest.fixture(scope="session", autouse=True)
+async def _purge_test_search_space(search_space_id: int):
+    """Delete stale documents from previous runs before the session starts."""
+    conn = await asyncpg.connect(_ASYNCPG_URL)
+    try:
+        result = await conn.execute(
+            "DELETE FROM documents WHERE search_space_id = $1",
+            search_space_id,
+        )
+        deleted = int(result.split()[-1])
+        if deleted:
+            print(
+                f"\n[purge] Deleted {deleted} stale document(s) "
+                f"from search space {search_space_id}"
+            )
+    finally:
+        await conn.close()
+    yield
+
+
+@pytest.fixture(autouse=True)
+async def _cleanup_documents(
+    client: httpx.AsyncClient,
+    headers: dict[str, str],
+    cleanup_doc_ids: list[int],
+):
+    """Delete test documents after every test (API first, DB fallback)."""
+    yield
+
+    remaining_ids: list[int] = []
+    for doc_id in cleanup_doc_ids:
+        try:
+            resp = await delete_document(client, headers, doc_id)
+            if resp.status_code == 409:
+                remaining_ids.append(doc_id)
+        except Exception:
+            remaining_ids.append(doc_id)
+
+    if remaining_ids:
+        conn = await asyncpg.connect(_ASYNCPG_URL)
+        try:
+            await conn.execute(
+                "DELETE FROM documents WHERE id = ANY($1::int[])",
+                remaining_ids,
+            )
+        finally:
+            await conn.close()
+
+
+# ---------------------------------------------------------------------------
+# Page-limit helpers (direct DB for setup, API for verification)
+# ---------------------------------------------------------------------------
+
+
+async def _get_user_page_usage(email: str) -> tuple[int, int]:
+    conn = await asyncpg.connect(_ASYNCPG_URL)
+    try:
+        row = await conn.fetchrow(
+            'SELECT pages_used, pages_limit FROM "user" WHERE email = $1',
+            email,
+        )
+        assert row is not None, f"User {email!r} not found in database"
+        return row["pages_used"], row["pages_limit"]
+    finally:
+        await conn.close()
+
+
+async def _set_user_page_limits(
+    email: str, *, pages_used: int, pages_limit: int
+) -> None:
+    conn = await asyncpg.connect(_ASYNCPG_URL)
+    try:
+        await conn.execute(
+            'UPDATE "user" SET pages_used = $1, pages_limit = $2 WHERE email = $3',
+            pages_used,
+            pages_limit,
+            email,
+        )
+    finally:
+        await conn.close()
+
+
+@pytest.fixture
+async def page_limits():
+    """Manipulate the test user's page limits (direct DB for setup only).
+
+    Automatically restores original values after each test.
+    """
+
+    class _PageLimits:
+        async def set(self, *, pages_used: int, pages_limit: int) -> None:
+            await _set_user_page_limits(
+                TEST_EMAIL, pages_used=pages_used, pages_limit=pages_limit
+            )
+
+    original = await _get_user_page_usage(TEST_EMAIL)
+    yield _PageLimits()
+    await _set_user_page_limits(
+        TEST_EMAIL, pages_used=original[0], pages_limit=original[1]
+    )
+
+
+# ---------------------------------------------------------------------------
+# Mock external system boundaries
+# ---------------------------------------------------------------------------
+
+
+@pytest.fixture(autouse=True)
+def _mock_external_apis(monkeypatch):
+    """Mock LLM, embedding, and chunking — these are external API boundaries."""
+    monkeypatch.setattr(
+        "app.indexing_pipeline.indexing_pipeline_service.summarize_document",
+        AsyncMock(return_value="Mocked summary."),
+    )
+    monkeypatch.setattr(
+        "app.indexing_pipeline.indexing_pipeline_service.embed_text",
+        MagicMock(return_value=[0.1] * _EMBEDDING_DIM),
+    )
+    monkeypatch.setattr(
+        "app.indexing_pipeline.indexing_pipeline_service.chunk_text",
+        MagicMock(return_value=["Test chunk content."]),
+    )
+
+
+@pytest.fixture(autouse=True)
+def _mock_redis_heartbeat(monkeypatch):
+    """Mock Redis heartbeat — Redis is an external infrastructure boundary."""
+    monkeypatch.setattr(
+        "app.tasks.celery_tasks.document_tasks._start_heartbeat",
+        lambda notification_id: None,
+    )
+    monkeypatch.setattr(
+        "app.tasks.celery_tasks.document_tasks._stop_heartbeat",
+        lambda notification_id: None,
+    )
+    monkeypatch.setattr(
+        "app.tasks.celery_tasks.document_tasks._run_heartbeat_loop",
+        AsyncMock(),
+    )

surfsense_backend/tests/integration/document_upload/test_document_upload.py

@@ -0,0 +1,337 @@
+"""
+Integration tests for the document upload HTTP API.
+
+Covers the API contract, auth, duplicate detection, and error handling.
+Pipeline internals are tested in the ``indexing_pipeline`` suite.
+
+Requires PostgreSQL + pgvector.
+"""
+
+from __future__ import annotations
+
+import shutil
+from pathlib import Path
+
+import httpx
+import pytest
+
+from tests.utils.helpers import (
+    FIXTURES_DIR,
+    poll_document_status,
+    upload_file,
+    upload_multiple_files,
+)
+
+pytestmark = pytest.mark.integration
+
+
+# ---------------------------------------------------------------------------
+# Upload smoke tests (one per distinct code-path: direct-read & ETL)
+# ---------------------------------------------------------------------------
+
+
+class TestTxtFileUpload:
+    """Upload a plain-text file (direct-read path) via the HTTP API."""
