cm-memory

List all memories. Supports search and filtering.

View a single memory with all fields and screenshot.

Create a new memory.

Request

{
  "created_by": "mart",           // required
  "created_from": "manual",       // required — see valid values below
  "reference": "https://...",     // optional URL
  "text": "Some text",             // optional
  "images": ["<base64>"],  // optional — JSON array of base64-encoded images
  "audio": "<base64 m4a>", // optional — base64 audio recording
  "people": ["mart", "john"],     // optional — associated people
  "location": "45.50,-73.57",     // optional — GPS coords or place name
  "status": "pending",            // optional — for todos: pending/done/cancelled
  "notifications": ["2026-04-01T09:00:00Z"], // optional — reminder datetimes
  "memory_refs": ["uuid-1"]       // optional — linked memory IDs
}

Common created_from values (any string accepted)

phone_screenshot · phone_url · raindrop · watch · manual · web_clip · twitter · reddit · note · interaction · todo · iphone_status

Response

{ "id": "uuid", "created_at": "2025-01-01T00:00:00+00:00" }

Update one or more fields of a memory. Updatable fields: created_by, created_from, reference, text, ai_summary, ai_tags, ai_title, images, ai_processed, ai_processed_at, ai_error, tags, people, location, notifications, status, memory_refs, audio, ai_audio_transcription, ai_image_description.

Response

{ "updated": true, "id": "uuid" }

Permanently delete a memory.

Response

{ "deleted": true, "id": "uuid" }

Serve an image file attached to a memory.

Serve an audio file attached to a memory.

View pending memory count and error list.

AI-process up to limit unprocessed memories. Delegates to cm-ai-router for transcription, image description, and analysis. Generates ai_title, ai_summary, and ai_tags.

Request

{ "limit": 10 }  // optional, default 10, max 50

Response

{ "processed": 8, "errors": [{ "id": "uuid", "error": "..." }] }

Reset all memories with ai_error back to unprocessed so they can be retried.

Response

{ "reset": 3 }

Fetch new bookmarks from Raindrop.io API. Requires RAINDROP_TOKEN env var.

Response

{ "imported": 12 }

Scrape Reddit saved posts via cm-ai-router Playwright. Requires AI_ROUTER_URL and REDDIT_USERNAME. Browser session in ai-router must be logged in to Reddit.

Response

{ "imported": 5 }

Scrape X/Twitter bookmarks via cm-ai-router Playwright. Requires AI_ROUTER_URL. Browser session in ai-router must be logged in to Twitter/X.

Response

{ "imported": 7 }

Run a read-only SQL query against the memories table. Only SELECT statements allowed. Designed for LLMs and external tools to search and analyze memory data.

Request

{
  "sql": "SELECT id, ai_title, status FROM memories WHERE status = ?",
  "params": ["pending"]
}

Response

{
  "columns": ["id", "ai_title", "status"],
  "rows": [{"id": "uuid", "ai_title": "Buy groceries", "status": "pending"}],
  "count": 1
}

View request audit log.