🚀 manadi: take your genai data with you

Capture, organize, and fine-tune from your GenAI interactions effortlessly.

---

📚 Overview

manadi automatically captures and preserves your interactions with Generative AI models—including your prompts, AI responses, and feedback—in structured formats suitable for fine-tuning your own personalized GenAI models.

---

✨ Features

• Seamless Interaction Logging: Automatically records your prompts, responses from GenAI bots, and your optional feedback.
• Structured Data Storage: Saves interactions in a JSONL file, ready-to-use for fine-tuning.
• Middleware and Browser Integration: Supports capturing interactions via API middleware or browser extensions.
• Flexible Export: Export your logged data in formats supported by major fine-tuning platforms (OpenAI, HuggingFace).

---

📖 How It Works

• Prompt & Response Capture: Every interaction is captured and tagged.
• Optional Feedback: Add corrections or feedback directly after each generation.
• Automatic Structuring: Captures are structured in JSONL format, ensuring compatibility with standard fine-tuning tools.

Example of structured data entry:

{
  "prompt": "What is consistent hashing?",
  "response": "Consistent hashing is a method used in distributed systems...",
  "feedback": "Great explanation, but please simplify further."
}

---

🛠️ Tech Stack

• Backend: Python (FastAPI)
• Frontend (optional): Streamlit, React, Chrome Extension
• Storage: SQLite / PostgreSQL / JSONL files

---

📦 Installation & Setup

Clone the repository:

git clone https://github.com/matmulai/manadi.git
cd manadiai

Install dependencies:

pip install -r requirements.txt

Start the server:

uvicorn main:app --reload

---

🛠️ Usage

• Interact normally with your GenAI bots through the provided interface.
• Your interactions are automatically logged.
• Export captured data when you're ready for fine-tuning.

---

🎯 Fine-tuning Your GenAI Model

• Use the generated interactions.jsonl file with platforms like:
  • OpenAI Fine-tuning API
  • HuggingFace Trainer

Example fine-tuning command:

openai api fine_tunes.create -t interactions.jsonl

---

🤝 Contributing

We welcome contributions!

• Fork this repository
• Create your branch (git checkout -b feature/my-feature)
• Commit your changes (git commit -m 'Add my feature')
• Push your changes (git push origin feature/my-feature)
• Create a pull request

---

📄 License

Distributed under the MIT License. See LICENSE for more information.