Pretzel aims to be a modern, open-source alternative to Jupyter Notebooks, enhancing the user experience with integrated AI capabilities for code generation, editing, and debugging. It targets data scientists and developers seeking an AI-augmented notebook environment with minimal disruption to their existing workflows.

How It Works

Pretzel is a fork of JupyterLab, incorporating AI features directly into the notebook interface. It leverages AI models for inline tab completion, code generation/editing via a Cmd+K shortcut, and a dedicated AI sidebar for chat and code search. The system automatically sends relevant notebook code as context to the AI, and users can reference variables using the '@' syntax. This approach aims to provide a seamless, AI-enhanced coding experience within a familiar Jupyter environment.

Quick Start & Requirements

• Install via pip: pip install pretzelai
• Run: pretzel lab
• Alternatively, use the hosted version: pretzelai.app
• For installation issues, ensure build tools (Microsoft Build Tools, build-essential python3-dev, or gcc on macOS) are installed.
• Docker installation instructions are provided.

Highlighted Details

• AI-powered inline tab completion and code generation/editing.
• AI sidebar for chat, code search, and debugging.
• Supports custom AI models (OpenAI, Anthropic, Ollama, Groq) via configuration.
• Roadmap includes real-time collaboration, SQL support, and a visual analysis builder.

Maintenance & Community

• Project is maintained by a two-person team.
• Feedback and bug reports are encouraged via GitHub Issues.
• Contact: founders@withpretzel.com
• Discord and Website links are provided in the README.

Licensing & Compatibility

• New code is licensed under AGPLv3. Original Jupyter code retains its BSD-3 license.
• AGPLv3 requires modifications and redistribution to be open-sourced under the same license.
• Commercial use is permitted; AGPLv3 obligations are triggered by public redistribution of modified versions. Contact founders@withpretzel.com for potential licensing discussions.

Limitations & Caveats

• The AGPLv3 license may be a concern for some organizations, though direct internal use is generally not affected.
• The project is actively developed by a small team, implying a potentially higher bus factor and faster iteration cycles.
• Azure Enterprise OpenAI models have not been tested.