Helix — AI Code Intelligence Platform

Understand, visualize, and reason over any software system.

🌐 Live Demo • 🔌 API Docs • Features • Architecture • Quick Start

---

🚀 What is Helix?

Modern software projects contain thousands of files, APIs, services, and dependencies. Understanding how a large codebase works is hard — for new developers, reviewers, and even the people who wrote it.

Helix transforms any codebase into a queryable knowledge graph. Upload a ZIP or connect a GitHub repo, and Helix parses every file with tree-sitter AST analysis, builds a Neo4j relationship graph, generates semantic embeddings, and lets you explore and query your code with natural language.

Unlike RAG-based code assistants that search text, Helix reasons over structure — function call graphs, class inheritance, module dependencies, and import chains — giving answers that understand how your code actually works.

---

🔗 Links

🌐 Frontend	https://helix-phi-beige.vercel.app
🔌 Backend API + Swagger	https://helix-iq1l.onrender.com/docs
📦 GitHub	https://github.com/harshitayadavv/helix

---

✨ Features

🔍 Intelligent Repository Analysis

• Upload ZIP or clone any public GitHub repository
• Automatic language detection across Python, JavaScript, TypeScript, Java, C++
• Real-time processing progress via WebSocket streaming
• Incremental indexing with status tracking

🌳 AST & Knowledge Graph

• Full tree-sitter AST parsing per language
• Extracts functions, classes, imports, exports, function calls
• Builds Neo4j graph with node types: File, Function, Class, Module
• Relationships: CONTAINS, CALLS, IMPORTS, INHERITS

🤖 AI Code Assistant

• LangGraph agent powered by Groq llama-3.3-70b-versatile
• Reasons over the knowledge graph, not just text
• Ask questions like:
  • "Explain the authentication flow"
  • "Where is JWT implemented?"
  • "Which files would break if I remove Redis?"
  • "Find all dead code"

🔎 Hybrid Semantic Search

• FAISS vector search + Neo4j keyword fallback
• Filter by node type (Function / Class / File / Module) and language
• Search history tracked per repository in Redis

⚠️ Impact Analysis

• Select any file or function → compute blast radius via Neo4j BFS
• Depth 1: direct dependents (red), Depth 2: indirect (orange), Depth 3: transitive (yellow)
• Risk score: Low / Medium / High / Critical
• Groq-generated plain-English explanation

🛡️ Security Analyzer

• Hardcoded secrets detection (API keys, passwords, tokens)
• SQL injection patterns, XSS vulnerabilities
• Unsafe imports (eval, exec, pickle, os.system)
• Weak auth patterns (MD5/SHA1 password hashing)

🧩 Code Smell Detection

• God Classes (>10 methods), Long Methods (>50 lines)
• Circular Dependencies (Neo4j cycle detection)
• Dead Code (functions with no incoming CALLS edges)
• Duplicate Logic (same function name across files)

📊 AI Project Health Score

• Overall score (0–100) with letter grade
• 6 sub-scores: Architecture · Maintainability · Complexity · Security · Performance · Documentation

⚡ Performance Analyzer

• N+1 query pattern detection, blocking calls in async functions
• Expensive nested loops, unnecessary object creation in loops

📖 Documentation Generator

• AI-generated README, API docs, Architecture overview, Onboarding guide
• Mermaid diagram for module dependencies — copy or download as .md

🔄 Repository Timeline

• Git log: commits, authors, timestamps, files changed
• Hotspot files + contributor graph

🎯 Interactive Graph Explorer

• React Flow canvas — zoomable, draggable, filterable
• BFS path tracing between any two nodes, node detail drawer on click

---

🏗️ Architecture

                    GitHub URL / ZIP Upload
                            │
                            ▼
                   Repository Processor
                            │
        ┌───────────────────┼───────────────────┐
        ▼                   ▼                   ▼
   AST Parser         Dependency Builder    Git Analyzer
  (tree-sitter)      (import resolver)     (GitPython)
        │                   │                   │
        └──────────────┬────┴───────────────────┘
                       ▼
             Knowledge Graph Builder
             (Neo4j + PostgreSQL)
                       │
        ┌──────────────┼──────────────┐
        ▼              ▼              ▼
 Vector Store     Search Engine    AI Agents
   (FAISS)       (Hybrid Search)  (LangGraph)
        └──────────────┬──────────────┘
                       ▼
                FastAPI Backend
                       │
               WebSocket Streaming
                       │
                       ▼
          Next.js Interactive Dashboard

