Workflow KnowledgeFS retrieval nodes ran through `retrieval-tests`, which
never wrote an AnswerTrace or an overview event, so their queries were
missing from the retrieval history and from the space overview; the history
also only ever showed each member their own traces.
- AnswerTrace gains a `source` (retrieval_test | workflow | service_api |
agent | mcp; migration 0051_answer_trace_source) derived from the
Capability v2 caller kind. The retrieval-tests route records one trace per
run (stages, evidence bundle, profile metadata), returns its id as
`answerTraceId`, and the workflow node's failed-retrieval capture attaches
to that trace instead of creating a second record.
- The quality trace list exposes and filters by `source`; traces from other
caller kinds are visible to any current reader of the space, and counts and
scores fall back to the evidence embedded in the trace when no bundle row
exists.
- Overview accounting: retrieval-tests and Research tasks now emit
`query.requested`, and the Research job state machine emits
`query.completed` / `query.failed` on terminal stages (wired for both the
in-process gateway and the durable runtime), so query volume, answer rate
and outcomes include every caller. Activity details keep `source` and
`taskKind`.
- Console and service trace routes accept a `source` filter; the retrieval
test page shows a source badge and an all / retrieval tests / workflow
filter, with translations for every locale.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_015zw5G5SX3HmVfnZof6YWAc
The KnowledgeFS retrieval node treated the query text as mandatory even though
a query image is a complete input on its own. Text and image are now
alternatives: the node needs at least one of them, never both.
- web: drop the required mark from the query text field; the checklist fails
only when neither variable is bound (new i18n message in every locale).
- api: make `query_variable_selector` optional with a model-level rule that
one of the two selectors must be set, and only map the query variable when
it is bound. At runtime an empty text without images is a configuration
error; an image-only query retrieves with an empty text, merges spaces on
their own scores instead of a text reranker, skips automatic metadata
extraction, and skips the text-based failed-query capture.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_015zw5G5SX3HmVfnZof6YWAc
Bring the LLM-driven "automatic" metadata filtering mode from the legacy
Knowledge Retrieval node to the KnowledgeFS retrieval node
(knowledge-retrieval-v2).
Backend:
- Accept `automatic` in `metadata_filtering_mode` and add
`metadata_model_config` to the node entity.
- Add `automatic_metadata_filter.py`, which reuses the legacy metadata
filter prompt templates, invokes the configured model, parses the JSON
answer and normalises operators/values against the KnowledgeFS
metadata catalog before building custom metadata conditions.
- Resolve the shared metadata catalog across the selected spaces through
the new `list_metadata_fields` app execution capability (paginated),
fail closed on missing/unavailable models and catalog admission errors,
fail open on extraction errors, and expose the outcome via
`metrics.metadata_filtering` and `llm_usage`.
- Inject the default extractor from the node factory.
- Fix an unescaped brace in the legacy completion prompt template that
broke `.format()`.
Frontend:
- Offer disabled/automatic/manual modes, normalise unknown persisted
modes to disabled, store the automatic model and completion params,
render the model selector through the shared MetadataFilter, and
require a model when saving in automatic mode.
Co-Authored-By: Claude Code <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WnUM1CdV5LpcxZEQ2vu9Xc
