Open-source OvertChat UI released for self-hosted vLLM deployments with integrated voice and agent support

OvertChat, an MIT-licensed open-source UI layer, has been released to simplify the deployment of self-hosted LLMs for non-technical users. Built to address the "last mile" friction of local AI, the tool connects to vLLM's OpenAI-compatible endpoint and bundles essential services including Parakeet STT, Kokoro TTS, and SearXNG for voice and search capabilities.

Key Features and Architecture

The project aims to replace complex setups like Open WebUI, which the author found too dense for family use. OvertChat provides a chat-first interface with native clients for web, Android, iOS, and desktop (macOS Apple Silicon/Intel and Linux). It includes an installer that wires together the separate services required for a complete local AI stack, reducing the deployment tax often associated with self-hosted voice and search integrations.

Agent Integration

A notable feature is Agent Connections, which allows users to hook coding agents such as Codex, Claude Code, OpenCode, or Hermes into the chat interface via a Host Connector. This enables users to drive coding agents running on their local machine or an SSH host directly from the chat UI, with visibility into plans, tool calls, and approvals. The author reports that the system has been in daily use by multiple non-technical users for months, running on an 8x 3090 vLLM box serving Qwen 3.8 Flash Next.