feat: implement agent caches and fix invalid prompt cache configs
Some checks are pending
Build and Push Docker Images / tag_release (push) Waiting to run
Build and Push Docker Images / build (./surfsense_backend, ./surfsense_backend/Dockerfile, backend, surfsense-backend, ubuntu-24.04-arm, linux/arm64, arm64) (push) Blocked by required conditions
Build and Push Docker Images / build (./surfsense_backend, ./surfsense_backend/Dockerfile, backend, surfsense-backend, ubuntu-latest, linux/amd64, amd64) (push) Blocked by required conditions
Build and Push Docker Images / build (./surfsense_web, ./surfsense_web/Dockerfile, web, surfsense-web, ubuntu-24.04-arm, linux/arm64, arm64) (push) Blocked by required conditions
Build and Push Docker Images / build (./surfsense_web, ./surfsense_web/Dockerfile, web, surfsense-web, ubuntu-latest, linux/amd64, amd64) (push) Blocked by required conditions
Build and Push Docker Images / create_manifest (backend, surfsense-backend) (push) Blocked by required conditions
Build and Push Docker Images / create_manifest (web, surfsense-web) (push) Blocked by required conditions

- Added a new function `_warm_agent_jit_caches` to pre-warm agent caches at startup, reducing cold invocation costs.
- Updated the `SurfSenseContextSchema` to include per-invocation fields for better state management during agent execution.
- Introduced caching mechanisms in various tools to ensure fresh database sessions are used, improving performance and reliability.
- Enhanced middleware to support new context features and improve error handling during connector and document type discovery.
This commit is contained in:
DESKTOP-RTLN3BA\$punk 2026-05-03 06:03:40 -07:00
parent 90a653c8c7
commit a34f1fb25c
60 changed files with 8477 additions and 5381 deletions

