🚀 Release Notes 🎉

Highlights ✨

• ✅ Fixes & Improvements: Improved provider status handling, error management, and API response processing for a smoother and more reliable experience.
• 🛠️ Refactor: Switched from see_stream to sse_stream for better streaming support across multiple providers.
• 🔄 Model Updates: Updated default and fallback models for LambdaChat and Gemini providers to ensure more accurate and up-to-date responses.
• ⚠️ Error Handling: Enhanced error detection and messaging in backend API and Kimi provider to better capture and report issues.
• 📚 Documentation: Fixed typos and updated links for better clarity (including correction from "Inference API" to "Interference API").
• 🧩 Optional Dependencies: Added conditional import for MarkItDown with graceful failure if missing.
• 💬 Chat Providers: Enabled more free providers and updated streaming defaults for improved user experience.
• 🖼️ PollinationsAI: Improved image-generation response handling and progress reporting.

Detailed Changes 📝

Providers Updates 🔧

• Free2GPT, Startnest, and Reka providers: Fixed provider status flags to correctly represent working state.
• LambdaChat: Changed default model from deepseek-v3-0324 to deepseek-r1 and cleaned up model aliases.
• Kimi: Replaced manual status checking with proper async error raising, added specific model "k2" for chat requests, improved exception catching.
• WeWordle: Simplified payload construction by removing unnecessary parameters.
• Reka: Set default streaming to True and modified cookie-fetching behavior for freshness.
• GeminiPro: Updated model defaults and fallback models for better reliability.
• Copilot: Cleaned up model aliases and streamlined providers' comparison logic.

Streaming and API 🔄

• Replaced all see_stream function calls with sse_stream for improved server-sent-event streaming compatibility.
• In backend API, enhanced exception handling including classy error messages with exception types for better debuggability.
• Added default "user": "unknown" field to cached data for traceability.

CLI & GUI Improvements 🖥️

• CLI: Added conditional import for the optional MarkItDown dependency, raising explicit errors when missing.
• GUI server backend API: Wrapped provider model queries in error catch to return HTTP 401 if authentication is missing, improving security feedback.

Others 🧹

• md2html.py tool now updates existing HTML files by replacing titles and text instead of rewriting everything, making docs generation more efficient.
• Fixed typo: "Inference API" → "Interference API" in README and updated related documentation links.
• Improved indentation and condition checks in OpenaiTemplate response parser for cleaner code and fewer parsing errors.
• PollinationsAI: Added progress and reasoning status updates during async image generation, improving transparency in UI/UX.

Thanks 🙏

Special thanks to all contributors, especially @hlohaus, for their continuous efforts improving g4f!

Upgrade now and experience the latest improvements and fixes! 🚀

Full Changelog: https://github.com/xtekky/gpt4free/compare/0.5.7.6...0.5.8.0