https://github.com/ndurner/oai_chat

Multi-modal Chatbot based on OpenAI
https://github.com/ndurner/oai_chat

chat chatbot gpt-4 gradio gradio-interface llm llm-inference mcp-client multimodal openai vision-language-model vlm

Last synced: 5 months ago
JSON representation

Multi-modal Chatbot based on OpenAI

• Host: GitHub
• URL: https://github.com/ndurner/oai_chat
• Owner: ndurner
• License: mit
• Created: 2023-11-12T20:41:55.000Z (over 2 years ago)
• Default Branch: main
• Last Pushed: 2025-08-03T14:54:29.000Z (10 months ago)
• Last Synced: 2025-08-03T16:33:24.423Z (10 months ago)
• Topics: chat, chatbot, gpt-4, gradio, gradio-interface, llm, llm-inference, mcp-client, multimodal, openai, vision-language-model, vlm
• Language: Python
• Homepage:
• Size: 133 KB
• Stars: 6
• Watchers: 3
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

          
---

title: OAI Chat

emoji: 🤖

colorFrom: yellow

colorTo: gray

sdk: gradio

sdk_version: 5.36.2

app_file: app.py

pinned: false

license: mit

---

# OAI Chat

Chat interface based on OpenAI transformer models. \

Features:

 * Image upload (support for vision via gpt-4-vision)

 * Word file (DOCX) upload

 * PDF file support (via image rendering & GPT-4V)

 * Plaintext file upload

 * chat history download

 * file download

   * example: download an ICS calendar file the model has created for you

* streaming chat

* image generation (via gpt-image-1, aka "ImageGen")

* MCP server support (both remote and local) via configurable registry

* optional UnrestrictedPython execution when `CODE_EXEC_UNRESTRICTED_PYTHON=1`

The MCP registry is looked up in the following order:

1. `$OAI_CHAT_MCP_REGISTRY` if set

2. `mcp_registry.json` in this repository

3. `~/.oai_chat/mcp_registry.json`

See `mcp_registry.sample.json` for an example configuration.

Headers and query parameters may reference environment variables using the `env:` prefix.

Use `"allowed_tools": ["*"]` to permit all tools from a server.

For local MCP servers, use the `command` and `args` fields to specify how to launch the server. Environment variables can be passed via the `env` field. For example:

```json

{

  "name": "exa_local",

  "command": "npx",

  "args": ["-y", "exa-mcp-server"],

  "env": {

    "EXA_API_KEY": "env:EXA_API_KEY"

  },

  "allowed_tools": ["*"]

}

```

When an MCP tool requires approval, the assistant will notify you in chat.

Reply with `y` to approve or `n` to deny the request, optionally adding a comment after the `y` or `n`.