CLI
Run from a checkout with ./auwai.
The same commands are available through python3 skills/auwai/scripts/auwai.py.
The installed package exposes the same interface as auwai.
auwai --version
Setup
| Command | Purpose |
|---|---|
init |
Print the interactive initialization brief |
init --check |
Report Python, SQLite/FTS5, paths, and existing state |
init --write |
Merge answers into config.toml and create the ledger |
init --write accepts --scope user|repo, --context-window,
--database-path, --journal-mode, --busy-timeout-ms, --allow-raw-read,
--cbm-auto, --cbm-bin, --hint, and --prime.
Coverage commands
| Command | Purpose |
|---|---|
index --repo PATH |
Build or rebuild the index and sync the ledger |
refresh --repo PATH |
Re-chunk changed files and sync the ledger |
survey --repo PATH |
Show scope, partition tree, and segment plan |
fullread plan |
Persist a full-read segment plan |
fullread next |
Show the next segment brief |
fullread read |
Return the segment bodies and mark the reads |
fullread done |
Close the active segment |
fullread status |
Show segment and coverage state |
fullread verify |
List unread first-party chunks |
query TERM |
Search the coverage index |
coverage |
Show coverage totals |
audit |
Check hashes against the working tree |
watch |
Native or polling change watch and ledger refresh |
watch --backend poll |
Force the portable polling backend |
guard |
Check the raw-read policy for a path or command |
Index and map commands
| Command | Purpose |
|---|---|
status --repo PATH |
Index state, providers, unread chunks |
map --repo PATH |
Print .auwai/map.md |
record ID |
Record a read and a finding |
recall TERM |
Restore durable findings |
delta --since REF |
Changed chunks and affected tests |
tests PATH |
Tests affected by a path |
doctor |
Verify language providers |
budget |
Resolve a read budget |
Session commands
The original task-scoped session remains:
./auwai plan "task" --write
./auwai begin --from-plan
./auwai next
./auwai read --next 3
./auwai recap
./auwai next-turn
This path uses the query budget. The full-read path uses the usable-context segment budget.
Common options
--repo PATH: repository root.--context-window N: model context.--fixed-prompt N: fixed prompt estimate for a full-read plan.--output-reserve N: answer and reasoning reserve.--safety N: safety margin.--synthesis none|final|partition: synthesis cadence.--plan ID: use a stored plan.--segment ID: use a stored segment.--json: machine-readable output.--limit N: query result limit.--once: one watcher scan.--interval SECONDS: watcher poll period.--repair: repair an audit drift.
Exit codes
0: success.1: incomplete coverage or a clean-check failure.2: usage, index, policy, or runtime error.
Related pages
Knowledge service
| Command | Purpose |
|---|---|
serve |
Run the optional shared facts service |
serve --check |
Report pending migrations without writing |
serve --migrate |
Apply pending migrations and exit |
serve --max-body-bytes N |
Set the request body limit |
serve --allow-anonymous |
Loopback-only anonymous mode |
sync --repo PATH |
Drain the local outbox |
sync --repo PATH --pull |
Pull shared facts |
facts build/search/stats |
Local deterministic facts layer |
fullread plan --share --jobs |
Register plans and full-read jobs |
fullread claim/read/done --remote |
Worker path over the service |
workspace fullread plan/next/read/done/status/verify |
Full read across a workspace |
serve requires AUWAI_KNOWLEDGE_TOKEN or --auth-file unless anonymous mode
is explicit. See knowledge service.