memo-mcp verify
memo-mcp verify --repair
| Check | Problem reported | --repair does |
|---|
| Every live document has passages | N live document(s) have no chunks | Reports only; re-ingest the source |
| Vectors point at existing passages | N vector(s) point at missing chunks | Reports only |
| Facts point at existing evidence | N fact(s) point at missing evidence chunks | Reports only |
| Every live passage has a vector for the current model | N live chunk(s) have no vector for model X | Queues an embed job; run memo-mcp backfill |
| FTS5 index integrity | <index> failed integrity-check | Rebuilds the index |
The command prints ok: no problems found, or one problem: line per finding and one
repaired: line per fix.
| Command | Embeds | When |
|---|
memo-mcp backfill | Passages queued for vectors, from --embed=false, a degraded write or verify --repair | After bulk loads, or after the model was unavailable |
memo-mcp reindex [--model id] | Every live passage lacking a vector for the model | After switching models |
Both are resumable and safe to interrupt. The server runs both in the background at start.
sqlite3 ~/.memo-mcp/kb/my-project.db 'PRAGMA integrity_check' # expect: ok
sqlite3 ~/.memo-mcp/kb/my-project.db 'PRAGMA user_version' # schema version
| After… | Run |
|---|
| A crash or power loss | verify, then PRAGMA integrity_check |
| Restoring a backup | verify |
Bulk ingest with --embed=false | backfill |
Changing MEMO_MODEL | reindex --model <id> |
| Upgrading across 1.0 to 1.1 | graph rebuild |
| Large redactions | VACUUM with no process attached |