+
+    async def test_upload_txt_returns_document_id(
+        self,
+        client: httpx.AsyncClient,
+        headers: dict[str, str],
+        search_space_id: int,
+        cleanup_doc_ids: list[int],
+    ):
+        resp = await upload_file(
+            client, headers, "sample.txt", search_space_id=search_space_id
+        )
+        assert resp.status_code == 200
+
+        body = resp.json()
+        assert body["pending_files"] >= 1
+        assert len(body["document_ids"]) >= 1
+        cleanup_doc_ids.extend(body["document_ids"])
+
+    async def test_txt_processing_reaches_ready(
+        self,
+        client: httpx.AsyncClient,
+        headers: dict[str, str],
+        search_space_id: int,
+        cleanup_doc_ids: list[int],
+    ):
+        resp = await upload_file(
+            client, headers, "sample.txt", search_space_id=search_space_id
+        )
+        assert resp.status_code == 200
+        doc_ids = resp.json()["document_ids"]
+        cleanup_doc_ids.extend(doc_ids)
+
+        statuses = await poll_document_status(
+            client, headers, doc_ids, search_space_id=search_space_id
+        )
+        for did in doc_ids:
+            assert statuses[did]["status"]["state"] == "ready"
+
+
+class TestPdfFileUpload:
+    """Upload a PDF (ETL extraction path) via the HTTP API."""
+
+    async def test_pdf_processing_reaches_ready(
+        self,
+        client: httpx.AsyncClient,
+        headers: dict[str, str],
+        search_space_id: int,
+        cleanup_doc_ids: list[int],
+    ):
+        resp = await upload_file(
+            client, headers, "sample.pdf", search_space_id=search_space_id
+        )
+        assert resp.status_code == 200
+        doc_ids = resp.json()["document_ids"]
+        cleanup_doc_ids.extend(doc_ids)
+
+        statuses = await poll_document_status(
+            client, headers, doc_ids, search_space_id=search_space_id, timeout=300.0
+        )
+        for did in doc_ids:
+            assert statuses[did]["status"]["state"] == "ready"
+
+
+# ---------------------------------------------------------------------------
+# Test D: Upload multiple files in a single request
+# ---------------------------------------------------------------------------
+
+
+class TestMultiFileUpload:
+    """Upload several files at once and verify the API response contract."""
+
+    async def test_multi_upload_returns_all_ids(
+        self,
+        client: httpx.AsyncClient,
+        headers: dict[str, str],
+        search_space_id: int,
+        cleanup_doc_ids: list[int],
+    ):
+        resp = await upload_multiple_files(
+            client,
+            headers,
+            ["sample.txt", "sample.md"],
+            search_space_id=search_space_id,
+        )
+        assert resp.status_code == 200
+
+        body = resp.json()
+        assert body["pending_files"] == 2
+        assert len(body["document_ids"]) == 2
+        cleanup_doc_ids.extend(body["document_ids"])
+
+
+# ---------------------------------------------------------------------------
+# Test E: Duplicate file upload (same file uploaded twice)
+# ---------------------------------------------------------------------------
+
+
+class TestDuplicateFileUpload:
+    """
+    Uploading the exact same file a second time should be detected as a
+    duplicate via ``unique_identifier_hash``.
+    """
+
+    async def test_duplicate_file_is_skipped(
+        self,
+        client: httpx.AsyncClient,
+        headers: dict[str, str],
+        search_space_id: int,
+        cleanup_doc_ids: list[int],
+    ):
+        resp1 = await upload_file(
+            client, headers, "sample.txt", search_space_id=search_space_id
+        )
+        assert resp1.status_code == 200
+        first_ids = resp1.json()["document_ids"]
+        cleanup_doc_ids.extend(first_ids)
+
+        await poll_document_status(
+            client, headers, first_ids, search_space_id=search_space_id
+        )
+
+        resp2 = await upload_file(
+            client, headers, "sample.txt", search_space_id=search_space_id
+        )
+        assert resp2.status_code == 200
+
+        body2 = resp2.json()
+        assert body2["skipped_duplicates"] >= 1
+        assert len(body2["duplicate_document_ids"]) >= 1
+        cleanup_doc_ids.extend(body2.get("document_ids", []))
+
+
+# ---------------------------------------------------------------------------
+# Test F: Duplicate content detection (different name, same content)
+# ---------------------------------------------------------------------------
+
+
+class TestDuplicateContentDetection:
+    """
+    Uploading a file with a different name but identical content should be
+    detected as duplicate content via ``content_hash``.
+    """
+
+    async def test_same_content_different_name_detected(
+        self,
+        client: httpx.AsyncClient,
+        headers: dict[str, str],
+        search_space_id: int,
+        cleanup_doc_ids: list[int],
+        tmp_path: Path,
+    ):
+        resp1 = await upload_file(
+            client, headers, "sample.txt", search_space_id=search_space_id
+        )
+        assert resp1.status_code == 200
+        first_ids = resp1.json()["document_ids"]
+        cleanup_doc_ids.extend(first_ids)
+        await poll_document_status(
+            client, headers, first_ids, search_space_id=search_space_id
+        )
+
+        src = FIXTURES_DIR / "sample.txt"
+        dest = tmp_path / "renamed_sample.txt"
+        shutil.copy2(src, dest)
+
+        with open(dest, "rb") as f:
+            resp2 = await client.post(
+                "/api/v1/documents/fileupload",
+                headers=headers,
+                files={"files": ("renamed_sample.txt", f)},
+                data={"search_space_id": str(search_space_id)},
+            )
+        assert resp2.status_code == 200
+        second_ids = resp2.json()["document_ids"]
+        cleanup_doc_ids.extend(second_ids)
+        assert second_ids, (
+            "Expected at least one document id for renamed duplicate content upload"
+        )
+
+        statuses = await poll_document_status(
+            client, headers, second_ids, search_space_id=search_space_id
+        )
+        for did in second_ids:
+            assert statuses[did]["status"]["state"] == "failed"
+            assert "duplicate" in statuses[did]["status"].get("reason", "").lower()
+
+
+# ---------------------------------------------------------------------------
+# Test G: Empty / corrupt file handling
+# ---------------------------------------------------------------------------
+
+
+class TestEmptyFileUpload:
+    """An empty file should be processed but ultimately fail gracefully."""
