G4Flow

A powerful, open-source AI chat assistant built with Streamlit and G4F

G4Flow is a versatile chat application that provides an intuitive interface for interacting with various AI models — text conversations, image generation, image variations, audio (TTS), video generation, and document analysis. It leverages the G4F (GPT4Free) framework to offer access to multiple AI models through a clean, user-friendly interface.

✨ Features

💬 Chat Capabilities

• Grouped Model Selection: Models organized by provider (OpenAI, Google, Meta, DeepSeek, Mistral, and more) for easy navigation across 100+ models
• System Message Templates: Customize AI behavior with predefined templates (Default, Creative, Academic, Programming, Friendly, Fitness) or create custom ones
• Image Analysis: Upload images for the AI to analyze and discuss using vision models
• Web Search Integration: Search the web using DuckDuckGo with source citations displayed as clickable references
• Response Streaming: View AI responses as they are generated in real-time
• Native Reasoning Display: Shows the reasoning process for DeepSeek-R1 and similar models with thinking duration, using g4f's native Reasoning API with regex fallback
• Conversation History Toggle: Enable/disable conversation history while maintaining system instructions
• JSON Mode: Force structured JSON output from models for programming use cases
• Max Tokens & Stop Sequences: Fine-tune response length and stop conditions
• Token Usage Tracking: Per-message and cumulative session token counts (prompt, completion, reasoning)
• Document Upload & Analysis: Upload TXT, PDF, DOCX, CSV, JSON, Markdown, and code files to analyze in chat

🖼️ Image Generation

• Multiple Image Models: 390+ image models including DALL-E 3, Midjourney, Flux, Stable Diffusion, and more
• Prompt Improvement: Automatically enhance image prompts for better results and translate non-English prompts to English
• Image History: View previously generated images in the chat history

🎨 Image Variation

• Upload & Vary: Upload an existing image and generate variations of it using any supported image model

🎵 Audio (TTS)

• Text-to-Speech: Convert text to audio using multiple voices (alloy, echo, fable, onyx, nova, shimmer)
• Transcript Display: View transcripts alongside generated audio when available

🎬 Video Generation

• AI Video: Generate videos from text prompts using 27+ video models
• Auto-Download: Optionally save generated videos locally

🔌 Custom Provider

• Connect Your Own LLM: Use Ollama, LM Studio, or any OpenAI-compatible API by entering a base URL and optional API key

💾 Media Management

• Save Generated Files: Toggle saving of generated images, audio, and videos to generated_media/ folder
• Chat Export: Export full chat history as JSON with embedded images and token usage data

🚀 Getting Started

Prerequisites

• Python 3.8 or higher
• pip (Python package installer)

Setup

1. Clone this repository:

   git clone https://github.com/kqlio67/g4flow.git
   cd g4flow

2. Create and activate a virtual environment (recommended):

   # On Windows
   python -m venv venv
   venv\Scripts\activate
   
   # On macOS/Linux
   python -m venv venv
   source venv/bin/activate

3. Install the required dependencies:

   pip install -r requirements.txt

4. Run the Streamlit application:

   streamlit run app.py

5. Open your web browser and navigate to the URL displayed in the terminal (typically http://localhost:8501).

📋 Repository Structure

g4flow/
├── app.py                  # Main Streamlit application code
├── requirements.txt        # Project dependencies
├── generated_media/        # Auto-generated media files (images, audio, video)
├── .gitignore             # Git ignore file
├── LICENSE                # MIT License
├── CONTRIBUTING.md        # Guidelines for contributors
└── README.md             # Project documentation

📦 Dependencies

Package	Purpose
streamlit	Web UI framework
g4f	AI model client (chat, images, video, audio)
g4f[search]	Web search via DuckDuckGo
pypdf2	PDF document reading
pdfplumber	PDF document reading (fallback)
python-docx	DOCX document reading
docx2txt	DOCX document reading (fallback)

🔧 Advanced Configuration

• System Messages: Add new templates in the SYSTEM_MESSAGES dictionary in app.py
• Custom Provider: Enter an OpenAI-compatible API URL in the Advanced Settings sidebar section
• Max Tokens / Stop Sequences: Configure under Advanced Settings in the sidebar
• JSON Mode: Enable in chat settings to force structured JSON output
• Styling: Edit the CSS in apply_custom_styling() in app.py

🤝 Contributing

Contributions are welcome! Please see our Contributing Guidelines for details.

📄 License

Licensed under the MIT License — see the LICENSE file for details.

👥 Contributors

🙏 Acknowledgments

• Built with the G4F framework
• Powered by Streamlit
• Thanks to all the contributors who have helped improve this project

⚠️ Disclaimer

This project is for educational purposes only. Please ensure you comply with the terms of service for all AI providers and APIs used through the G4F framework.

---

If you find this project useful, please consider giving it a star ⭐

Made with ❤️ by kqlio67