Contributing

Contributing

Overview

We welcome contributions! TokenPrint is an ambitious project that requires expertise in React, WebGL, Python, and Deep Learning.

How to get started

1. Read the Docs: Familiarize yourself with the Architecture and Code Style.
2. Check the Ideas List: Look at docs/contributing-ideas.md in the repository. It contains a curated list of tasks sorted by difficulty, complete with file pointers and verification steps.
3. Pick an Issue: Comment on an open issue on GitHub so others know you are working on it.

Core Rules for PRs

• No Fabricated Data: You cannot use Math.random() or hardcode synthetic tensors to make a UI look nice. Your feature must read from the backend or the GGUF file.
• Pass the CI: Your code must pass black formatting for Python, Prettier/TypeScript checks for the frontend, and the trace_diff.py regression tests.
• Update Documentation: If you add a new API endpoint or a new geometric shape, you must update the relevant Markdown files in this Wiki and the docs/ folder.

Reporting Bugs

• Wrong Numbers: If you spot a number in the UI that contradicts the real model (e.g., the formula says LayerNorm but the model is Llama), this is a top-priority bug. Open an issue immediately.
• Security: If you find a security vulnerability (e.g., path traversal in the GGUF parser), do not open a public issue. See SECURITY.md for private reporting instructions.

Code of Conduct

Please adhere to the CODE_OF_CONDUCT.md found in the root of the repository. Be respectful, constructive, and scientifically rigorous.

Navigation

Previous	Home	Next
FAQ	Home	None