AI Code Intelligence

Chat with Any GitHub Repository

Understand architecture, dependencies, implementation details, and code structures instantly through source-cited codebase search.

No login required

Source-cited answers

Commit references

RAG Pipeline Visualization

Multi-Layered Code Intelligence

repository-root/

├──src/

├──auth.ts

└──router.ts

└──package.json

Git Branch: main

ast-chunker-embeddings

Chunk #12

class AuthController...

Embedding

[0.12, -0.45, 0.89...]

ChromaDB Indexed

768 DimensionsGemini Embedding

gemini-query-synthesis

Q: How is security handled?

A: JWT checks inside auth.ts:25

Source-Cited Verification

Pipeline Flow

How RepoChat Works

Convert any public GitHub repository into a semantic knowledge base in three simple steps.

01Connect Repository

Provide a public GitHub link. RepoChat clones the repository code sandboxed and maps the file tree.

02Index & Vectorize

Extract source classes and modules into chunks, computing vector representations using Gemini Embeddings.

03Source-Cited Chat

Ask codebase questions. Receive complete descriptions complete with exact file locations and line references.

Capabilities

Premium Developer Intelligence

Explore complex or unfamiliar codebases in minutes using precise AI semantic retrieval.

Automatic Repository Summarization

Receive a structured README-style overview detailing exactly what the codebase does, its folder structure, external dependencies, and execution entrypoints.

Readme generationModule maps

Browser-Only Privacy

Your chat logs and history remain cached locally in your browser storage. No proprietary code details are persisted on the server db.

Commit-Anchored Citations

All source code references point to immutable Git commits. Links stay stable even if developers push code changes to GitHub later.

Context-Aware AI Chat Assistant

Ask specific implementation questions ("Where is user auth handled?", "How does the db transaction session pool work?"). The assistant searches vector space and explains concepts with actual snippets.

Gemini Pro APISource-grounded synthesis

Interactive Demo

Test Drive Curated Codebases

Select one of our curated public GitHub repositories below to instantly pre-fill the search tool.

Ingestion Guidelines

✓Small repositories (< 100 files) analyze in seconds and are optimal for test chats.

✓Large codebases might hit public free-tier Gemini API limitations during vectorizing.

✓Supported file extensions: python (.py), javascript (.js, .jsx), typescript (.ts, .tsx), go (.go), etc.

✓For large private enterprise repositories, run RepoChat locally using your own private API key.

Python Packaging Demo

pypa/sampleproject

A tiny official sample Python project for testing package metadata, dependencies, and structure.

Compact Python Library

pallets/markupsafe

A small Python library useful for exploring package layout, dependency structure, and source citations.

Python CLI Framework

pallets/click

A clean CLI framework repository useful for testing architecture, modules, and implementation questions.

Testing Plugin Example

pytest-dev/pytest-cov

A focused Python plugin repository useful for exploring test configuration, package structure, and coverage tooling.

Python Packaging Utility

pypa/packaging

A practical Python packaging utility repository useful for dependency and versioning questions.

Example Inquiries to Try

What does this project do?

How is the repository structured?

What dependencies are required?

Where is authentication implemented?

How does the database layer work?

What files are most important?

How is configuration handled?

How is versioning managed?

Design Blueprint

RAG Ingestion & Query Pipeline

RepoChat coordinates frontend requests, sandboxed background workers, vector search indexes, and Gemini synthesis.

Cloning & Parsing

Worker clones public repo source tree and uses AST parsers to split files into document chunks.

Vector Embedding

Code segments are translated into dense vector points via the Gemini Embeddings API.

Semantic Indexing

Embeddings and file tree metadata are indexed into an isolated local ChromaDB collection.

RAG Query Chat

Gemini LLM synthesizes code responses by context-matching the user question against retrieved vector chunks.

System Isolation Promise: Each repository ingestion creates a distinct SQLite record and ChromaDB collection. Your workspace remains entirely isolated, with chat histories stored on your browser client.

Technologies

Built on a State-of-the-Art AI Stack

RepoChat utilizes modern full-stack web architectures to deliver instant AI code search and analysis.

Frontend Interface

Next.js 14React 18TypeScriptTailwind CSS

Backend Service

FastAPIPython 3.11SQLAlchemySQLite3

AI & Retrieval

Gemini ProGemini EmbeddingsChromaDBRAG Pipeline

UX & Infrastructure

Lucide IconsCSS VariablesAsync WorkersRate Limiting

Gemini Free-Tier Environment Limits

This public deployment utilizes Google's free-tier Gemini endpoints. Please take note of the following runtime constraints to ensure smooth operations:

Repository SizeUnder 500 filesRecommended to prevent token budget exhaustions.

API Quota limitsShared Free TierHeavy concurrent user traffic may cause transient 429 quota retries.

Hourly limit60 Queries / hourProtects resources from bot abuse or scraping queries.

Need unlimited requests or analyzing massive code repositories? Run RepoChat locally in your terminal. Follow the startup guide in our GitHub repository to connect your private Google Gemini API key.

Recently Indexed Repositories

Continue exploring repositories you have already analyzed in this browser.

No indexed repositories found in this browser. Enter a public GitHub repository link above to analyze your first codebase.