RepoRadar

From Repos to Readable Insights.

Comment

Inspiration

As developers, we’ve all faced the frustration of diving into a new or legacy repository without documentation — jumping between files, tracing imports, and piecing together logic line by line.
We wanted to make this process effortless. RepoRadar was inspired by that idea — a “radar” for repositories that scans, understands, and explains code intelligently.

What it does

RepoRadar automatically analyzes any Python codebase to:

  • Parse all files, functions, and classes using static analysis
  • Generate AI-based summaries and docstring suggestions
  • Build an interactive dependency graph showing how code components connect
  • Let users explore summaries and visualizations directly through a Streamlit interface

In short — it helps you understand a repository in minutes instead of hours.

How we built it

We built RepoRadar by combining AST parsing, LLM summarization, and interactive visualization:

  • 🧩 Parsed Python files using ast and asttokens to extract structure and function relationships
  • 🤖 Used OpenRouter’s LLM API (like Microsoft Phi-3.5 Mini and GPT-4o Mini) to summarize each code chunk
  • 📊 Visualized dependencies with PyVis and NetworkX
  • 🖥️ Built a stylish minimalistic and responsive UI with Streamlit
  • 💾 Added caching for summaries to improve performance

Challenges we ran into

Handling complex nested imports and circular dependencies.

  • Fixing the PyVis template rendering error (template.render) when generating graphs.
  • Avoiding cached summaries from older runs when switching between files.
  • Managing branch merges and dependency updates across development environments.

Each of these pushed us to refine the design and error handling to make RepoRadar functional and scalable.

Accomplishments that we're proud of

  • Fully integrated OpenRouter’s LLMs for automated summarization.
  • Solved caching and file-specific summarization issues.
  • Fixed PyVis rendering and dependency graph export.
  • Delivered a clean, minimal, and intuitive UI experience.
  • Built a project that genuinely helps developers save time and think smarter.

What we learned

  • Designing prompt templates for precise LLM code summarization.
  • Using AST-based parsing to analyze and extract Python code structures.
  • Debugging Streamlit state management for smoother interactivity.
  • Managing collaborative workflows with Git branching and merges.

What's next for RepoRadar

  • Extend support beyond Python — to JavaScript, Java, and C++.
  • Automate git checkout to compare two branches of the same repository and evaluate the differences
  • Integrate analytics to measure repository complexity and maintainability.

Built With

Share this project:

Updates