Cli surface

CLI surface¶

All commands live under codeclone memory and accept --root (default .).

Command	Purpose
`init [--refresh] [--dry-run] [--from-report PATH] [--no-docs] [--no-tests]`	Create or refresh the memory store
`status`	Schema version, counts, last ingest metadata
`for-path PATH [--limit N]`	Records linked to a repo-relative path
`search QUERY [--match any\\|all] [--semantic] [--active-only] [--limit N]`	FTS search; optional semantic blend
`semantic status`	Index availability, provider, row counts
`semantic rebuild`	Rebuild LanceDB sidecar from memory + audit + trajectory
`semantic probe [--exact-tokens] [--json]`	Projection length stats (index-unit aware for trajectory)
`semantic search QUERY [--limit N]`	Search with semantic ranking (index required)
`stale [--limit N]`	List stale records and reasons
`vacuum`	Retention purge per config (no dry-run flag)
`coverage --scope PATH [PATH...]`	Scope coverage metrics
`review-candidates [--limit N]`	List draft records awaiting human review
`approve RECORD_ID [--by NAME] [--i-know-what-im-doing]`	Promote draft → active
`reject RECORD_ID [--by NAME] [--reason TEXT] [--i-know-what-im-doing]`	Reject draft
`archive RECORD_ID [--by NAME] [--i-know-what-im-doing]`	Archive record
`trajectory status\\|rebuild\\|list\\|search\\|show\\|agents\\|anomalies\\|dashboard\\|export`	Trajectory projection, passport analytics, and export
`jobs status\\|enqueue\\|run-once\\|list`	Trajectory + semantic + Experience projection queue

Init flags¶

Flag	Effect
`--dry-run`	Build ingest batch without writing the store
`--refresh`	Re-ingest and run staleness pass on drifted system records
`--from-report`	Load a canonical report JSON path instead of cached/latest analysis
`--no-docs`	Skip document-link ingest
`--no-tests`	Skip test-anchor ingest

Governance (`approve` / `reject` / `archive`)¶

Direct CLI governance is disabled by default. The preferred path is the CodeClone VS Code Memory view (IDE governance channel over MCP with --ide-governance-channel).

For explicit human break-glass outside the IDE channel, pass --i-know-what-im-doing on approve, reject, or archive. Attributions use --by (default human), not --verified-by.

MCP agents cannot call approve/reject/archive on manage_engineering_memory.

Trajectory analytics flags¶

Subcommand	Extra flags
`list`	`--limit N`
`search`	`QUERY`, `--limit N`, `--match any\\|all`
`agents`	`--include-routine`, `--json`
`anomalies`	`--limit N`, `--include-routine`, `--json`
`dashboard`	`--limit N`, `--include-routine`, `--json`
`export`	`--profile NAME`, `--out PATH`, `--allow-external-out`, `--force`, `--json`

--include-routine includes routine analysis-only trajectories in aggregates (default excludes them). Maps to MCP filters.include_routine on trajectory analytics modes.

Projection jobs flags¶

Subcommand	Extra flags
`enqueue`	`--force` (enqueue even when policy off or stimulus unchanged), `--no-spawn`
`run-once`	`--not-before ISO-8601-UTC` (defer until coalesce window elapses)
`list`	`--limit N`, `--json`

Refs:

codeclone/surfaces/cli/memory.py
codeclone/surfaces/cli/memory_render.py