+
+    async def test_empty_pdf_fails(
+        self,
+        client: httpx.AsyncClient,
+        headers: dict[str, str],
+        search_space_id: int,
+        cleanup_doc_ids: list[int],
+    ):
+        resp = await upload_file(
+            client, headers, "empty.pdf", search_space_id=search_space_id
+        )
+        assert resp.status_code == 200
+
+        doc_ids = resp.json()["document_ids"]
+        cleanup_doc_ids.extend(doc_ids)
+        assert doc_ids, "Expected at least one document id for empty PDF upload"
+
+        statuses = await poll_document_status(
+            client, headers, doc_ids, search_space_id=search_space_id, timeout=120.0
+        )
+        for did in doc_ids:
+            assert statuses[did]["status"]["state"] == "failed"
+            assert statuses[did]["status"].get("reason"), (
+                "Failed document should include a reason"
+            )
+
+
+# ---------------------------------------------------------------------------
+# Test H: Upload without authentication
+# ---------------------------------------------------------------------------
+
+
+class TestUnauthenticatedUpload:
+    """Requests without a valid JWT should be rejected."""
+
+    async def test_upload_without_auth_returns_401(
+        self,
+        client: httpx.AsyncClient,
+        search_space_id: int,
+    ):
+        file_path = FIXTURES_DIR / "sample.txt"
+        with open(file_path, "rb") as f:
+            resp = await client.post(
+                "/api/v1/documents/fileupload",
+                files={"files": ("sample.txt", f)},
+                data={"search_space_id": str(search_space_id)},
+            )
+        assert resp.status_code == 401
+
+
+# ---------------------------------------------------------------------------
+# Test I: Upload with no files attached
+# ---------------------------------------------------------------------------
+
+
+class TestNoFilesUpload:
+    """Submitting the form with zero files should return a validation error."""
+
+    async def test_no_files_returns_error(
+        self,
+        client: httpx.AsyncClient,
+        headers: dict[str, str],
+        search_space_id: int,
+    ):
+        resp = await client.post(
+            "/api/v1/documents/fileupload",
+            headers=headers,
+            data={"search_space_id": str(search_space_id)},
+        )
+        assert resp.status_code in {400, 422}
+
+
+# ---------------------------------------------------------------------------
+# Test K: Searchability after upload
+# ---------------------------------------------------------------------------
+
+
+class TestDocumentSearchability:
+    """After upload reaches ready, the document must appear in the title search."""
+
+    async def test_uploaded_document_appears_in_search(
+        self,
+        client: httpx.AsyncClient,
+        headers: dict[str, str],
+        search_space_id: int,
+        cleanup_doc_ids: list[int],
+    ):
+        resp = await upload_file(
+            client, headers, "sample.txt", search_space_id=search_space_id
+        )
+        assert resp.status_code == 200
+        doc_ids = resp.json()["document_ids"]
+        cleanup_doc_ids.extend(doc_ids)
+
+        await poll_document_status(
+            client, headers, doc_ids, search_space_id=search_space_id
+        )
+
+        search_resp = await client.get(
+            "/api/v1/documents/search",
+            headers=headers,
+            params={"title": "sample", "search_space_id": search_space_id},
+        )
+        assert search_resp.status_code == 200
+
+        result_ids = [d["id"] for d in search_resp.json()["items"]]
+        assert doc_ids[0] in result_ids, (
+            f"Uploaded document {doc_ids[0]} not found in search results: {result_ids}"
+        )

surfsense_backend/tests/integration/document_upload/test_page_limits.py

@@ -0,0 +1,332 @@
+"""
+Integration tests for page-limit enforcement during document upload.
+
+These tests manipulate the test user's ``pages_used`` / ``pages_limit``
+columns directly in the database (setup only) and then exercise the upload
+pipeline to verify that:
+
+  - Uploads are rejected *before* ETL when the limit is exhausted.
+  - ``pages_used`` increases after a successful upload (verified via API).
+  - A ``page_limit_exceeded`` notification is created on rejection.
+  - ``pages_used`` is not modified when a document fails processing.
+
+All tests reuse the existing small fixtures (``sample.pdf``, ``sample.txt``)
+so no additional processing time is introduced.
+
+Prerequisites:
+  - PostgreSQL + pgvector
+"""
+
+from __future__ import annotations
+
+import httpx
+import pytest
+
+from tests.utils.helpers import (
+    get_notifications,
+    poll_document_status,
+    upload_file,
+)
+
+pytestmark = pytest.mark.integration
+
+
+# ---------------------------------------------------------------------------
+# Helper: read pages_used through the public API
+# ---------------------------------------------------------------------------
+
+
+async def _get_pages_used(client: httpx.AsyncClient, headers: dict[str, str]) -> int:
+    """Fetch the current user's pages_used via the /users/me API."""
+    resp = await client.get("/users/me", headers=headers)
+    assert resp.status_code == 200, (
+        f"GET /users/me failed ({resp.status_code}): {resp.text}"
+    )
+    return resp.json()["pages_used"]
+
+
+# ---------------------------------------------------------------------------
+# Test A: Successful upload increments pages_used
+# ---------------------------------------------------------------------------
+
+
+class TestPageUsageIncrementsOnSuccess:
+    """After a successful PDF upload the user's ``pages_used`` must grow."""
