the cli¶
commands¶
bower curate <text>¶
store a memory. when GEMINI_API_KEY is set, the text is processed by an LLM that extracts structured facts, patterns, decisions, and procedures. without a key, it's stored as a single note.
bower query <text>¶
hybrid search with automatic Tier 2 → Tier 3 escalation.
| flag | effect |
|---|---|
--limit N |
max results (default 10, max 50) |
--fast |
BM25 only, no vector embeddings |
--deep |
force Tier 3 LLM re-rank + synthesis |
--no-cache |
bypass all cache tiers for this call |
--json |
structured JSON output |
bower query "how do we handle authentication"
bower query --deep "explain the auth architecture end to end"
bower query --fast --limit 3 "redis version"
bower tree [prefix]¶
walk the on-disk markdown tree.
bower ls <path>¶
list a single subtree level.
bower mv <id> <path>¶
move a memory to a different folder. preserves all metadata and relations.
bower supersede <old-id> <new-id>¶
mark an old memory as superseded by a new one. retrieval will prefer the new one but the old is preserved for historical lookups.
bower export <path>¶
export a subtree to stdout as a single concatenated markdown stream. useful for git-style review or piping to another tool.
bower import <file>¶
inverse of export.
bower dream¶
offline analysis: link discovery, deduplication, maturity promotion, contradiction detection.
| sub-command | effect |
|---|---|
dream |
run full pipeline, auto-apply high-confidence changes |
dream-scan |
dry-run, emit JSON of candidate changes |
dream-apply |
apply pre-scanned candidates |
dream-undo |
roll back the last dream pass |
bower reindex¶
force a full rebuild of the sqlite cache from the on-disk tree. usually unnecessary because reconciliation runs lazily; useful if you've batch-edited files outside of bower and want eager pickup.
bower status¶
memory count, db size, index health, recent query stats, cache hit ratios.
bower serve [--port N]¶
run as a local HTTP daemon. see http api.
bower settings¶
configuration management.