Summary

• Add OpenAI-compatible API endpoint support as 4th AI provider option
• Configurable endpoint URL, API key, and model settings
• Supports any service using OpenAI chat completions format
• Uses standard OpenAI chat completions API structure
• Consistent error handling and response processing

Test plan

• Verify "OpenAI-Compatible" appears in AI model dropdown
• Test with OpenAI API endpoint and valid API key
• Test with any custom OpenAI-compatible endpoint
• Verify error handling for invalid endpoints/API keys
• Test with various compatible model names

Implementation Details

• Added preference keys for endpoint URL, API key, and model
• Implemented OpenAI-compatible API call logic following chat completions format
• Updated AI model selection to include option 4
• Consistent prompt template and generation config structure
• Uses same categorization logic as other AI providers

Compatibility

Works with any service supporting OpenAI chat completions format:

• OpenAI API (default endpoint)
• Local LLM servers (LM Studio, Oobabooga, etc.)
• Custom endpoints that follow OpenAI's API structure
• Users can configure any compatible endpoint manually

Note: This is a generic OpenAI-compatible implementation - users need to configure specific endpoints manually if they want to use services other than OpenAI.

💘 Generated with Crush CLI using GLM-4.6 to vibe code these features. Manual review is recommended as I'm not a programmer.