- Expose model provider header (#100)
- anthropic: drop top_p when both temperature and top_p are set + flatten response (#99)
- add base_url support for OpenAI provider in managed mode (#98)
- distroless base image reducing CVEs from 89 to 10 (#96)
- typo in hub_mode env var (#90)
- vertexai: support API key auth via Gemini Developer API (#87)
- helm: deploy using helm chart (#83)
- remove deploy to old staging CD phase (#84)
- czn: manual bump (#82)
- helm: deployment using helm chart (#80)
- ci: lock version (#75)
- openai: add tool_call_id request param (#73)
- sync versions to 0.7.2 and migrate to .cz.yaml config (#70)
- max_tokens change to max_completion_tokens (#68)
- tracing: add vendor reporting for OTEL tracing (#64)
- fix bump package (#67)
- revert tower http request (#66)
- add reasoning support (#63)
- make effort optional for gemini (#65)
- gemini enum support (#62)
- deps: revert commitizen tower version bump (#61)
- add gemini structure output (#60)
- gemini system prompt (#59)
- models: add filtered model info retrieval and response structures (#51)
- management API (#39)
- config: allow env vars (#58)
- Dockerfile: specify compatible sqlx-cli version for edition2021 (#54)
- simplify string formatting to remove clippy warnings (#53)
- deps: revert commitizen unwanted chrono version change (#49)
- bedrock: handle ARN and inference profile identifiers without transformation (#48)
- bedrock: support IAM role auth (#47)
- make general optional again (#43)
- tracing: support disabling tracing of prompts and completions (#42)
- openai: support custom base URL (#40)
- azure: add support for custom base URL in AzureProvider endpoint (#41)
- provider: add Google VertexAI support (#24)
- support AWS bedrock base models (#25)
- add max_completion_tokens to ChatCompletionRequest (#36)
- support structured output (#33)
- replace eprintln with tracing info for API request errors in Azure and OpenAI providers (#37)
- make optional json_schema field to ResponseFormat (#35)
- add logprobs and top_logprobs options to ChatCompletionRequest (#27)
- cd: correct docker hub secrets (#31)
- azure: embeddings structs improvement (#29)
- add proper error logging for azure and openai calls (#18)
- anthropic: separate system from messages (#17)
- tool call support (#16)
- restructure providers, separate request/response conversion (#15)
- openai: support streaming (#10)
- add prometheus metrics (#13)
- cd: deploy to traceloop on workflow distpatch (#11)
- config file path from env var instead of command argument (#12)
- otel support (#7)
- implement pipeline steering logic (#5)
- dynamic pipeline routing (#4)
- azure openai provider (#3)
- initial completions and embeddings routes with openai and anthropic providers (#1)
- dockerfile and release pipeline (#2)
- make anthropic work (#8)
- cleanups; lint warnings fail CI (#9)
- missing model name in response; 404 for model not found (#6)