life-memory-media-ingestion

Life Memory Media Ingestion

Use this skill when an agent needs to submit a URL, file, image, spreadsheet, PDF, document, audio, or video into the Life Memory System.

The rule is simple: media is a source artifact first, not a permanent fact. Create a media artifact, let the system extract source evidence, then inspect the artifact and governance status.

Required MCP Tools

Confirm these tools exist before submitting media:

• memory_create_link_artifact
• memory_upload_media_base64
• memory_list_media_artifacts
• memory_get_media_artifact
• memory_extract_media_artifact

If any tool is missing, call memory_map if available, refresh the MCP connection, and report that the client is connected to an older Life Memory MCP server.

Link Workflow

Use memory_create_link_artifact for public http or https links.

Required fields:

• url: the public URL.
• source_channel: use mcp unless the caller has a more specific channel.

Optional fields:

• source_text: short context from the user or task.
• extract: default true.
• sync: default false.

Do not fetch the URL yourself and paste the webpage into memory_after_end. The system handles SSRF checks, extraction, failure status, and RawEvent/work-case evidence creation.

Example:

{
  "url": "https://example.com/article",
  "source_text": "User asked me to preserve this article as a note source.",
  "source_channel": "mcp",
  "extract": true,
  "sync": false
}

File Workflow

Use memory_upload_media_base64 for files, images, spreadsheets, documents, audio, and video.

Required fields:

• filename: original filename or a safe descriptive name.
• content_base64: base64 file bytes.

Recommended fields:

• mime_type: explicit MIME type when known.
• media_type: one of image, spreadsheet, pdf, document, audio, video, or file.
• source_channel: use mcp.
• extract: default true.
• sync: default false.

Example:

{
  "filename": "meeting-notes.csv",
  "mime_type": "text/csv",
  "media_type": "spreadsheet",
  "source_channel": "mcp",
  "content_base64": "<base64>",
  "extract": true,
  "sync": false
}