SurfSense/surfsense_backend/app/agents/shared/context.py
CREDO23 28b13ed25b refactor(agents): move context schema + state reducers to app/agents/shared (slice 2)
Continue promoting the shared agent toolkit out of `new_chat` into the
cross-agent `app/agents/shared` kernel.

- state_reducers.py: clean move (no single-agent importer); all 7 importers
  flipped to app.agents.shared.state_reducers.
- context.py: moved to app.agents.shared.context; flipped the multi-agent,
  app, automations, chat-flows and monolith importers. A thin re-export shim
  remains at new_chat/context.py because the not-yet-retired single-agent
  (chat_deepagent) and the new_chat package __init__ still import it; the shim
  goes away with the single-agent deletion.
- Updated the stream parity test's annotation normalizer to strip the new
  app.agents.shared.context. prefix (SurfSenseContextSchema.__module__ changed
  with the move), keeping monolith<->flows signature parity intact.

Behavior-preserving: definitions unchanged; only import paths move. 1219 tests green.
2026-06-04 12:19:22 +02:00

71 lines
3 KiB
Python

"""
Context schema definitions for SurfSense agents.
This module defines the per-invocation context object passed to the SurfSense
deep agent via ``agent.astream_events(..., context=ctx)`` (LangGraph >= 0.6).
The agent's compiled graph is the same across invocations (and cached by
``agent_cache``), so anything that varies per turn — the user mentions a
specific document, the front-end issues a unique ``request_id``, etc. —
MUST live on this context object instead of being captured into a
middleware ``__init__`` closure. Middlewares read fields back via
``runtime.context.<field>``; tools read them via ``runtime.context``.
This object is read inside both ``KnowledgePriorityMiddleware`` (for
``mentioned_document_ids``) and any future middleware that needs
per-request state without invalidating the compiled-agent cache.
"""
from __future__ import annotations
from dataclasses import dataclass, field
from typing import TypedDict
class FileOperationContractState(TypedDict):
intent: str
confidence: float
suggested_path: str
timestamp: str
turn_id: str
@dataclass
class SurfSenseContextSchema:
"""
Per-invocation context for the SurfSense deep agent.
Defaults are chosen so the dataclass can be safely default-constructed
(LangGraph's ``Runtime.context`` itself defaults to ``None`` if no
context is supplied — see ``langgraph.runtime.Runtime``). All fields
are optional; consumers must None-check before reading.
Phase 1.5 fields:
search_space_id: Search space the request is scoped to.
mentioned_document_ids: KB documents the user @-mentioned this turn.
Read by ``KnowledgePriorityMiddleware`` to seed its priority
list. Stays out of the compiled-agent cache key — that's the
whole point of putting it here.
mentioned_folder_ids: KB folders the user @-mentioned this turn
(cloud filesystem mode). Surfaced as ``[USER-MENTIONED]``
entries in ``<priority_documents>`` so the agent prioritises
walking those folders with ``ls`` / ``find_documents``.
file_operation_contract: One-shot file operation contract emitted
by ``FileIntentMiddleware`` for the upcoming turn.
turn_id / request_id: Correlation IDs surfaced by the streaming
task; populated for telemetry.
Phase 2 will extend with: thread_id, user_id, visibility,
filesystem_mode, anon_session_id, available_connectors,
available_document_types, created_by_id (everything currently captured
by middleware ``__init__`` closures).
"""
search_space_id: int | None = None
mentioned_document_ids: list[int] = field(default_factory=list)
mentioned_folder_ids: list[int] = field(default_factory=list)
mentioned_connector_ids: list[int] = field(default_factory=list)
mentioned_connectors: list[dict[str, object]] = field(default_factory=list)
file_operation_contract: FileOperationContractState | None = None
turn_id: str | None = None
request_id: str | None = None