---

🖥️ Tech Stack

Layer	Technology
Frontend	Next.js 14, TypeScript, Tailwind CSS, React Flow, Framer Motion, ShadCN UI
Backend	FastAPI, Python 3.11, Celery, Uvicorn
Graph DB	Neo4j AuraDB
Relational DB	PostgreSQL (SQLAlchemy async)
Cache / Queue	Redis
AI Agent	LangGraph, Groq llama-3.3-70b-versatile
Parsing	tree-sitter (Python, JS, TS, Java, C++)
Embeddings	sentence-transformers (all-MiniLM-L6-v2)
Vector Search	FAISS
Git Analysis	GitPython
Hosting	Vercel (frontend) · Render (backend)
Infra	Docker, Docker Compose

---

⚡ Quick Start

Prerequisites

• Docker + Docker Compose
• Python 3.11 (conda recommended)
• Node.js 18+
• Groq API key (free at console.groq.com)

1. Clone the repo

git clone https://github.com/harshitayadavv/helix.git
cd helix

2. Backend setup

cd helix-backend
cp .env.example .env
# Edit .env — add GROQ_API_KEY and NEO4J credentials

docker compose up -d neo4j postgres redis   # start infrastructure

conda create -n helix python=3.11 -y
conda activate helix
pip install -r requirements.txt

# Terminal 1
uvicorn app.main:app --reload --port 8001

# Terminal 2
celery -A celery_worker.celery_app worker --loglevel=info --pool=solo

3. Frontend setup

cd helix-frontend
cp .env.local.example .env.local
# Set NEXT_PUBLIC_API_URL=http://localhost:8001
npm install && npm run dev

4. Open Helix

Service	URL
Frontend	http://localhost:3000
API + Swagger	http://localhost:8001/docs
Neo4j Browser	http://localhost:7474

---

📡 API Reference

All endpoints require X-API-Key header after registration.

# Register
curl -X POST https://helix-iq1l.onrender.com/api/v1/auth/register \
  -H "Content-Type: application/json" \
  -d '{"email": "you@example.com", "password": "yourpass", "name": "Your Name"}'

# Upload a repository
curl -X POST https://helix-iq1l.onrender.com/api/v1/repositories/upload \
  -H "X-API-Key: your_key" \
  -F "file=@myrepo.zip"

# Ask the AI
curl -X POST https://helix-iq1l.onrender.com/api/v1/ai/ask \
  -H "X-API-Key: your_key" \
  -H "Content-Type: application/json" \
  -d '{"repo_id": "your_repo_id", "question": "Explain the authentication flow"}'

Full docs: https://helix-iq1l.onrender.com/docs

---

📁 Project Structure

helix/
├── helix-backend/
│   ├── app/
│   │   ├── api/routes/       # FastAPI route handlers
│   │   ├── core/
│   │   │   ├── ai/           # LangGraph agent, embeddings, prompts
│   │   │   ├── analysis/     # Security, smells, health, performance, impact
│   │   │   ├── docs/         # Documentation generator
│   │   │   ├── git/          # Git analyzer (GitPython)
│   │   │   ├── graph/        # Neo4j client, graph builder
│   │   │   ├── parser/       # tree-sitter AST parser
│   │   │   └── search/       # Hybrid FAISS + Neo4j search
│   │   ├── db/               # PostgreSQL + SQLAlchemy
│   │   └── services/         # Repo processor, Celery, WebSocket
│   └── docker-compose.yml
│
└── helix-frontend/
    └── src/
        ├── app/
        │   ├── repo/[id]/    # Per-repo workspace
        │   ├── dashboard/    # Repo list + uploader
        │   └── auth/         # Login + signup
        ├── components/       # 20+ reusable components
        └── lib/              # api.ts, websocket.ts, utils.ts

---

🗺️ Roadmap

• Multi-agent architecture reviewer
• PR review assistant
• AI Change Simulator
• VS Code extension
• Voice interaction
• Team Knowledge Hub

---

👩‍💻 Author

Built by Harshita Yadav — B.Tech, IIIT Kota

---

_{Built with ❤️ by Harshita — Helix is not just a project. It's a platform.}