Changelog
Features
- 8374d53 feat(azure): forward gpt-6-luna decisions to Foundry's Decisions API (#1320)
- ea3b062 feat(azure): serve Microsoft-Decision-1 through decisions and System One (#1319)
- 6e197ad feat(azure): support Azure AI Foundry v1 base URLs (#1317)
- 6e4da41 feat(azure): use the GA realtime websocket on v1 base URLs (#1318)
- 63365a4 feat(chatgpt): serve chat completions via Responses translation (#1313)
- 8088368 feat(embeddings): support configured model failover (#1309)
- 16cc74f feat(http): trust private CAs and client certificates on every outbound connection (#876)
- 3c6b6b7 feat(pluginapi): add a range input for bounded numeric fields (#1244)
- 7455aaf feat(server): advertise models the way vLLM discovery expects (#1245)
- e3086be feat(server): serve native TLS with hot-reloaded cert files and optional mTLS (#1326)
- 535bfa7 feat(vertex): add anthropic api_mode (#1322)
Bug Fixes
- 9e453a0 fix(gemini): translate Gemini 4 requests and repair the OpenAI-compatible surface (#1329)
- 8c611be fix(passthrough): proxy provider-native error bodies, clean decode errors, cap stop sequences (#966)
- f5ce0db fix(server): keep skip paths with a query or fragment unmatched (#1331)
- 73ac234 fix(tls): address TLS and certificate review findings (#1330)
- fecf3b1 fix(usage): record embedding token usage from Gemini and partial reports (#1315)
- d85952c fix(virtualmodels): stop narrower policies from hiding a model-wide slowdown (#1325)
- b9ae943 fix: close the 0.2.0 pre-release findings (#1332)
Full Changelog: v0.1.109...v0.2.0