View file

@ -6,6 +6,7 @@ from langchain_core.tools import tool
from sqlalchemy.ext.asyncio import AsyncSession
from app.agents.new_chat.tools.hitl import request_approval
from app.db import async_session_maker
from ._auth import LUMA_API, get_api_key, get_luma_connector, luma_headers
@ -17,6 +18,23 @@ def create_create_luma_event_tool(
search_space_id: int | None = None,
user_id: str | None = None,
):
"""
Factory function to create the create_luma_event tool.
The tool acquires its own short-lived ``AsyncSession`` per call via
:data:`async_session_maker` so the closure is safe to share across
HTTP requests by the compiled-agent cache. Capturing a per-request
session here would surface stale/closed sessions on cache hits.
Args:
db_session: Reserved for registry compatibility. Per-call sessions
are opened via :data:`async_session_maker` inside the tool body.
Returns:
Configured create_luma_event tool
"""
del db_session # per-call session — see docstring
@tool
async def create_luma_event(
name: str,
@ -40,83 +58,86 @@ def create_create_luma_event_tool(
IMPORTANT:
- If status is "rejected", the user explicitly declined. Do NOT retry.
"""
if db_session is None or search_space_id is None or user_id is None:
if search_space_id is None or user_id is None:
return {"status": "error", "message": "Luma tool not properly configured."}
try:
connector = await get_luma_connector(db_session, search_space_id, user_id)
if not connector:
return {"status": "error", "message": "No Luma connector found."}
async with async_session_maker() as db_session:
connector = await get_luma_connector(
db_session, search_space_id, user_id
)
if not connector:
return {"status": "error", "message": "No Luma connector found."}
result = request_approval(
action_type="luma_create_event",
tool_name="create_luma_event",
params={
"name": name,
"start_at": start_at,
"end_at": end_at,
"description": description,
"timezone": timezone,
},
context={"connector_id": connector.id},
)
if result.rejected:
return {
"status": "rejected",
"message": "User declined. Event was not created.",
}
final_name = result.params.get("name", name)
final_start = result.params.get("start_at", start_at)
final_end = result.params.get("end_at", end_at)
final_desc = result.params.get("description", description)
final_tz = result.params.get("timezone", timezone)
api_key = get_api_key(connector)
headers = luma_headers(api_key)
body: dict[str, Any] = {
"name": final_name,
"start_at": final_start,
"end_at": final_end,
"timezone": final_tz,
}
if final_desc:
body["description_md"] = final_desc
async with httpx.AsyncClient(timeout=20.0) as client:
resp = await client.post(
f"{LUMA_API}/event/create",
headers=headers,
json=body,
result = request_approval(
action_type="luma_create_event",
tool_name="create_luma_event",
params={
"name": name,
"start_at": start_at,
"end_at": end_at,
"description": description,
"timezone": timezone,
},
context={"connector_id": connector.id},
)
if resp.status_code == 401:
return {
"status": "auth_error",
"message": "Luma API key is invalid.",
"connector_type": "luma",
}
if resp.status_code == 403:
return {
"status": "error",
"message": "Luma Plus subscription required to create events via API.",
}
if resp.status_code not in (200, 201):
return {
"status": "error",
"message": f"Luma API error: {resp.status_code}{resp.text[:200]}",
}
if result.rejected:
return {
"status": "rejected",
"message": "User declined. Event was not created.",
}
data = resp.json()
event_id = data.get("api_id") or data.get("event", {}).get("api_id")
final_name = result.params.get("name", name)
final_start = result.params.get("start_at", start_at)
final_end = result.params.get("end_at", end_at)
final_desc = result.params.get("description", description)
final_tz = result.params.get("timezone", timezone)
return {
"status": "success",
"event_id": event_id,
"message": f"Event '{final_name}' created on Luma.",
}
api_key = get_api_key(connector)
headers = luma_headers(api_key)
body: dict[str, Any] = {
"name": final_name,
"start_at": final_start,
"end_at": final_end,
"timezone": final_tz,
}
if final_desc:
body["description_md"] = final_desc
async with httpx.AsyncClient(timeout=20.0) as client:
resp = await client.post(
f"{LUMA_API}/event/create",
headers=headers,
json=body,
)
if resp.status_code == 401:
return {
"status": "auth_error",
"message": "Luma API key is invalid.",
"connector_type": "luma",
}
if resp.status_code == 403:
return {
"status": "error",
"message": "Luma Plus subscription required to create events via API.",
}
if resp.status_code not in (200, 201):
return {
"status": "error",
"message": f"Luma API error: {resp.status_code}{resp.text[:200]}",
}
data = resp.json()
event_id = data.get("api_id") or data.get("event", {}).get("api_id")
return {
"status": "success",
"event_id": event_id,
"message": f"Event '{final_name}' created on Luma.",
}
except Exception as e:
from langgraph.errors import GraphInterrupt

