llms.py v4 Released: Unified AI Gateway with Projects, Profiles, and Public Showcase
Key Takeaways
- ▸Unified gateway supports 530+ models from 24 cloud and local providers in a single self-hosted interface
- ▸New Projects feature provides secure workspace isolation restricting AI agent filesystem access to designated folders
- ▸One-click publishing enables users to share generated images, audio, and interactive applications to a public gallery
Summary
llms.py, an open-source self-hosted AI gateway, released version 4 with major new features designed to enhance productivity and enable content sharing. The update introduces Projects for secure workspace isolation, Agent Profiles for specialized AI workflows, and a public Gallery that allows users to publish generated content with a single click to a live showcase at ai.llmspy.org.
Version 4 consolidates access to 530+ AI models from 24 providers (including OpenAI, Google Gemini, OpenRouter, and others) through a unified interface supporting multimodal capabilities—text generation, image creation, text-to-speech, and voice-to-text. The release adds server tools for web search, Model Context Protocol (MCP) support, skills extensions, and eight customizable themes alongside a modern ChatGPT-like web UI with voice input via microphone.
The standout feature is one-click publishing to the public Gallery, enabling users to share AI-generated images, audio, and playable games and applications. The Gallery provides community browsing by category and tags, with generated apps ready to play directly. llms.py remains free and open-source, supporting both CLI usage and self-hosted deployment.
- Multimodal support for text chat, image generation, text-to-speech, and voice-to-text transcription with 24+ providers
- Eight customizable themes, Agent Profiles for specialized workflows, and extensibility via MCP and Skills plugins
