The README feature list and the overview stopped at the analysis capability.
Both examples and the app backend composed agents without the capabilities the
documentation recommends alongside an evidence capability.
custom_agent.py ran each input as an independent agent run, so it needed a state
dict and a carried history before compaction could mean anything there: without
state the evidence record is empty, and earlier evidence would reduce to
receipts retaining nothing.
`_compact_old_tool_returns`, `PRIOR_TURN_NOTICE` and `turn_start` leave
`RAGCapabilityBase`, along with its `wrap_model_request` hook. The evidence
capabilities now retrieve and validate, and nothing else. Registering the compaction
capability is what rewrites a request; leaving it out sends the transcript untouched,
which was never a choice a host could make before.
The boundary is the recorded question identity rather than message shape, so a
resumption compacts what lies below the question in progress instead of switching
compaction off for the whole run. The newest earlier evidence return carries the
capsule and every other becomes a receipt, so one capsule exists by construction and
every return stays paired with its call.
Pictures of cited evidence are fetched through the capability that retrieved them and
re-attached beside the capsule with fresh labels. Ownership of a picture on the wire
requires the machine tag we write and an image directly after it, since neither
position nor prose is proof: several tools' results can arrive in one request, and a
user quoting our wording above their own picture had it removed. A picture that cannot
be fetched or decoded is emitted with neither its image nor its label.
The chat TUI and the example backend register the compactor, being multi-turn.
`client.ask`, `client.analyze` and the MCP tools do not: a single-shot question has
nothing earlier to compact.
visualize_chunk re-expanded chunks from scratch to recover their refs,
which could not faithfully reproduce the original merge, scores, and
clip — so a visualization could highlight different pages than the
citation covered. Carry the cited items on Citation.doc_item_refs and
resolve bounding boxes from them directly; re-expansion remains only as
the fallback for callers with no stored context (CLI, inspector). Chat,
inspector, the app endpoint, and the frontend pass the refs through.
Convert all LanceDB operations from sync calls wrapped in async
functions to the native async API (connect_async, AsyncConnection,
AsyncTable, AsyncQuery). Database I/O no longer blocks the event loop.
- Store and HaikuRAG use async context managers (async with). Store
initialization is deferred to __aenter__; direct construction
without async with is no longer supported.
- Index creation uses config objects (FTS, BTree, IvfPq) instead of
string-based index_type parameter.
- Upgrade callbacks are async.
- HaikuRAG tracks background vacuum tasks and awaits them in __aexit__
and before destructive rebuild operations to avoid races with
concurrent table mutations.
- temp_db_path fixture uses pytest's tmp_path for reliable async
cleanup.
Apply the same find_dotenv(usecwd=True) fix from cli.py to:
- evaluations/evaluations/benchmark.py
- app/backend/main.py
Co-Authored-By: Tianyi Cui <contact@tianyicui.com>
Sends SSE comment every 15 seconds during long LLM operations
to keep the connection alive and prevent body timeout errors.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
