feat(feedback): add hydradb feedback, and surface the request id it needs

[gunmay-hydradb]

What

hydradb feedback REQUEST_ID — report whether a query's results were actually useful.

It correlates on one key, the request_id from that query's response meta. Nothing else about the original query is re-sent, so nothing has to be trusted from the client.

hydradb query "contract terms"
#   … results …
#   request_id: 8f1c0e8a-…  ·  rate it: hydradb feedback 8f1c0e8a-… --feedback "..."

hydradb feedback 8f1c0e8a-… -f "returned the 2023 policy, not the current one" --rating negative

This is two changes, because a feedback command alone would have shipped unusable

The request id was unreachable from the CLI. _unwrap returns .data and drops meta, so a successful query discarded its own request id before any caller saw it — you could run a query and then have nothing to give feedback on.

So context.query now carries the id into its payload. That is additive — /query's data has no request_id of its own — so the documented --output json shape gains a key and loses none:

hydradb --output json query "contract terms" | jq -r .request_id

The human output prints a copy-pasteable line, including on an empty result. A query that found nothing is the case most worth reporting, and the one with no chunk ids to fall back on.

FeedbackSubmitRequestFeedback.model_fields    # ['feedback']
FeedbackSubmitRequestGroundTruth.model_fields # ['ground_truth']

It does work today, and I verified that against staging — the models are extra="allow", so undeclared fields ride along to the wire. But that means every field this endpoint needs travels as an accident of pydantic config, at two levels of nesting, and sending an answer together with source ids means picking one branch and smuggling the other past it. A generated model that cannot name its own required field is equivalent to no model.

When the spec is fixed so the branches keep their siblings, this is a one-file change back to the SDK.

Note: the TS SDK has the same defect but strips unknown fields instead of allowing them, so hydradb-mcp had no choice at all there. Same spec bug, two different failure modes.

Testing

11 new tests, including that request_id actually reaches the wire and that an answer and source ids travel together — the two things the SDK model cannot do.

Verified end to end against staging: query → request id surfaced → feedback recorded, both the prose and ground-truth paths, with real feedback_ids returned.

• pytest — 342 passed
• pytest conformance — 33 passed
• make lint — clean

Related

• MCP sibling (Group 2): feat(feedback): add hydradb_feedback, and surface the request id it needs [PRO-1965] hydradb-mcp#81
• Backend: hydra-db/hydradb-application#706, #822

[greptile-apps]

Greptile Summary

This PR adds a complete feedback workflow to the CLI and exposes each query’s request ID so users can correlate feedback with an existing query.

• Registers hydradb feedback with prose, rating, ground-truth, source, and scope options.
• Adds a raw authenticated POST /feedback client resource with request normalization, source-ID deduplication, limits, and consistent response/error handling.
• Preserves query metadata by adding meta.request_id to the query payload and displays a copyable feedback hint, including for empty results.
• Adds wrapper and command-level tests for request encoding, validation, result rendering, and request-ID propagation.
• Greptile automatically discovered a related ticket that helped explain the purpose of this PR: provide CLI and MCP feedback support while surfacing the query request ID needed for correlation.

Confidence Score: 5/5

The PR appears safe to merge; the previous findings are fully addressed and no new actionable defects were identified.

Source validation now rejects unsupported values before network access, and the new command-level tests cover the previously untested public feedback surface. The additional validation and tests introduced since the previous review preserve the shared CLI execution, output, scope, and error contracts.

Important Files Changed

Filename	Overview
src/hydradb_cli/hydra/client.py	Adds request-ID propagation and a normalized raw feedback resource with authentication, scope, validation, and response handling.
src/hydradb_cli/commands/_impl.py	Adds feedback validation, execution, result rendering, and query feedback hints; the previous source-validation concern is fixed.
src/hydradb_cli/commands/canonical.py	Defines the public feedback command options and maps them through the shared command implementation.
src/hydradb_cli/main.py	Registers the new top-level feedback command.
tests/test_cli_commands.py	Adds command-level coverage that resolves the previous testing finding and protects validation, mapping, output, and exit behavior.
tests/test_wrapper.py	Covers feedback wire payloads, request normalization, source-ID handling, defaults, limits, and query request-ID propagation.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart LR
    User[CLI user] --> Query[hydradb query]
    Query --> Context[HydraDB context client]
    Context --> API[HydraDB API]
    API --> Envelope[Query data plus meta.request_id]
    Envelope --> Payload[Query payload with request_id]
    Payload --> User
    User --> Feedback[hydradb feedback REQUEST_ID]
    Feedback --> Validation[Validate signal, rating, source, and scope]
    Validation --> FeedbackClient[HydraDB feedback client]
    FeedbackClient --> Post[POST /feedback]
    Post --> Result[Recorded or not-durably-stored result]
    Result --> User

Loading

_{Reviews (2): Last reviewed commit: "fix(feedback): validate --source, and re..." | Re-trigger Greptile}