+
+    async def test_pages_used_increases_after_pdf_upload(
+        self,
+        client: httpx.AsyncClient,
+        headers: dict[str, str],
+        search_space_id: int,
+        cleanup_doc_ids: list[int],
+        page_limits,
+    ):
+        await page_limits.set(pages_used=0, pages_limit=1000)
+
+        resp = await upload_file(
+            client, headers, "sample.pdf", search_space_id=search_space_id
+        )
+        assert resp.status_code == 200
+        doc_ids = resp.json()["document_ids"]
+        cleanup_doc_ids.extend(doc_ids)
+
+        statuses = await poll_document_status(
+            client, headers, doc_ids, search_space_id=search_space_id, timeout=300.0
+        )
+        for did in doc_ids:
+            assert statuses[did]["status"]["state"] == "ready"
+
+        used = await _get_pages_used(client, headers)
+        assert used > 0, "pages_used should have increased after successful processing"
+
+
+# ---------------------------------------------------------------------------
+# Test B: Upload rejected when page limit is fully exhausted
+# ---------------------------------------------------------------------------
+
+
+class TestUploadRejectedWhenLimitExhausted:
+    """
+    When ``pages_used == pages_limit`` (zero remaining) the document
+    should reach ``failed`` status with a page-limit reason.
+    """
+
+    async def test_pdf_fails_when_no_pages_remaining(
+        self,
+        client: httpx.AsyncClient,
+        headers: dict[str, str],
+        search_space_id: int,
+        cleanup_doc_ids: list[int],
+        page_limits,
+    ):
+        await page_limits.set(pages_used=100, pages_limit=100)
+
+        resp = await upload_file(
+            client, headers, "sample.pdf", search_space_id=search_space_id
+        )
+        assert resp.status_code == 200
+        doc_ids = resp.json()["document_ids"]
+        cleanup_doc_ids.extend(doc_ids)
+
+        statuses = await poll_document_status(
+            client, headers, doc_ids, search_space_id=search_space_id, timeout=300.0
+        )
+        for did in doc_ids:
+            assert statuses[did]["status"]["state"] == "failed"
+            reason = statuses[did]["status"].get("reason", "").lower()
+            assert "page limit" in reason, (
+                f"Expected 'page limit' in failure reason, got: {reason!r}"
+            )
+
+    async def test_pages_used_unchanged_after_limit_rejection(
+        self,
+        client: httpx.AsyncClient,
+        headers: dict[str, str],
+        search_space_id: int,
+        cleanup_doc_ids: list[int],
+        page_limits,
+    ):
+        await page_limits.set(pages_used=50, pages_limit=50)
+
+        resp = await upload_file(
+            client, headers, "sample.pdf", search_space_id=search_space_id
+        )
+        assert resp.status_code == 200
+        doc_ids = resp.json()["document_ids"]
+        cleanup_doc_ids.extend(doc_ids)
+
+        await poll_document_status(
+            client, headers, doc_ids, search_space_id=search_space_id, timeout=300.0
+        )
+
+        used = await _get_pages_used(client, headers)
+        assert used == 50, (
+            f"pages_used should remain 50 after rejected upload, got {used}"
+        )
+
+
+# ---------------------------------------------------------------------------
+# Test C: Page-limit notification is created on rejection
+# ---------------------------------------------------------------------------
+
+
+class TestPageLimitNotification:
+    """A ``page_limit_exceeded`` notification must be created when upload
+    is rejected due to the limit."""
+
+    async def test_page_limit_exceeded_notification_created(
+        self,
+        client: httpx.AsyncClient,
+        headers: dict[str, str],
+        search_space_id: int,
+        cleanup_doc_ids: list[int],
+        page_limits,
+    ):
+        await page_limits.set(pages_used=100, pages_limit=100)
+
+        resp = await upload_file(
+            client, headers, "sample.pdf", search_space_id=search_space_id
+        )
+        assert resp.status_code == 200
+        doc_ids = resp.json()["document_ids"]
+        cleanup_doc_ids.extend(doc_ids)
+
+        await poll_document_status(
+            client, headers, doc_ids, search_space_id=search_space_id, timeout=300.0
+        )
+
+        notifications = await get_notifications(
+            client,
+            headers,
+            type_filter="page_limit_exceeded",
+            search_space_id=search_space_id,
+        )
+        assert len(notifications) >= 1, (
+            "Expected at least one page_limit_exceeded notification"
+        )
+
+        latest = notifications[0]
+        assert (
+            "page limit" in latest["title"].lower()
+            or "page limit" in latest["message"].lower()
+        ), (
+            f"Notification should mention page limit: title={latest['title']!r}, "
+            f"message={latest['message']!r}"
+        )
+
+
+# ---------------------------------------------------------------------------
+# Test D: Successful upload creates a completed document_processing notification
+# ---------------------------------------------------------------------------
+
+
+class TestDocumentProcessingNotification:
+    """A ``document_processing`` notification with ``completed`` status must
+    exist after a successful upload."""
+
+    async def test_processing_completed_notification_exists(
+        self,
+        client: httpx.AsyncClient,
+        headers: dict[str, str],
+        search_space_id: int,
+        cleanup_doc_ids: list[int],
+        page_limits,
+    ):
+        await page_limits.set(pages_used=0, pages_limit=1000)
+
+        resp = await upload_file(
+            client, headers, "sample.txt", search_space_id=search_space_id
+        )
+        assert resp.status_code == 200
+        doc_ids = resp.json()["document_ids"]
+        cleanup_doc_ids.extend(doc_ids)
+
+        await poll_document_status(
+            client, headers, doc_ids, search_space_id=search_space_id
+        )
+
+        notifications = await get_notifications(
+            client,
+            headers,
+            type_filter="document_processing",
+            search_space_id=search_space_id,
+        )
+        completed = [
+            n
+            for n in notifications
+            if n.get("metadata", {}).get("processing_stage") == "completed"
+        ]
+        assert len(completed) >= 1, (
+            "Expected at least one document_processing notification with 'completed' stage"
+        )
+
+
+# ---------------------------------------------------------------------------
+# Test E: pages_used unchanged when a document fails for non-limit reasons
+# ---------------------------------------------------------------------------
+
+
+class TestPagesUnchangedOnProcessingFailure:
+    """If a document fails during ETL (e.g. empty/corrupt file) rather than
+    a page-limit rejection, ``pages_used`` should remain unchanged."""
