Self-hosted LLM chat that brings enterprise plumbing to Ollama

What it does

ChatOllama is a Nuxt 3-based chat interface that connects to a laundry list of language model providers—Ollama, OpenAI, Anthropic, Gemini, Groq, and others—through a single web UI. It adds team-oriented features like knowledge-base RAG, real-time voice chat via Gemini 2.0 Flash, and AI agents that can call external tools through the Model Context Protocol. The app also includes a full user-management hierarchy with super admins, admins, and regular users, plus access-control lists for who gets to configure those external tools.

The interesting bit

The project treats a personal LLM hobby like an enterprise SaaS deployment: it recently migrated from SQLite to PostgreSQL, supports feature flags to toggle entire modules at runtime, and lets you lock down MCP server configuration so only admins can add new tools. That makes it less of a quick personal script and more of a self-hosted team appliance for developers who want ChatGPT-like capabilities without sending everything to the cloud.

Key highlights

• Plugs into Ollama, OpenAI, Anthropic, Gemini, Groq, Moonshot, and generic OpenAI-compatible endpoints
• Knowledge bases with document upload and vector search via Chroma or Milvus
• Realtime voice chat using Gemini 2.0 Flash
• MCP integration for tool-using agents, with STDIO, SSE, and Streamable HTTP transport support
• Role-based access control for MCP server management and user administration
• Modular feature flags to enable or disable MCP, knowledge bases, voice chat, and model management

Caveats

• The new “Deep Agents” feature currently requires an Anthropic API key, even if you are otherwise running local models via Ollama
• The README warns that environment variable naming differs between Docker runtime (NUXT_ prefix) and local builds, which is an easy trap when toggling features
• It demands a full stack of dependencies: PostgreSQL, a vector database, and optionally Ollama or multiple commercial API keys

Verdict

Worth a look if you need a multi-user, self-hosted chat interface with RAG and tool access for a team. Skip it if you just want a lightweight personal front-end for Ollama—this one comes with database migrations and an admin panel.