dma123/gpt-js

GPT JS Chat

An HTML/JS-based chat application that interacts with AI models via the OpenAI-compatible API.

Warning: This tool is in development, made as a spare time project, and no warranty is given for any functionality.

Key capabilities include:

• —AI Interactions: Stream responses from AI models. Generate text, tables, code, formulas (via LaTeX), and simple SVG images/charts.
• —Tool Support: Integrate external tools (e.g., web search) using MCP for real-time data fetching.
• —Formatting & Usability: Render Markdown, highlight code, display math equations, and add citations for tool outputs.
• —Chat Management: Multiple chats with save/load, editing/deleting messages, alternative responses, and branching conversations.

Usage

Online Demo

Test a (sometimes outdated version) at: https://huggingface.co/spaces/dma123/gpt-js.

Local Setup

1. 1.Clone the repo: git clone https://github.com/dmatscheko/gpt-js.git
2. 2.Run a simple HTTP server: python -m http.server 8000 (or use any static file server).
3. 3.Open http://localhost:8000 in your browser.

With MCP (for Tools)

To enable advanced tools like web/X search:

1. 1.Clone the repo: git clone https://github.com/dmatscheko/gpt-js.git
2. 2.Install dependencies: uv sync (requires uv; alternatively, use pip for fastmcp).
3. 3.Customize tools via mcp_config.json.
4. 4.Run: uv run main.py (or python main.py). This starts a local MCP proxy at http://127.0.0.1:3000/mcp and auto-configures the app.

Controls

• —Input: Type messages; use Shift+Enter (or Ctrl/Alt+Enter) to submit. Press Esc to abort AI responses.
• —Chats: Sidebar for multiple chats; edit titles, add/delete messages, navigate alternatives.
• —Settings: Tune sampling (temperature/top-p), manage API endpoints, refresh models, select role (user/system/assistant/tool).
• —Avatars: Click message avatars to upload custom images (stored locally).

Screenshot

This screenshot was "randomly selected" because its output was ok-ish ;) [image]