OSS ChatGPT WebUI v4 Launches with Projects, Agent Profiles, and 1-Click Publishing
Key Takeaways
- ▸Projects and Agent Profiles add workflow management and AI specialization capabilities, enabling users to isolate workspaces and customize agent behavior
- ▸Supports 530+ models from 24 providers with daily model updates, spanning OpenAI, Google Gemini, OpenRouter, and others
- ▸One-click publishing to public gallery enables creators to share generated content and discover others' creations with community features (ratings, tags, categories)
Summary
Version 4 of llms.py, an open-source and OpenAI-compatible AI gateway, introduces major new features for managing AI workflows and sharing generated content at scale. The self-hosted platform unifies access to 530+ models across 24 providers, supporting multimodal interactions including text chat, image generation, and text-to-speech synthesis. The latest release prioritizes extensibility and user experience, featuring an enhanced plugin system, Model Context Protocol (MCP) support, and expanded provider integrations that integrate daily updates through models.dev.
V4 brings four standout capabilities: Projects for secure workspace isolation with restricted AI filesystem access, Agent Profiles that bundle custom prompts and tools for specialized use cases, Server Tools for leveraging provider-hosted capabilities like web search, and one-click publishing to a public gallery at ai.llmspy.org. Additional improvements include AI-powered browser automation, voice input/output, integrated code execution environments, and support for over 200 system prompts. The community gallery enables users to publish generated images, audio files, and playable applications (games, tools) with community ratings, tags, and infinite-scroll discovery.
- Enhanced extensibility via plugins, MCP support, and server tools creates a flexible foundation for custom integrations and specialized workflows
Editorial Opinion
The v4 release demonstrates how open-source tooling can match—and in some ways exceed—commercial AI platforms by prioritizing user control, data privacy through self-hosting, and extensibility. The addition of community publishing features signals a shift toward treating open-source LLM platforms as creative tools, not just technical infrastructure. The challenge ahead is managing discoverability across 530 models and ensuring the interface remains intuitive as features accumulate.
