AIR

Semantic Quality Report

Compare embedding models using the opt-in backend semantic quality sweep.

Run + load

Running a sweep requires Docker (testcontainers) and can take a few minutes. The backend will run the existing pytest-based integration sweep and return a report JSON.

Dense models (comma-separated)

Sparse model

Custom dense model JSON (optional)

Load existing JSON file

Or fetch from URL

Load a report JSON generated by the backend test runner.