# Changelog

## v188.8 — deploy-clean release

This deployable package consolidates the v188.1 → v188.8 fixes into a clean runtime archive.

### Highlights

- Partial validation states no longer render as full success.
- `/health` validation is correctly reflected in inference gate records.
- Generation smoke tests are schema-aware and can use `INFERENCE_CONTRACT.json` smoke payloads.
- Smoke-test failures are classified by owner/type, including validator request errors vs generated app runtime errors.
- Anonymous eval records include explicit verdict, outcome, and validation fields.
- Backend eval archive publication is observable via `eval_publish_status.json`.
- Active Run UI better reflects post-run states, recovery, fallback hardware, latency, documents, and eval archive status.
- Space log collection is more robust, with SDK/REST fallbacks and structured diagnostics.
- Runtime versioning is centralized in `src/version.py`.

### Sanitization notes

The deploy-clean archive removes development-only tests, old version-specific changelog files, Python caches, and temporary files.

Operational/debug assets are preserved, including `worker/hello_worker.py`.