CLI reference¶

All commands are invoked as auto-lorebook <command>. Each command links to the page that explains its semantics.

Ingest¶

auto-lorebook ingest <url-or-path> \
  [--source-url <url>] \
  [--source-id <id>] \
  [--session-date <YYYY-MM-DD>] \
  [--perspective <text>] \
  [--source-nature <kind>] \
  [--setting <name>] \
  [--no-interactive]

Fetch a source, gather context, and (by default) run the reading pipeline. Interactive prompts by default; flags skip their corresponding prompts. See context pipeline.

auto-lorebook configure-context <source_id>

Re-run context prompts for an existing source — fills in skipped fields or corrects mistakes.

Reading¶

auto-lorebook generate-reading <source_id>

Run Stage 1a → 1b on a source whose reading wasn't generated at ingest time. See Stage 1.

auto-lorebook regenerate-reading <source_id> --from={structure|summarize} \
  [--segments <id1,id2,...>]

Re-run from a given substage. --segments is valid only with --from=summarize.

auto-lorebook approve-reading <source_id>
auto-lorebook readings list
auto-lorebook readings show <source_id>

auto-lorebook plan <source_id>
auto-lorebook extract <source_id>

plan runs Stage 2 (planner): routes claim bullets to entities and writes pending/<id>/plan.yaml. Refuses to run unless the wiki-side reading.md exists (run approve-reading first). See Stage 2.

extract runs Stage 3 (extractor): locates verbatim transcript spans and writes proposal YAMLs under pending/<id>/proposals/. Refuses to run unless pending/<id>/plan.yaml exists (run plan first). See Stage 3.

Plans¶

auto-lorebook plans list
auto-lorebook plans show <ingest_id>
auto-lorebook replan <ingest_id>

Plans are intermediate — no approval gate. replan re-runs the planner and extractor on unreviewed proposals; approved proposals are unaffected. See Stage 2.

Review¶

auto-lorebook review <ingest_id>

Walk through proposals one at a time. Approve, edit, reject, or play each. See fact review.

Corrections¶

auto-lorebook promote-correction "<from>" "<to>"

Promote a per-source name correction to the global .transcription-corrections.yaml. See entity model.

Entities¶

auto-lorebook entities list [--category <cat>] [--created-by <ingest_id>]
auto-lorebook entities show <slug-or-name>
auto-lorebook entities new --category <cat> --name <name> [--slug <slug>]
auto-lorebook entities rebuild-index

list prints a category/name/slug/alias-count table; filters compose. show resolves the query against slugs first, then canonical names (case-insensitive), then aliases. new writes a minimum-viable stub; slug defaults to a slugified --name. rebuild-index is a placeholder until an on-disk cache exists — today the in-memory index is rebuilt on every command anyway. See hand-creating entities.

auto-lorebook reject-ingest <ingest_id>

Remove all facts, alias additions, and empty entity stubs created by a given ingest. See audit trail.

Wiki¶

auto-lorebook wiki list [--category <cat>]
auto-lorebook wiki show <entity>
auto-lorebook wiki rebuild

wiki rebuild regenerates all summaries from YAML. Skips files whose recorded inputs haven't changed; add --force to regenerate unconditionally. See staleness.

Sources¶

auto-lorebook sources list          # flags sources with missing session_date
auto-lorebook sources show <source_id>

QA¶

auto-lorebook seed-ingest --at={structure|summarize|approve|plan} \
  [--fixture <name>] \
  [--source-id <id>]

Mints a fresh qa-<hex> source_id and lays down a synthetic ingest up through the chosen stage, so the next pipeline stage can be exercised in isolation without re-running prior stages or hitting the LLM. The default fixture is tiny-aldara; additional fixtures live under src/auto_lorebook/_qa_fixtures/. Pair with reject-ingest <id> to clean up. See QA seeding.

Web UI¶

auto-lorebook serve [--port 8080]

Available from Phase 6 onward. See roadmap.

Utility¶

auto-lorebook version
auto-lorebook --version

Display the installed package version. version is a subcommand; --version is a top-level flag. Both print the same string.