jeevesagent.memory.auto_extract

Auto-extract wrapper — runs the Consolidator on every remembered episode so the bot extracts and stores structured facts automatically as conversations happen.

The wrapper is what turns Agent(memory="sqlite:./bot.db") into a “my bot just remembers things” experience: the user says “I prefer dark mode” and a turn later the framework has a Fact(subject="alice", predicate="prefers", object="dark_mode") in its store, partitioned by user_id, ready to surface in future runs via recall_facts.

Wiring:

• Wraps any Memory whose .facts is not None.

• On every remember(episode) call: writes the episode through, then runs the configured Consolidator on JUST that episode (single-episode batch), letting the consolidator append any extracted facts to inner.facts.

• Extraction is best-effort: a failing extract (model error, malformed JSON, rate limit) NEVER breaks the run. The wrapper logs and moves on; the underlying episode write already succeeded.

• Every other Memory protocol method forwards straight through to the inner backend.

The Agent builds one of these automatically when auto_extract=True (the default) and the resolved memory has a fact store. Users who want the today’s behaviour pass auto_extract=False to Agent(...) — the wrapper simply isn’t applied.

Classes

AutoExtractMemory

Wraps a Memory and runs auto fact extraction on every

Module Contents

class jeevesagent.memory.auto_extract.AutoExtractMemory(inner: jeevesagent.core.protocols.Memory, consolidator: jeevesagent.memory.consolidator.Consolidator, *, on_extract_error: collections.abc.Callable[[BaseException], collections.abc.Awaitable[None]] | None = None, telemetry: jeevesagent.core.protocols.Telemetry | None = None, auto_picked: bool = False)

Wraps a Memory and runs auto fact extraction on every remember call.

Construct via the Agent auto_extract= kwarg; this class isn’t normally instantiated by user code. The wrapped memory must expose a .facts attribute (a FactStore) for extraction to do anything — when inner.facts is None, the wrapper still installs cleanly but every extraction is a no-op.

async append_block(name: str, content: str, *, user_id: str | None = None) → None

async consolidate() → None

async export(*, user_id: str | None = None) → jeevesagent.core.types.MemoryExport

async forget(*, user_id: str | None = None, session_id: str | None = None, before: datetime.datetime | None = None) → int

async profile(*, user_id: str | None = None) → jeevesagent.core.types.MemoryProfile

async recall(query: str, *, kind: str = 'episodic', limit: int = 5, time_range: tuple[datetime.datetime, datetime.datetime] | None = None, user_id: str | None = None) → list[jeevesagent.core.types.Episode]

async recall_facts(query: str, *, limit: int = 5, valid_at: datetime.datetime | None = None, user_id: str | None = None) → list[jeevesagent.core.types.Fact]

async remember(episode: jeevesagent.core.types.Episode) → str

Persist the episode, then run auto-extraction.

The episode write happens first and is the contract — the function returns its id even when extraction fails. So the consolidator’s fragility never leaks into the agent’s own durability guarantees.

async session_messages(session_id: str, *, user_id: str | None = None, limit: int = 20) → list[jeevesagent.core.types.Message]

async update_block(name: str, content: str, *, user_id: str | None = None) → None

async working(*, user_id: str | None = None) → list[jeevesagent.core.types.MemoryBlock]

property facts: Any

Forward the inner backend’s fact store. Reading this gives callers the same access to the bi-temporal store the consolidator writes into.

property inner: jeevesagent.core.protocols.Memory

The wrapped backend. Power-user introspection — most call sites just use the protocol methods.