All notable changes to this project will be documented in this file.
The format is based on Keep a Changelog, and this project follows Semantic Versioning.
- Breaking V2 config migration: replaced layered model configuration (
preset, groupedmodels,model,*_model,ai_provider) withruntime + flat modelscontract. - Public runtime values are now
claude_codeandopen_code. - Runtime/model selection is accepted only inside
configfor build/execute flows.
- Legacy configuration keys for model/provider selection (
ai_provider,preset,model, grouped model maps, and all*_modelrequest keys).
- Initial public release of SWE-AF
- Multi-agent SWE orchestration with plan/execute/verify flow
- Docker and local execution modes
- Agent-level API endpoints and artifact persistence
- OpenCode provider support for open-source models (DeepSeek, Qwen, Llama, MiniMax via OpenRouter)
- Multiple AI provider support: Claude (Anthropic) and OpenCode (OpenRouter/OpenAI/Google)