ingest_source()
KeyStores a raw source, chunks it, embeds it, and FTS-indexes it. Returns the source_id and chunk summaries. Extraction is Claude's job afterward — the server stores honestly, it does not interpret.
ingest_source(content, title, source_type)→ SOURCE_ID + SUMMARIESCallable over MCP as ingest_source(content, title, source_type). REST mirror POST /v1/sources Planned — alpha is MCP-only
When something worth remembering arrives — a call transcript, a pricing review, a board memo — the agent stores it raw. The server chunks, embeds, and indexes; the agent then extracts the facts and beliefs worth proposing. Storage and interpretation stay separate on purpose.
Over MCP, your agent invokes it on its own when the moment calls for it — just describe the situation. A Python SDK for headless runtimes is planned; this is what the call will look like.
| Param | Type | Description |
|---|---|---|
| contentREQUIRED | string | The raw source text to store. |
| title | string | Human-readable source title. |
| source_type | string | e.g. "note", "document", "transcript". |

