STIP Docs

Changelog

Version history and updates for LLM Gateway API

Changelog

Track the evolution of the LLM Gateway API.

This changelog has been inferred from git commit history. Some entries may represent internal refactoring or infrastructure changes.

Recent Updates

Latest (HEAD)

  • Added GitHub Actions workflow for CI/CD
  • Updated project requirements and dependencies
  • Fixed FastAPI version compatibility

Refactoring: LLM Gateway

  • Complete refactoring of LLM inference functionality
  • Improved architecture with factory pattern
  • Enhanced code organization and modularity
  • Better separation of concerns (routers, services, core)

Dependency Updates

  • Updated to Pydantic v2.0.0
  • Added Loguru for enhanced logging
  • Updated to FastAPI[standard] v0.113.0
  • Fixed common models requirements

Testing & CI

  • Added comprehensive test suite (unit and integration tests)
  • Implemented CI pipeline
  • Test improvements and better coverage
  • Added Docker build automation

Provider Support

  • OpenAI integration
  • Groq integration (including Llama 3.3)
  • Google integration
  • Local LLM server support

Features

  • Text generation endpoint
  • Classification endpoint with structured output
  • Summarization endpoint with structured output
  • Streaming support via SSE
  • Token usage tracking
  • API key authentication
  • Comprehensive error handling

Version Tags

v0.1.6

  • LLM base model constraint generation improvements
  • Enhanced client functionality

v0.1.5

  • Configuration updates

v0.1.4

  • Hotfix for configuration issues

v0.1.3

  • Configuration hotfixes

Known Issues

This section documents known limitations and issues with the current version.

Streaming Limitations

  • Streaming is only supported for text generation tasks
  • Classification and summarization tasks do not support streaming
  • See the API documentation for details

Provider-Specific Behavior

  • Some models may have different token limits than specified
  • Response format may vary slightly between providers
  • Rate limits are enforced by individual providers

Upcoming

Check the project repository for planned features and improvements.

Potential Enhancements

  • Additional provider support (Anthropic, Cohere, etc.)
  • Batch request support
  • Response caching
  • Request queue management
  • Advanced retry strategies
  • Metrics and monitoring endpoints

Migration Notes

Breaking Changes

No breaking changes in recent versions. The API maintains backward compatibility.

Deprecations

No deprecated features at this time.

Contributing

To contribute to the changelog:

  1. Document all significant changes in commit messages
  2. Tag releases with semantic versioning
  3. Update this file when preparing releases

On this page