llms.py v4 Released: Open-Source ChatGPT Alternative with Projects, Agent Profiles, and Gallery Publishing

Summary

llms.py v4 has been released, delivering a self-hosted, OpenAI-compatible AI gateway that supports 530+ models across 24 providers including OpenAI, Google Gemini, and others. The new version introduces major features including secure Projects for workspace isolation, specialized Agent Profiles with custom prompts and tools, Publisher functionality for sharing AI-generated content to a public gallery, and expanded server tools support with built-in web search capabilities.

The lightweight CLI and web UI provide a privacy-focused ChatGPT alternative for both local and cloud LLMs, supporting text, image, and audio generation in a single interface. Notable new capabilities include multimodal support with image generation via Gemini, OpenAI, and Nvidia, high-quality text-to-speech synthesis, and first-class Python function calling. The platform also features MCP (Model Context Protocol) support, AI-powered browser automation, and voice-to-text input.

The Publishing feature is particularly significant, allowing users to create and publish AI-generated games, images, audio, and projects directly to a live showcase at ai.llmspy.org. The new gallery supports browsing by category, tags, ratings, and infinite scroll, making it easy for the community to discover and play self-contained AI-generated applications without requiring authentication.

• Publisher system lets users share AI-generated content (images, audio, games) to public gallery at ai.llmspy.org with one-click publishing and free anonymous accounts
• Expanded multimodal support including image generation across 6+ providers, TTS with Gemini, voice input, and playable self-contained AI-generated apps