Repository navigation
Evaluation: Add dataset filter option #1137
Description
Is your feature request related to a problem?
The GET /api/v1/evaluations endpoint currently returns all evaluation runs in a project without the ability to filter by a specific dataset. This requires clients to fetch and filter results client-side, leading to unreliable data viewing due to potential issues with pagination.
Describe the solution you'd like
- Add an optional
dataset_idquery parameter toGET /api/v1/evaluations. - Filter runs in SQL through
list_evaluation_runsinapp/crud/evaluations/core.pybefore applying ordering/limit/offset. - Maintain compatibility with the existing
typefilter and pagination. - If the parameter is omitted, retain the current behavior of showing runs across all datasets in the project.
- Ensure
dataset_idfrom another project returns no runs. - Update
app/api/docs/evaluation/list_evaluations.mdto include the new parameter documentation.
Original issue
Is your feature request related to a problem?
GET /api/v1/evaluations lists every evaluation run in a project with no way to scope the result to a single evaluation dataset. A client that only cares about one dataset's runs has to fetch runs and filter them client-side, which breaks against limit/offset — the runs it wants may not be on the page it fetched. Any per-dataset view (e.g. "runs for this dataset") is therefore unreliable today.
Describe the solution you'd like
- Add an optional
dataset_idquery param toGET /api/v1/evaluations. - Thread
dataset_idthroughlist_evaluation_runsinapp/crud/evaluations/core.pyso filtering happens in SQL, applied before ordering/limit/offset. - Compose with the existing
typefilter and pagination rather than replacing them. - Omitting the param keeps current behavior: runs across all datasets in the project.
- Keep results org/project-scoped — a
dataset_idbelonging to another project returns no runs. - Update
app/api/docs/evaluation/list_evaluations.mdto document the param.
Comments: Implemented in #1136
Original issue
Backend: GET /api/v1/evaluations (list evaluation runs) has no way to scope results to a single evaluation dataset. A client that wants only one dataset's runs must page through every run in the project and filter client-side, which breaks against limit/offset — the runs they want may not be on the page fetched. Ask: add an optional dataset_id query param to the list endpoint (and the underlying crud list_evaluation_runs), composing with the existing type filter and pagination. Repo: ProjectTech4DevAI/kaapi-backend. Related PR: #1136
Metadata
Metadata
Assignees
Labels
Type
Projects
- Status
Activity
[-]Evaluation: Add dataset_id filter to evaluation runs list[/-][+]Evaluation: Add dataset filter option[/+]github-actions commentedon Aug 18, 2026
🎉 This issue has been resolved in version 1.5.0-main.5 🎉
The release is available on GitHub release
Your semantic-release bot 📦🚀