* fix(rerank): support DashScope nested request/response envelope
OpenAIRerankClient sent a flat request body ({"model", "query",
"documents"}) and parsed "results" at the top level of the response.
DashScope (qwen3-rerank) requires a nested envelope:
Request: {"model", "input": {"query", "documents"}, "parameters": ...}
Response: {"output": {"results": [...]}, "request_id", "usage"}
This caused DashScope rerank to silently fail — the response had no
top-level "results" key, so the client returned None.
Changes:
- Add _is_dashscope() to detect DashScope endpoints by host marker.
- Add _build_request_body() that produces the nested envelope for
DashScope and the flat body for standard OpenAI/Cohere services.
- Add _extract_results() that reads output.results for DashScope and
top-level results for standard services.
- Accept both "relevance_score" (singular, DashScope) and
"relevance_scores" (plural, some providers) in result items.
- Add 13 tests covering host detection, body construction, response
parsing, end-to-end mocked flows for both providers, plural key
handling, empty documents, and sparse results.
Fixes#3459
* fix(rerank): detect DashScope protocol by URL path, not hostname
Reviewer noted the previous hostname-based switch broke the documented
qwen3-rerank compatible-api endpoint (/compatible-api/v1/reranks), which
must use the flat OpenAI-style body and top-level results.
Switch to path-based detection: only /api/v1/services/rerank uses the
native nested input/output envelope; everything else (including the
DashScope compatible-api and generic OpenAI/Cohere gateways) keeps the
flat protocol. Rename _is_dashscope -> _uses_nested_envelope for clarity.
Add regression tests covering the compatible-api flat path and reconcile
the existing native-path fixtures to the nested envelope.
* docs(rerank): use qwen3-rerank for compatible-api example
The compatible-api/v1/reranks endpoint uses the flat OpenAI-compatible
protocol; qwen3-vl-rerank is a native-envelope model served at
/api/v1/services/rerank. Align the example model with the endpoint the
implementation selects by URL path.
---------
Co-authored-by: zhangyu.34 <zhangyu.34@bytedance.com>
* fix(task): recover add-resource jobs after restart
Persist asynchronous add-resource work in QueueFS so interrupted jobs can resume instead of leaving tasks running forever.
* fix(queue): omit parser args from prepared jobs
* fix(queue): fail when semantic source is missing
* fix(session): return in-flight archive messages in get_session_context (#3129)
Seed latest_completed_index from 0 instead of commit_count so that
archives whose Phase 1 has completed (messages written, commit_count
advanced) but whose Phase 2 is still running (.done not yet written)
are treated as pending rather than already completed.
The release/0.3.x implementation seeded from 0 and did not have
this bug; the regression was introduced when commit_count was
adopted as the seed value.
* test(session): deterministic regression for pending archive context (#3129)
Replace the monkey-patched commit_async test with a direct
filesystem-state test that sets up the post-Phase-1 archive
(messages.jsonl present, commit_count advanced, no .done marker)
and asserts get_session_context still surfaces the archived
messages.
This is deterministic regardless of the queue-worker architecture
because it creates the archive state directly via the mock AGFS
and loads a fresh session from that state, never calling
commit_async or touching the session compressor.
* perf(vectordb): coalesce auto cuVS rebuilds during bulk ingest
Add an opt-in bulk-ingest maintenance scope that coalesces Auto cuVS background rebuilds across multiple write batches.
- defer derived GPU maintenance until the outermost bulk scope exits while keeping native writes and persistence visible per call
- harden the background worker against debounce, generation, shutdown, and stale-candidate races
- preserve suspension across index replacement and retire replaced workers
- wait for the final Auto GPU snapshot before vectordb_perf records search QPS
- document that the scope is non-transactional and only schedules readiness on exit
Auto cuVS and background rebuild remain disabled by default. Native CPU and remote backends use no-op hooks, so their existing behavior and dtype are unchanged.
* fix(vectordb): reject stale index replacements
* fix(vectordb): harden bulk rebuild lifecycle
---------
Co-authored-by: Yuanqing Zhao <2604121+yuanqingz@users.noreply.github.com>
LiteLLMRerankClient.rerank_batch wrapped each document as {"text": d} and read
result items via getattr(item, ...). That works for Cohere-style object results
but breaks Voyage through litellm: Voyage's rerank API rejects dict-wrapped
documents (400: 'documents' is not a valid string), and litellm returns Voyage
results as plain dicts, so getattr(item, "index") misses and rerank silently
falls back to a no-op.
- Pass documents as plain strings (litellm.rerank expects List[str]).
- Add _result_field() to read index/relevance_score from dict- or object-shaped
result items.
Verified end-to-end against voyage/rerank-2.5 (scores now applied). Adds
tests/unit/models/rerank/test_litellm_rerank.py covering the plain-string
documents contract and both dict- and object-shaped results.
Co-authored-by: Michael Tarleton <mtarleton@istation.com>
FastMCP derives tool input schemas from Python type hints, so Optional/
Union parameters become anyOf nodes with no top-level type, and nested
models become $ref/$defs. Valid JSON Schema, but clients that forward
these schemas verbatim to strict function-calling APIs break: Gemini's
OpenAPI 3.0 subset rejects the whole request (400: schema didn't specify
the schema type field), and n8n's JSON-schema-to-Zod conversion can
silently fall back and drop tool arguments entirely.
Rewrite the advertised schemas after registration: drop null branches,
collapse unions to their most general branch, inline $refs, and ensure
every node carries an explicit type. Runtime argument validation still
uses the original function signatures, so union parameters keep
accepting every branch (e.g. read still takes a bare URI string even
though the schema advertises an array).
Also type recall's other_peer_penalty honestly as
Optional[Union[float, Dict[str, float]]] instead of Optional[Any],
which produced an empty {} schema node.
* feat(connector): delegate add_resource imports to external Connector
Opt-in integration that routes add_resource data fetching and parsing
to external Connector service; the Connector stages source data and
calls back into OV through the standard add_resource pipeline.
- add ConnectorClient wrapping the control plane's inner doc/add and
task/info endpoints
- add [connector] config section: enable, connector/tracker endpoint
URLs, timeout_seconds, poll_interval_ms, allowed_add_types
- route add_resource via Connector when enabled and args.add_type is
in allowed_add_types; otherwise fall back to the standard pipeline
with an info log
- track imports as connector_import TaskRecords and poll Connector
task status in the background until terminal state or timeout
* fix(connector): delegate add_resource imports to external Connector