View file

@ -5,6 +5,8 @@ import httpx
from langchain_core.tools import tool
from sqlalchemy.ext.asyncio import AsyncSession
from app.db import async_session_maker
from ._auth import LUMA_API, get_api_key, get_luma_connector, luma_headers
logger = logging.getLogger(__name__)
@ -15,6 +17,23 @@ def create_list_luma_events_tool(
search_space_id: int | None = None,
user_id: str | None = None,
):
"""
Factory function to create the list_luma_events tool.
The tool acquires its own short-lived ``AsyncSession`` per call via
:data:`async_session_maker` so the closure is safe to share across
HTTP requests by the compiled-agent cache. Capturing a per-request
session here would surface stale/closed sessions on cache hits.
Args:
db_session: Reserved for registry compatibility. Per-call sessions
are opened via :data:`async_session_maker` inside the tool body.
Returns:
Configured list_luma_events tool
"""
del db_session # per-call session — see docstring
@tool
async def list_luma_events(
max_results: int = 25,
@ -28,77 +47,80 @@ def create_list_luma_events_tool(
Dictionary with status and a list of events including
event_id, name, start_at, end_at, location, url.
"""
if db_session is None or search_space_id is None or user_id is None:
if search_space_id is None or user_id is None:
return {"status": "error", "message": "Luma tool not properly configured."}
max_results = min(max_results, 50)
try:
connector = await get_luma_connector(db_session, search_space_id, user_id)
if not connector:
return {"status": "error", "message": "No Luma connector found."}
async with async_session_maker() as db_session:
connector = await get_luma_connector(
db_session, search_space_id, user_id
)
if not connector:
return {"status": "error", "message": "No Luma connector found."}
api_key = get_api_key(connector)
headers = luma_headers(api_key)
api_key = get_api_key(connector)
headers = luma_headers(api_key)
all_entries: list[dict] = []
cursor = None
all_entries: list[dict] = []
cursor = None
async with httpx.AsyncClient(timeout=20.0) as client:
while len(all_entries) < max_results:
params: dict[str, Any] = {
"limit": min(100, max_results - len(all_entries))
}
if cursor:
params["cursor"] = cursor
async with httpx.AsyncClient(timeout=20.0) as client:
while len(all_entries) < max_results:
params: dict[str, Any] = {
"limit": min(100, max_results - len(all_entries))
}
if cursor:
params["cursor"] = cursor
resp = await client.get(
f"{LUMA_API}/calendar/list-events",
headers=headers,
params=params,
resp = await client.get(
f"{LUMA_API}/calendar/list-events",
headers=headers,
params=params,
)
if resp.status_code == 401:
return {
"status": "auth_error",
"message": "Luma API key is invalid.",
"connector_type": "luma",
}
if resp.status_code != 200:
return {
"status": "error",
"message": f"Luma API error: {resp.status_code}",
}
data = resp.json()
entries = data.get("entries", [])
if not entries:
break
all_entries.extend(entries)
next_cursor = data.get("next_cursor")
if not next_cursor:
break
cursor = next_cursor
events = []
for entry in all_entries[:max_results]:
ev = entry.get("event", {})
geo = ev.get("geo_info", {})
events.append(
{
"event_id": entry.get("api_id"),
"name": ev.get("name", "Untitled"),
"start_at": ev.get("start_at", ""),
"end_at": ev.get("end_at", ""),
"timezone": ev.get("timezone", ""),
"location": geo.get("name", ""),
"url": ev.get("url", ""),
"visibility": ev.get("visibility", ""),
}
)
if resp.status_code == 401:
return {
"status": "auth_error",
"message": "Luma API key is invalid.",
"connector_type": "luma",
}
if resp.status_code != 200:
return {
"status": "error",
"message": f"Luma API error: {resp.status_code}",
}
data = resp.json()
entries = data.get("entries", [])
if not entries:
break
all_entries.extend(entries)
next_cursor = data.get("next_cursor")
if not next_cursor:
break
cursor = next_cursor
events = []
for entry in all_entries[:max_results]:
ev = entry.get("event", {})
geo = ev.get("geo_info", {})
events.append(
{
"event_id": entry.get("api_id"),
"name": ev.get("name", "Untitled"),
"start_at": ev.get("start_at", ""),
"end_at": ev.get("end_at", ""),
"timezone": ev.get("timezone", ""),
"location": geo.get("name", ""),
"url": ev.get("url", ""),
"visibility": ev.get("visibility", ""),
}
)
return {"status": "success", "events": events, "total": len(events)}
return {"status": "success", "events": events, "total": len(events)}
except Exception as e:
from langgraph.errors import GraphInterrupt

