memory-ingestlisted

# memory-ingest For bulk import or one-off file ingestion, use the shell command rather than an MCP tool: ```bash memory ingest --folder ~/Documents/old-notes/ --dry-run # preview memory ingest --folder ~/Documents/old-notes/ # actually write memory ingest --file ~/path/to/single-note.md ``` ## What the kit does for each file 1. **Generates a stable ID** from the file path (SHA-1 prefix) 2. **Wraps with default frontmatter** if missing: - `type: observation` (safe default; user can promote later) - `entities: []` (you'll want to add wikilinks after ingest) - `tags: [imported]` - `importance: 0.5` - `created: <today>` 3. **Lints** the new file — surfaces dead wikilinks, missing fields ## After bulk import — what to do next The mass-imported memories will be low-signal until you add entity wikilinks. Recommend the user run `memory-refresh`-style sweeps over them: - Read each one - Identify the people / companies / projects mentioned - Add `[[Wikilinks]]` to the entities array OR use the LLM-assisted entity tagger (if available) to do this in batch: ```bash memory tag-entities --since "2026-01-01" # not yet implemented ``` ## When NOT to use - For a single memory generated mid-conversation → use `memory-save` instead - For ingesting from a connected MCP source → use `memory-refresh` instead