Update documentation

[TanGentleman]

Here's a quick sample:

YouTwo

YouTwo is a fully MCP-compatible memory layer system that enables seamless ingestion, indexing, and retrieval of digital memories—such as documents, notes, and audio—for individuals and teams. Designed for integration with modern AI workflows, YouTwo exposes its API as LLM tool calls, making it easy to build advanced knowledge management and productivity solutions.

Features

• MCP Compatibility: Exposes all core functions as LLM tool calls for easy integration with agent frameworks and AI assistants.
• Efficient Data Ingestion & Indexing: Supports rapid ingestion and indexing of heterogeneous data (documents, notes, audio).
• Hybrid RAG Search: Integrates Vectara API for high-quality, hybrid retrieval-augmented generation (RAG) over your data.
• Knowledge Graph Management: Dynamically manages entities and relationships, supporting updates and real-time queries.
• Serverless & Scalable: Built on Convex for serverless backend and efficient, scalable data operations.
• User-Friendly Frontend: Gradio-based interface for intuitive memory search, knowledge graph visualization, and demoing capabilities.
• Agent Integration: Utilizes Huggingface smolagents for agent-based tool calling and dynamic user interaction.

Tech Stack

• Backend & Serverless: Convex (TypeScript)
• Frontend: Gradio (Python)
• AI/ML: Huggingface smolagents (Python), Vectara API (hybrid RAG)
• Knowledge Graph: Custom management and update logic

Getting Started

Prerequisites

• Node.js & npm (for Convex backend)
• Python 3.8+ (for Gradio frontend and AI components)
• Vectara API key
• MCP-compatible LLM or agent framework (optional, for advanced integration)

Installation

1. Clone the repository:
   git clone https://github.com/yourusername/youtwo.git
   cd youtwo

2. Backend (Convex):

   • Install dependencies and set up your Convex project.
   • Configure environment variables for database and Vectara API.

3. Frontend (Gradio):

   • Install Python dependencies:
     pip install -r requirements.txt

   • Run the Gradio app:
     python app.py

4. AI/ML Components:

   • Set up Huggingface smolagents and configure tool calls as needed.

Usage

• Ingest Data: Upload documents, notes, or audio via the Gradio interface or API.
• Query Memories: Use the frontend or API to search, retrieve, and visualize your knowledge graph.
• LLM Tool Calls: Integrate with any MCP-compatible agent or LLM to expose YouTwo’s functions as tool calls.

Example: Exposing a Function as an LLM Tool Call

Example: Registering a memory search function for LLM tool calls. Start the MCP server using run_mcp.py.
from youtwo.api (currently convex_mcp) import search_memories

@app.tool
def find_meeting_notes(query: str):
return search_memories(query)

Contributing

Contributions are welcome! Please open an issue or submit a pull request.

License

MIT License

Contact

For questions or support, please create visit our GitHub issues page.

---

Let me know if you want to add usage examples, API docs, or a demo GIF!