Hi Spring AI team!
I'm working on a project that uses Spring AI for various AI integrations (TTS, STT, chat models). Currently using Spring AI's excellent Anthropic integration for Claude API.
However, I'm also using Claude Code CLI - Anthropic's new command-line coding assistant that runs locally. It provides additional capabilities like MCP server integration, local tool execution, and coding-specific features.
Currently I'm calling Claude Code CLI directly via ProcessBuilder, but it would be fantastic to have it as a proper Spring AI ChatModel provider to leverage: - Unified interface with other models - Spring AI's retry mechanisms and observability - Easy switching between Claude API and Claude Code CLI - Testing abstractions
Questions: 1. Are there any plans to add Claude Code CLI support as a ChatModel provider? 2. Would the team be open to a PR implementing this? 3. Any guidance on best practices for process-based ChatModel implementations (since it's CLI-based, not HTTP API)?
There's a community wrapper making it OpenAI-compatible (https://github.com/RichardAtCT/claude-code-openai-wrapper), but native support would be cleaner.
Thanks for the amazing work on Spring AI!
🤖 Generated by Claude Code