View file

@ -5,6 +5,8 @@ import httpx
from langchain_core.tools import tool
from sqlalchemy.ext.asyncio import AsyncSession
from app.db import async_session_maker
from ._auth import LUMA_API, get_api_key, get_luma_connector, luma_headers
logger = logging.getLogger(__name__)
@ -15,6 +17,23 @@ def create_read_luma_event_tool(
search_space_id: int | None = None,
user_id: str | None = None,
):
"""
Factory function to create the read_luma_event tool.
The tool acquires its own short-lived ``AsyncSession`` per call via
:data:`async_session_maker` so the closure is safe to share across
HTTP requests by the compiled-agent cache. Capturing a per-request
session here would surface stale/closed sessions on cache hits.
Args:
db_session: Reserved for registry compatibility. Per-call sessions
are opened via :data:`async_session_maker` inside the tool body.
Returns:
Configured read_luma_event tool
"""
del db_session # per-call session — see docstring
@tool
async def read_luma_event(event_id: str) -> dict[str, Any]:
"""Read detailed information about a specific Luma event.
@ -26,60 +45,63 @@ def create_read_luma_event_tool(
Dictionary with status and full event details including
description, attendees count, meeting URL.
"""
if db_session is None or search_space_id is None or user_id is None:
if search_space_id is None or user_id is None:
return {"status": "error", "message": "Luma tool not properly configured."}
try:
connector = await get_luma_connector(db_session, search_space_id, user_id)
if not connector:
return {"status": "error", "message": "No Luma connector found."}
api_key = get_api_key(connector)
headers = luma_headers(api_key)
async with httpx.AsyncClient(timeout=15.0) as client:
resp = await client.get(
f"{LUMA_API}/events/{event_id}",
headers=headers,
async with async_session_maker() as db_session:
connector = await get_luma_connector(
db_session, search_space_id, user_id
)
if not connector:
return {"status": "error", "message": "No Luma connector found."}
if resp.status_code == 401:
return {
"status": "auth_error",
"message": "Luma API key is invalid.",
"connector_type": "luma",
}
if resp.status_code == 404:
return {
"status": "not_found",
"message": f"Event '{event_id}' not found.",
}
if resp.status_code != 200:
return {
"status": "error",
"message": f"Luma API error: {resp.status_code}",
api_key = get_api_key(connector)
headers = luma_headers(api_key)
async with httpx.AsyncClient(timeout=15.0) as client:
resp = await client.get(
f"{LUMA_API}/events/{event_id}",
headers=headers,
)
if resp.status_code == 401:
return {
"status": "auth_error",
"message": "Luma API key is invalid.",
"connector_type": "luma",
}
if resp.status_code == 404:
return {
"status": "not_found",
"message": f"Event '{event_id}' not found.",
}
if resp.status_code != 200:
return {
"status": "error",
"message": f"Luma API error: {resp.status_code}",
}
data = resp.json()
ev = data.get("event", data)
geo = ev.get("geo_info", {})
event_detail = {
"event_id": event_id,
"name": ev.get("name", ""),
"description": ev.get("description", ""),
"start_at": ev.get("start_at", ""),
"end_at": ev.get("end_at", ""),
"timezone": ev.get("timezone", ""),
"location_name": geo.get("name", ""),
"address": geo.get("address", ""),
"url": ev.get("url", ""),
"meeting_url": ev.get("meeting_url", ""),
"visibility": ev.get("visibility", ""),
"cover_url": ev.get("cover_url", ""),
}
data = resp.json()
ev = data.get("event", data)
geo = ev.get("geo_info", {})
event_detail = {
"event_id": event_id,
"name": ev.get("name", ""),
"description": ev.get("description", ""),
"start_at": ev.get("start_at", ""),
"end_at": ev.get("end_at", ""),
"timezone": ev.get("timezone", ""),
"location_name": geo.get("name", ""),
"address": geo.get("address", ""),
"url": ev.get("url", ""),
"meeting_url": ev.get("meeting_url", ""),
"visibility": ev.get("visibility", ""),
"cover_url": ev.get("cover_url", ""),
}
return {"status": "success", "event": event_detail}
return {"status": "success", "event": event_detail}
except Exception as e:
from langgraph.errors import GraphInterrupt