RAAN

[R]eliable [A]gentic [A]I [N]otebook

This repository contains the specifications for the .RAAN project, which aims to create a reliable agentic AI notebook file system. The specifications outline the architecture of the file extension.

RAAN Specifications

Take a look at the SPECS.md file for detailed specifications of the .RAAN project, including file extension, usage instructions, and the technology stack used.

Usage

To use the RAAN notebook system, follow these steps:

1. Create a new file with the .RAAN extension (follow the specifications).
2. Structure the file according to the specifications outlined above.
3. Use a compatible text editor or IDE to edit the file.
4. Save the file to maintain the structure and metadata.

TODO - Tech Debt

• Update /scripts to use python scripts instead of bash scripts.
• Update ollama to localai in jupyter notebooks.
• Use XML tags for metadata in the RAAN notebook files?
• Replace HTML data-tag with XML tags for template.
• Research the best practices for structuring file extensions for AI notebooks.
• Complete the specifications for the file extension, check what is the "industry standard" for AI notebooks.
• Implement a plugin for VS Code to improve the editing experience.
• Create artifacts for the RAAN project, including example notebooks and documentation.
• Integrate with Nutrino Notebooks for MCP capabilities (PoC this first!).
• Setup python lintering and formatting.
• Setup python testing framework.
• Create a custom template webpdf to export notebook as PDF with custom styles (e.g. fix the margins!).
• Handle context to use vector databases for knowledge bases, instead of just using the knowledge base as a list of strings.

Resources

• Jupyter Notebook Templates
• Mermaid Documentation