v0.48.0 Release Recap
v0.48.0 is a non-breaking ranking observability release that adds JSON eval snapshots for the representative corpus.
What Changed
| Area | New in v0.48.0 | Why it matters |
|---|---|---|
| CLI ranking snapshot | contextdb eval ranking | Emits a repeatable JSON report for representative corpus queries |
| Score drift data | Per-query top-k, expected rank, reciprocal rank, and MRR | Makes release-to-release ranking movement reviewable |
| Score breakdowns | Raw component scores and weighted score breakdowns | Shows why each result ranked where it did |
| Feature metadata | ranking-eval-snapshots | Lets live deployments advertise ranking eval snapshot support |
Compatibility
This release only adds a local evaluation command. It does not change retrieval scoring, corpus fixtures, public query APIs, or write behavior.
Operator Flow
bash
contextdb eval ranking --out ranking-eval.json --reportUse the generated snapshot before and after ranking changes to compare expected rank, MRR, result order, and score component movement.
Verification
Release gates for this slice:
| Gate | Status |
|---|---|
go test -count=1 ./... | Passed |
npm run docs:build | Passed |
git diff --check | Passed |