CodeSahayatri is a repository-structure visualization and explanation tool designed to help interns and early-career developers understand large codebases faster.

It lets a user enter a GitHub repository URL, fetches the repository tree and recent commit history from the GitHub API, builds a client-side graph of directories and files, and provides an interactive chat assistant that explains architecture using the repository structure as context.

This project is a useful example of adjacent experience for long-context and complex-reasoning tooling work because it already combines:

• large-repository ingestion from external systems
• graph-based structural representations of codebases
• multi-file context assembly
• commit and contributor metadata analysis
• LLM-powered repository explanation grounded in code structure

While this project is not a benchmark dataset or an AST-based static analysis pipeline, it demonstrates practical work on repository understanding, architecture surfacing, and context construction for code reasoning workflows.

The current application flow is:

1. Repository URL input
   • The user submits a GitHub repository URL through the frontend.
2. Repository and commit retrieval
   • The app fetches the repository tree and recent commit details from the GitHub API.
3. Graph construction
   • File and directory paths are converted into nodes and parent-child links.
4. Interactive visualization
   • The repository structure is rendered as an interactive force-directed graph.
5. Context-aware explanation
   • The chat assistant builds repository context from the graph and selected node relationships, then asks an LLM to explain the architecture.

If you are positioning this project in a proposal, the most credible framing is:

• it shows experience navigating large repositories programmatically
• it shows experience turning repository structure into machine-readable context
• it shows experience building developer-facing interfaces for code understanding
• it shows experience connecting structural code context to LLM reasoning

That makes it a strong related-work example for projects involving:

• repository curation
• long-context code understanding
• benchmark task design around multi-file reasoning
• analysis of how agents fail when context spans many files and subsystems

To keep the project description accurate, this repository currently does not include:

• Python AST static analysis
• a Flask backend
• import or call-graph dependency extraction
• benchmark dataset generation or evaluation pipeline integration

1. Copy .env.example to .env:

cp .env.example .env

2. Add your API keys to .env:

• VITE_GITHUB_TOKEN: Your GitHub personal access token
• VITE_OPENAI_API_KEY: Your OpenAI API key

npm install
npm run dev