+
+    async def test_pages_used_stable_on_etl_failure(
+        self,
+        client: httpx.AsyncClient,
+        headers: dict[str, str],
+        search_space_id: int,
+        cleanup_doc_ids: list[int],
+        page_limits,
+    ):
+        await page_limits.set(pages_used=10, pages_limit=1000)
+
+        resp = await upload_file(
+            client, headers, "empty.pdf", search_space_id=search_space_id
+        )
+        assert resp.status_code == 200
+        doc_ids = resp.json()["document_ids"]
+        cleanup_doc_ids.extend(doc_ids)
+
+        if doc_ids:
+            statuses = await poll_document_status(
+                client, headers, doc_ids, search_space_id=search_space_id, timeout=120.0
+            )
+            for did in doc_ids:
+                assert statuses[did]["status"]["state"] == "failed"
+
+        used = await _get_pages_used(client, headers)
+        assert used == 10, f"pages_used should remain 10 after ETL failure, got {used}"
+
+
+# ---------------------------------------------------------------------------
+# Test F: Second upload rejected after first consumes remaining quota
+# ---------------------------------------------------------------------------
+
+
+class TestSecondUploadExceedsLimit:
+    """Upload one PDF successfully, consuming the quota, then verify a
+    second upload is rejected."""
+
+    async def test_second_upload_rejected_after_quota_consumed(
+        self,
+        client: httpx.AsyncClient,
+        headers: dict[str, str],
+        search_space_id: int,
+        cleanup_doc_ids: list[int],
+        page_limits,
+    ):
+        await page_limits.set(pages_used=0, pages_limit=1)
+
+        resp1 = await upload_file(
+            client, headers, "sample.pdf", search_space_id=search_space_id
+        )
+        assert resp1.status_code == 200
+        first_ids = resp1.json()["document_ids"]
+        cleanup_doc_ids.extend(first_ids)
+
+        statuses1 = await poll_document_status(
+            client, headers, first_ids, search_space_id=search_space_id, timeout=300.0
+        )
+        for did in first_ids:
+            assert statuses1[did]["status"]["state"] == "ready"
+
+        resp2 = await upload_file(
+            client,
+            headers,
+            "sample.pdf",
+            search_space_id=search_space_id,
+            filename_override="sample_copy.pdf",
+        )
+        assert resp2.status_code == 200
+        second_ids = resp2.json()["document_ids"]
+        cleanup_doc_ids.extend(second_ids)
+
+        statuses2 = await poll_document_status(
+            client, headers, second_ids, search_space_id=search_space_id, timeout=300.0
+        )
+        for did in second_ids:
+            assert statuses2[did]["status"]["state"] == "failed"
+            reason = statuses2[did]["status"].get("reason", "").lower()
+            assert "page limit" in reason, (
+                f"Expected 'page limit' in failure reason, got: {reason!r}"
+            )

surfsense_backend/tests/integration/document_upload/test_upload_limits.py

@@ -0,0 +1,145 @@
+"""
+Integration tests for backend file upload limit enforcement.
+
+These tests verify that the API rejects uploads that exceed:
+  - Max files per upload (10)
+  - Max per-file size (50 MB)
+  - Max total upload size (200 MB)
+
+The limits mirror the frontend's DocumentUploadTab.tsx constants and are
+enforced server-side to protect against direct API calls.
+
+Prerequisites:
+  - PostgreSQL + pgvector
+"""
+
+from __future__ import annotations
+
+import io
+
+import httpx
+import pytest
+
+pytestmark = pytest.mark.integration
+
+
+# ---------------------------------------------------------------------------
+# Test A: File count limit
+# ---------------------------------------------------------------------------
+
+
+class TestFileCountLimit:
+    """Uploading more than 10 files in a single request should be rejected."""
+
+    async def test_11_files_returns_413(
+        self,
+        client: httpx.AsyncClient,
+        headers: dict[str, str],
+        search_space_id: int,
+    ):
+        files = [
+            ("files", (f"file_{i}.txt", io.BytesIO(b"test content"), "text/plain"))
+            for i in range(11)
+        ]
+        resp = await client.post(
+            "/api/v1/documents/fileupload",
+            headers=headers,
+            files=files,
+            data={"search_space_id": str(search_space_id)},
+        )
+        assert resp.status_code == 413
+        assert "too many files" in resp.json()["detail"].lower()
+
+    async def test_10_files_accepted(
+        self,
+        client: httpx.AsyncClient,
+        headers: dict[str, str],
+        search_space_id: int,
+        cleanup_doc_ids: list[int],
+    ):
+        files = [
+            ("files", (f"file_{i}.txt", io.BytesIO(b"test content"), "text/plain"))
+            for i in range(10)
+        ]
+        resp = await client.post(
+            "/api/v1/documents/fileupload",
+            headers=headers,
+            files=files,
+            data={"search_space_id": str(search_space_id)},
+        )
+        assert resp.status_code == 200
+        cleanup_doc_ids.extend(resp.json().get("document_ids", []))
+
+
+# ---------------------------------------------------------------------------
+# Test B: Per-file size limit
+# ---------------------------------------------------------------------------
+
+
+class TestPerFileSizeLimit:
+    """A single file exceeding 50 MB should be rejected."""
