User-visible improvements:
- Native Agent now shows AI responses in the chat interface
- Uses configured default model from settings instead of random selection
- Streams responses in real-time as the model generates them
Technical changes:
- Implemented response stream forwarding from Thread to AcpThread
- Created Session struct to manage Thread and AcpThread together
- Added proper SessionUpdate handling for text chunks and tool calls
- Fixed model selection to use LanguageModelRegistry's default
- Added comprehensive logging for debugging model interactions
- Removed unused cwd parameter - native agent captures context differently than external agents
- Add public session_id() method to AcpThread to enable testing
- Fix ModelSelector methods to use async move closures properly to avoid borrow conflicts
- Add test_agent_connection that verifies:
- Model selector is available for agent2
- Can list available models
- Can create threads with default model
- Can query selected model for a session
- Can send prompts using the selected model
- Can cancel sessions
- Handles errors for invalid sessions
- Remove unnecessary mut keywords from async closures
- Add ModelSelector trait to acp_thread crate with list_models, select_model, and selected_model methods
- Extend AgentConnection trait with optional model_selector() method returning Option<Rc<dyn ModelSelector>>
- Implement ModelSelector for agent2's AgentConnection using LanguageModelRegistry
- Make selected_model field mandatory on Thread struct
- Update Thread::new to require a default_model parameter
- Update agent2 to fetch default model from registry when creating threads
- Fix prompt method to use the thread's selected model directly
- All methods use &mut AsyncApp for async-friendly operations
This will prepare us for running the protocol over MCP
Release Notes:
- N/A
---------
Co-authored-by: Ben Brandt <benjamin.j.brandt@gmail.com>
Co-authored-by: Conrad Irwin <conrad.irwin@gmail.com>
Co-authored-by: Richard Feldman <oss@rtfeldman.com>
