Memory Types - Flumes AI

Flumes supports a small set of canonical memory types. Use these to store and filter structured knowledge.

Type	When it’s used	Example
`fact`	Declarative information believed to be true	”Paris is the capital of France.”
`event`	Time-stamped occurrence	”Chatted about Rome museums on 2024-07-01.”
`document`	Longer-form text or chunked content	An article or PDF excerpt
`note`	Freeform note	”Follow up about tickets.”
`profile`	Attributes about an entity	”ent:user_123 has premium=true.”
`chunk`	Arbitrary indexed snippet	Any raw passage

Filtering by type

curl -s "https://api.flumes.ai/v0/memories?type=fact&status=active&namespace=default&limit=20" \
  -H "Authorization: Bearer $FLUMES_API_KEY"

Use with tags, namespace, and time bounds (e.g., before/after) for precise dataset slices.

Structured examples

Fact

{
  "type": "fact",
  "subject": "ent:user_123",
  "predicate": "likes_city",
  "object_text": "Rome",
  "object_type": "string",
  "keys": ["ent:user_123|likes_city"],
  "confidence": 0.9,
  "status": "active"
}

Event

{
  "type": "event",
  "actor": "ent:user_123",
  "verb": "chatted",
  "target": "ent:assistant",
  "timestamp": "2024-07-01T12:34:56Z",
  "namespace": "default"
}