+
+    async def test_oversized_file_returns_413(
+        self,
+        client: httpx.AsyncClient,
+        headers: dict[str, str],
+        search_space_id: int,
+    ):
+        oversized = io.BytesIO(b"\x00" * (50 * 1024 * 1024 + 1))
+        resp = await client.post(
+            "/api/v1/documents/fileupload",
+            headers=headers,
+            files=[("files", ("big.pdf", oversized, "application/pdf"))],
+            data={"search_space_id": str(search_space_id)},
+        )
+        assert resp.status_code == 413
+        assert "per-file limit" in resp.json()["detail"].lower()
+
+    async def test_file_at_limit_accepted(
+        self,
+        client: httpx.AsyncClient,
+        headers: dict[str, str],
+        search_space_id: int,
+        cleanup_doc_ids: list[int],
+    ):
+        at_limit = io.BytesIO(b"\x00" * (50 * 1024 * 1024))
+        resp = await client.post(
+            "/api/v1/documents/fileupload",
+            headers=headers,
+            files=[("files", ("exact50mb.txt", at_limit, "text/plain"))],
+            data={"search_space_id": str(search_space_id)},
+        )
+        assert resp.status_code == 200
+        cleanup_doc_ids.extend(resp.json().get("document_ids", []))
+
+
+# ---------------------------------------------------------------------------
+# Test C: Total upload size limit
+# ---------------------------------------------------------------------------
+
+
+class TestTotalSizeLimit:
+    """Multiple files whose combined size exceeds 200 MB should be rejected."""
+
+    async def test_total_size_over_200mb_returns_413(
+        self,
+        client: httpx.AsyncClient,
+        headers: dict[str, str],
+        search_space_id: int,
+    ):
+        chunk_size = 45 * 1024 * 1024  # 45 MB each
+        files = [
+            (
+                "files",
+                (f"chunk_{i}.txt", io.BytesIO(b"\x00" * chunk_size), "text/plain"),
+            )
+            for i in range(5)  # 5 x 45 MB = 225 MB > 200 MB
+        ]
+        resp = await client.post(
+            "/api/v1/documents/fileupload",
+            headers=headers,
+            files=files,
+            data={"search_space_id": str(search_space_id)},
+        )
+        assert resp.status_code == 413
+        assert "total upload size" in resp.json()["detail"].lower()

surfsense_backend/tests/integration/indexing_pipeline/adapters/test_file_upload_adapter.py

@@ -0,0 +1,99 @@
+import pytest
+from sqlalchemy import select
+
+from app.db import Chunk, Document, DocumentStatus
+from app.indexing_pipeline.adapters.file_upload_adapter import index_uploaded_file
+
+pytestmark = pytest.mark.integration
+
+
+@pytest.mark.usefixtures(
+    "patched_summarize", "patched_embed_text", "patched_chunk_text"
+)
+async def test_sets_status_ready(db_session, db_search_space, db_user, mocker):
+    """Document status is READY after successful indexing."""
+    await index_uploaded_file(
+        markdown_content="## Hello\n\nSome content.",
+        filename="test.pdf",
+        etl_service="UNSTRUCTURED",
+        search_space_id=db_search_space.id,
+        user_id=str(db_user.id),
+        session=db_session,
+        llm=mocker.Mock(),
+    )
+
+    result = await db_session.execute(
+        select(Document).filter(Document.search_space_id == db_search_space.id)
+    )
+    document = result.scalars().first()
+
+    assert DocumentStatus.is_state(document.status, DocumentStatus.READY)
+
+
+@pytest.mark.usefixtures(
+    "patched_summarize", "patched_embed_text", "patched_chunk_text"
+)
+async def test_content_is_summary(db_session, db_search_space, db_user, mocker):
+    """Document content is set to the LLM-generated summary."""
+    await index_uploaded_file(
+        markdown_content="## Hello\n\nSome content.",
+        filename="test.pdf",
+        etl_service="UNSTRUCTURED",
+        search_space_id=db_search_space.id,
+        user_id=str(db_user.id),
+        session=db_session,
+        llm=mocker.Mock(),
+    )
+
+    result = await db_session.execute(
+        select(Document).filter(Document.search_space_id == db_search_space.id)
+    )
+    document = result.scalars().first()
+
+    assert document.content == "Mocked summary."
+
+
+@pytest.mark.usefixtures(
+    "patched_summarize", "patched_embed_text", "patched_chunk_text"
+)
+async def test_chunks_written_to_db(db_session, db_search_space, db_user, mocker):
+    """Chunks derived from the source markdown are persisted in the DB."""
+    await index_uploaded_file(
+        markdown_content="## Hello\n\nSome content.",
+        filename="test.pdf",
+        etl_service="UNSTRUCTURED",
+        search_space_id=db_search_space.id,
+        user_id=str(db_user.id),
+        session=db_session,
+        llm=mocker.Mock(),
+    )
+
+    result = await db_session.execute(
+        select(Document).filter(Document.search_space_id == db_search_space.id)
+    )
+    document = result.scalars().first()
+
+    chunks_result = await db_session.execute(
+        select(Chunk).filter(Chunk.document_id == document.id)
+    )
+    chunks = chunks_result.scalars().all()
+
+    assert len(chunks) == 1
+    assert chunks[0].content == "Test chunk content."
+
+
+@pytest.mark.usefixtures(
+    "patched_summarize_raises", "patched_embed_text", "patched_chunk_text"
+)
+async def test_raises_on_indexing_failure(db_session, db_search_space, db_user, mocker):
+    """RuntimeError is raised when the indexing step fails so the caller can fire a failure notification."""
