v0.27.1 v0.7.0
tashfeenahmed/freellmapiAug 11, 2026 → Aug 11, 20260 days between releases
What's New
Major release adds OpenAI-compatible API and advanced routing reliability.
Breaking Changes
- OpenAI-compatible API surfaces aligned with the official Responses API spec.
New Features
- Per-model routing weight overrides via `MODEL_ROUTING_OVERRIDES` env var
- OpenAI-compatible API surfaces aligned with the official Responses API spec
- Models that repeatedly fail are auto-cooled-down
- Enhanced key management with per-key proxy overrides
- Security hardening with owner-only database permissions
Improvements
- Failed providers are skipped as a whole on 5xx/timeout errors
- Streaming responses record real token usage for accurate analytics
- Provider back-off stated in the error body is honored
- Exploration probe respects weight-0 overrides
- Opt-in schema validation of tool-call arguments
Bug Fixes
- Provider back-off stated in the error body is now honored
Upgrade Notes
Configure the new `MODEL_ROUTING_OVERRIDES` environment variable for advanced routing control. Be aware that OpenAI-compatible API surfaces are now aligned with the official Responses API spec, which may require client adjustments. Review key management settings for new bulk actions and proxy overrides.
The Two Releases
v0.27.1
Aug 11, 2026· 2 months ago
Minimal release notes: "This is a patch release on top of v0.27.0. - Support quantized DSpark Markov heads (#50424)"
Full release notesv0.7.0
Aug 11, 2026· 2 months ago
This is the largest release since v0.6.0, introducing 39 commits covering routing, OpenAI-compatible surfaces, key management, analytics, security hardening, and the desktop app. It focuses on reliability, API alignment, and improved user control over API keys.
Full release notes