This is a fork of the original Reflex Chat with enhanced multi-provider support.

A user-friendly, highly customizable Python web app designed to demonstrate LLMs in a ChatGPT format with support for multiple LLM providers. This fork serves as an excellent example for developers who want to learn how to implement and integrate different LLM providers in their applications.

• Added Ollama support - Run local models with real-time streaming
• Added Google Gemini support - Including Vertex AI integration with streaming
• Enhanced provider architecture - Clean, extensible factory pattern for adding new providers
• Unified streaming - All providers now support real-time response streaming
• uv support - Lightning-fast dependency management

This fork is designed to help developers understand:

• How to implement multiple LLM provider integrations
• Async streaming patterns for real-time chat
• Clean architecture patterns for provider management
• Environment-based configuration for different services
• Error handling and fallback mechanisms

The app supports three LLM providers. Configure the one you want to use:

You'll need a valid OpenAI subscription. Copy .env.example to .env and configure:

cp .env.example .env

Then edit .env with your OpenAI credentials:

LLM_PROVIDER=openai
OPENAI_API_KEY=your_openai_api_key_here
OPENAI_MODEL=gpt-4o  # or gpt-4o-mini, gpt-5-mini, etc.

Run models locally using Ollama:

1. Install Ollama
2. Start Ollama server: ollama serve
3. Pull a model: ollama pull gemma3:4b
4. Configure your .env:

LLM_PROVIDER=ollama
OLLAMA_HOST=http://localhost:11434  # optional, defaults to this
OLLAMA_MODEL=gemma3:4b  # required: must be a model you have installed

Use Google's Gemini models. Configure your .env:

LLM_PROVIDER=gemini
GEMINI_API_KEY=your_gemini_api_key_here
GEMINI_MODEL=gemini-2.0-flash  # or gemini-1.5-flash, gemini-2.0-pro, etc.

For Vertex AI (Enterprise Google Cloud):

LLM_PROVIDER=gemini
GEMINI_API_KEY=your_gemini_api_key_here
GOOGLE_USE_VERTEXAI=true
GOOGLE_PROJECT_ID=your-gcp-project-id
GOOGLE_LOCATION=us-central1  # optional
GEMINI_MODEL=gemini-2.0-flash

git clone https://github.com/reflex-dev/reflex-chat.git
cd reflex-chat

To get started with Reflex, you'll need Python 3.10+. You can install dependencies using either pip (traditional) or uv (recommended for faster installs).

2. Install and sync dependencies:

   uv sync

Install all dependencies with the provided requirements.txt:

pip install -r requirements.txt

Initialize and run the app:

reflex init
reflex run

• OpenAI: OPENAI_MODEL (your chosen model, e.g., gpt-5-mini, gpt-4o, etc.)
• Ollama: OLLAMA_MODEL (your local model, e.g., gemma3:4b, llama3.2, etc.)
  • Note: Ollama requests use num_ctx=4096 by default for optimal context window
• Google: GEMINI_MODEL (your chosen model, e.g., gemini-2.0-flash, gemini-2.5-pro, etc.)

• 100% Python-based, including the UI, using Reflex
• Create and delete chat sessions
• Real-time streaming responses for all LLM providers (OpenAI, Ollama, Gemini)
• Multiple LLM support: OpenAI, Ollama (local), Google Gemini with Vertex AI
• Fully customizable - no web dev knowledge required
  • See https://reflex.dev/docs/styling/overview for more details

The following repo is licensed under the MIT License.