+    with pytest.raises(RuntimeError):
+        await index_uploaded_file(
+            markdown_content="## Hello\n\nSome content.",
+            filename="test.pdf",
+            etl_service="UNSTRUCTURED",
+            search_space_id=db_search_space.id,
+            user_id=str(db_user.id),
+            session=db_session,
+            llm=mocker.Mock(),
+        )

surfsense_backend/tests/integration/indexing_pipeline/test_index_document.py

@@ -0,0 +1,341 @@
+import pytest
+from sqlalchemy import select
+
+from app.config import config as app_config
+from app.db import Chunk, Document, DocumentStatus
+from app.indexing_pipeline.indexing_pipeline_service import IndexingPipelineService
+
+_EMBEDDING_DIM = app_config.embedding_model_instance.dimension
+
+pytestmark = pytest.mark.integration
+
+
+@pytest.mark.usefixtures(
+    "patched_summarize", "patched_embed_text", "patched_chunk_text"
+)
+async def test_sets_status_ready(
+    db_session,
+    db_search_space,
+    make_connector_document,
+    mocker,
+):
+    """Document status is READY after successful indexing."""
+    connector_doc = make_connector_document(search_space_id=db_search_space.id)
+    service = IndexingPipelineService(session=db_session)
+
+    prepared = await service.prepare_for_indexing([connector_doc])
+    document = prepared[0]
+    document_id = document.id
+
+    await service.index(document, connector_doc, llm=mocker.Mock())
+
+    result = await db_session.execute(
+        select(Document).filter(Document.id == document_id)
+    )
+    reloaded = result.scalars().first()
+
+    assert DocumentStatus.is_state(reloaded.status, DocumentStatus.READY)
+
+
+@pytest.mark.usefixtures(
+    "patched_summarize", "patched_embed_text", "patched_chunk_text"
+)
+async def test_content_is_summary_when_should_summarize_true(
+    db_session,
+    db_search_space,
+    make_connector_document,
+    mocker,
+):
+    """Document content is set to the LLM-generated summary when should_summarize=True."""
+    connector_doc = make_connector_document(search_space_id=db_search_space.id)
+    service = IndexingPipelineService(session=db_session)
+
+    prepared = await service.prepare_for_indexing([connector_doc])
+    document = prepared[0]
+    document_id = document.id
+
+    await service.index(document, connector_doc, llm=mocker.Mock())
+
+    result = await db_session.execute(
+        select(Document).filter(Document.id == document_id)
+    )
+    reloaded = result.scalars().first()
+
+    assert reloaded.content == "Mocked summary."
+
+
+@pytest.mark.usefixtures(
+    "patched_summarize", "patched_embed_text", "patched_chunk_text"
+)
+async def test_content_is_source_markdown_when_should_summarize_false(
+    db_session,
+    db_search_space,
+    make_connector_document,
+):
+    """Document content is set to source_markdown verbatim when should_summarize=False."""
+    connector_doc = make_connector_document(
+        search_space_id=db_search_space.id,
+        should_summarize=False,
+        source_markdown="## Raw content",
+    )
+    service = IndexingPipelineService(session=db_session)
+
+    prepared = await service.prepare_for_indexing([connector_doc])
+    document = prepared[0]
+    document_id = document.id
+
+    await service.index(document, connector_doc, llm=None)
+
+    result = await db_session.execute(
+        select(Document).filter(Document.id == document_id)
+    )
+    reloaded = result.scalars().first()
+
+    assert reloaded.content == "## Raw content"
+
+
+@pytest.mark.usefixtures(
+    "patched_summarize", "patched_embed_text", "patched_chunk_text"
+)
+async def test_chunks_written_to_db(
+    db_session,
+    db_search_space,
+    make_connector_document,
+    mocker,
+):
+    """Chunks derived from source_markdown are persisted in the DB."""
+    connector_doc = make_connector_document(search_space_id=db_search_space.id)
+    service = IndexingPipelineService(session=db_session)
+
+    prepared = await service.prepare_for_indexing([connector_doc])
+    document = prepared[0]
+    document_id = document.id
+
+    await service.index(document, connector_doc, llm=mocker.Mock())
+
+    result = await db_session.execute(
+        select(Chunk).filter(Chunk.document_id == document_id)
+    )
+    chunks = result.scalars().all()
+
+    assert len(chunks) == 1
+    assert chunks[0].content == "Test chunk content."
+
+
+@pytest.mark.usefixtures(
+    "patched_summarize", "patched_embed_text", "patched_chunk_text"
+)
+async def test_embedding_written_to_db(
+    db_session,
+    db_search_space,
+    make_connector_document,
+    mocker,
+):
+    """Document embedding vector is persisted in the DB after indexing."""
+    connector_doc = make_connector_document(search_space_id=db_search_space.id)
+    service = IndexingPipelineService(session=db_session)
+
+    prepared = await service.prepare_for_indexing([connector_doc])
+    document = prepared[0]
+    document_id = document.id
+
+    await service.index(document, connector_doc, llm=mocker.Mock())
+
+    result = await db_session.execute(
+        select(Document).filter(Document.id == document_id)
+    )
+    reloaded = result.scalars().first()
+
+    assert reloaded.embedding is not None
+    assert len(reloaded.embedding) == _EMBEDDING_DIM
+
+
+@pytest.mark.usefixtures(
+    "patched_summarize", "patched_embed_text", "patched_chunk_text"
+)
+async def test_updated_at_advances_after_indexing(
+    db_session,
+    db_search_space,
+    make_connector_document,
+    mocker,
+):
+    """updated_at timestamp is later after indexing than it was at prepare time."""
