Memory

Extract durable facts from a conversation (Mem0-parity)

Accepts a raw multi-turn conversation and runs LLM fact extraction + kNN-dedup + create/update/archive against the contexts store. Conversations longer than MEMORY_ASYNC_THRESHOLD messages (default 8) run asynchronously — the response is { jobId, statusUrl }, and the caller polls GET /api/ingest-jobs/{jobId}. Shorter conversations return the full result inline.

post/api/memory

Request

  • Base URL: http://localhost:38200
  • URL: http://localhost:38200/api/memory
  • Auth: one of:
    • HTTP bearer
    • API key in cookie token

Request body

subjectIdstring
agentSlugstring
workspacestring
projectstring
sessionIdstring
dryRunboolean

Preview extraction without writing to the store.

maxClaimsinteger
asyncboolean

Force the async (true) or sync (false) path. Omit to let the server auto-pick.

Response

Extraction ran synchronously; full result returned inline.

object required

Changes

No recorded changes to this operation across all 1 revision of this API.