Episodic Memory

Who the node is. The id is the key the DOT pins.

id

episodic-memory

name

Episodic Memory

type

tool

phases

none declared

What it does, and the prose the agent is handed when the graph runs.

action29 words

Append each stage's inputs, outputs and decisions to the episode log as an immutable entry, and return the k most relevant episodes whenever a later step asks for context.

spec93 words

Append every event you are handed, a stage's inputs, its outputs and the decisions it took to the episode log as one new entry, stamped with its run id and stage id. Entries are append-only: never rewrite or remove one, because the log is the record of what each stage actually did. When a recall query arrives, answer on context with the five nearest episodes scoring above 0.75 similarity; when no query arrives, answer with the five most recent instead. Keep thirty runs of history and let anything older fall off the back.in full above

model

whatever the graph supplies

agent

not named

skill

skills/episodic-memory.md

tools

Vector store, File I/O

mcp

qdrant, filesystem

params

k: 5, retention_runs: 30, similarity_threshold: 0.75

What arrives, what leaves, which nodes it expects to hear from, and what may not.

inputs

event : json, query : text

outputs

context : json

dependencies

ingest-stage, transform-stage

cannot

no type is refused

will_not

rewrite or remove an entry

The keys the static analysis reads. Nothing here instructs the agent.

risk_markers

none

notes51 words

Entries are append-only, which is what makes the log double as an audit trail of exactly what each stage did. A recovery line reads it by recency, the newest valid entry, but k and similarity_threshold stay declared because the same store answers semantic recall for lines that query it that way.

The card's own version, and who wrote it.

version

1.0.0

author

autogen

provenance

not stated