+    connector_doc = make_connector_document(search_space_id=db_search_space.id)
+    service = IndexingPipelineService(session=db_session)
+
+    prepared = await service.prepare_for_indexing([connector_doc])
+    document = prepared[0]
+    document_id = document.id
+
+    result = await db_session.execute(
+        select(Document).filter(Document.id == document_id)
+    )
+    updated_at_pending = result.scalars().first().updated_at
+
+    await service.index(document, connector_doc, llm=mocker.Mock())
+
+    result = await db_session.execute(
+        select(Document).filter(Document.id == document_id)
+    )
+    updated_at_ready = result.scalars().first().updated_at
+
+    assert updated_at_ready > updated_at_pending
+
+
+@pytest.mark.usefixtures(
+    "patched_summarize", "patched_embed_text", "patched_chunk_text"
+)
+async def test_no_llm_falls_back_to_source_markdown(
+    db_session,
+    db_search_space,
+    make_connector_document,
+):
+    """When llm=None and no fallback_summary, content falls back to source_markdown."""
+    connector_doc = make_connector_document(
+        search_space_id=db_search_space.id,
+        should_summarize=True,
+        source_markdown="## Fallback content",
+    )
+    service = IndexingPipelineService(session=db_session)
+
+    prepared = await service.prepare_for_indexing([connector_doc])
+    document = prepared[0]
+    document_id = document.id
+
+    await service.index(document, connector_doc, llm=None)
+
+    result = await db_session.execute(
+        select(Document).filter(Document.id == document_id)
+    )
+    reloaded = result.scalars().first()
+
+    assert DocumentStatus.is_state(reloaded.status, DocumentStatus.READY)
+    assert reloaded.content == "## Fallback content"
+
+
+@pytest.mark.usefixtures(
+    "patched_summarize", "patched_embed_text", "patched_chunk_text"
+)
+async def test_fallback_summary_used_when_llm_unavailable(
+    db_session,
+    db_search_space,
+    make_connector_document,
+):
+    """fallback_summary is used as content when llm=None and should_summarize=True."""
+    connector_doc = make_connector_document(
+        search_space_id=db_search_space.id,
+        should_summarize=True,
+        source_markdown="## Full raw content",
+        fallback_summary="Short pre-built summary.",
+    )
+    service = IndexingPipelineService(session=db_session)
+
+    prepared = await service.prepare_for_indexing([connector_doc])
+    document_id = prepared[0].id
+
+    await service.index(prepared[0], connector_doc, llm=None)
+
+    result = await db_session.execute(
+        select(Document).filter(Document.id == document_id)
+    )
+    reloaded = result.scalars().first()
+
+    assert DocumentStatus.is_state(reloaded.status, DocumentStatus.READY)
+    assert reloaded.content == "Short pre-built summary."
+
+
+@pytest.mark.usefixtures(
+    "patched_summarize", "patched_embed_text", "patched_chunk_text"
+)
+async def test_reindex_replaces_old_chunks(
+    db_session,
+    db_search_space,
+    make_connector_document,
+    mocker,
+):
+    """Re-indexing a document replaces its old chunks rather than appending."""
+    connector_doc = make_connector_document(
+        search_space_id=db_search_space.id,
+        source_markdown="## v1",
+    )
+    service = IndexingPipelineService(session=db_session)
+
+    prepared = await service.prepare_for_indexing([connector_doc])
+    document = prepared[0]
+    document_id = document.id
+
+    await service.index(document, connector_doc, llm=mocker.Mock())
+
+    updated_doc = make_connector_document(
+        search_space_id=db_search_space.id,
+        source_markdown="## v2",
+    )
+    re_prepared = await service.prepare_for_indexing([updated_doc])
+    await service.index(re_prepared[0], updated_doc, llm=mocker.Mock())
+
+    result = await db_session.execute(
+        select(Chunk).filter(Chunk.document_id == document_id)
+    )
+    chunks = result.scalars().all()
+
+    assert len(chunks) == 1
+
+
+@pytest.mark.usefixtures(
+    "patched_summarize_raises", "patched_embed_text", "patched_chunk_text"
+)
+async def test_llm_error_sets_status_failed(
+    db_session,
+    db_search_space,
+    make_connector_document,
+    mocker,
+):
+    """Document status is FAILED when the LLM raises during indexing."""
+    connector_doc = make_connector_document(search_space_id=db_search_space.id)
+    service = IndexingPipelineService(session=db_session)
+
+    prepared = await service.prepare_for_indexing([connector_doc])
+    document = prepared[0]
+    document_id = document.id
+
+    await service.index(document, connector_doc, llm=mocker.Mock())
+
+    result = await db_session.execute(
+        select(Document).filter(Document.id == document_id)
+    )
+    reloaded = result.scalars().first()
+
+    assert DocumentStatus.is_state(reloaded.status, DocumentStatus.FAILED)
+
+
+@pytest.mark.usefixtures(
+    "patched_summarize_raises", "patched_embed_text", "patched_chunk_text"
+)
+async def test_llm_error_leaves_no_partial_data(
+    db_session,
+    db_search_space,
+    make_connector_document,
+    mocker,
+):
+    """A failed indexing attempt leaves no partial embedding or chunks in the DB."""
+    connector_doc = make_connector_document(search_space_id=db_search_space.id)
+    service = IndexingPipelineService(session=db_session)
+
+    prepared = await service.prepare_for_indexing([connector_doc])
+    document = prepared[0]
+    document_id = document.id
+
+    await service.index(document, connector_doc, llm=mocker.Mock())
+
+    result = await db_session.execute(
+        select(Document).filter(Document.id == document_id)
+    )
+    reloaded = result.scalars().first()
+
+    assert reloaded.embedding is None
+    assert reloaded.content == "Pending..."
+
+    chunks_result = await db_session.execute(
+        select(Chunk).filter(Chunk.document_id == document_id)
+    )
+    assert chunks_result.scalars().all() == []