Advanced / Payload builder
Importance & Ranking
The fields you send with a write decide how findable and maintainable the memory is later. Build the payload deliberately instead of sending bare text.
Builder
Build the write payload
POST /memory/add
{
"text": "The user prefers concise answers and dark mode.",
"source": "chat",
"tags": ["preference","ui"],
"importance": 0.80,
"metadata": { "conversation_id": "conv-42" }
}Scope headers travel separately: the API key identifies the organization, and X-Project-ID plus X-End-User-ID select the scope this memory belongs to.
Reference
What each field buys you
- text
- The fact itself. Short, self-contained statements survive better than transcripts.
- source
- Where it came from. Useful for auditing and for cleaning up one integration later.
- tags
- The one field you can filter on at query time, so choose a small controlled vocabulary.
- importance
- A 0–1 hint you supply. Treat it as your own priority signal, and keep the scale consistent.
- metadata
- Your own identifiers for joining a memory back to your data. Never put secrets here.
- session_id
- Groups related writes and reads for one conversation.
Copy and run
Copy a working example
ms.add("The user prefers concise answers and dark mode.",source="chat",tags=["preference", "ui"],importance=0.8,session_id="conv-42",metadata={"conversation_id": "conv-42"},)
Was this page helpful?