DevDocs AI 🤖

A production-grade AI system that analyzes Python code, generates high-quality documentation, and evaluates its quality with actionable feedback.

🚀 Try Live Demo

---

🎯 What It Does

Upload Python code or a .py file and get:

• A structured project-style README
• Clear documentation of functions and logic
• A quality score (0–100)
• Concrete suggestions to improve code documentation

Example Flow:

User uploads .py file
System parses code using AST (no execution)
System analyzes structure + semantics
System generates documentation via LLM
System evaluates quality across 6 dimensions
System returns docs + score + suggestions

---

✨ Features

✅ AST-Based Code Analysis – Safe parsing without executing code
✅ AI Documentation Generator – README-style output using Groq + Llama 3.3
✅ Quality Scoring Engine – 0–100 score with detailed feedback
✅ File Upload Support – Upload .py files directly
✅ REST API – Built with FastAPI + interactive Swagger docs
✅ Production Architecture – Modular, typed, validated, and deployable

---

🏗️ Architecture Overview

The system is structured into clean, single-responsibility modules:

• Code Parser (code_parser.py) – AST-based structural code analysis
• LLM Service (llm_service.py) – Handles prompting and Groq API interaction
• Quality Scorer (quality_scorer.py) – Evaluates documentation quality
• Analysis Service (analysis_service.py) – Orchestrates full pipeline
• Schemas (responses.py) – Type-safe API contracts using Pydantic
• API Layer (main.py) – FastAPI app and routing

---

Data Flow

User Uploads Code
↓
🧠 AST Parser (structure extraction)
↓
📋 Analysis Service (orchestration)
↓
🤖 LLM Service (documentation generation)
↓
⭐ Quality Scorer (0–100 evaluation)
↓
✅ Final Output: Docs + Score + Suggestions

---

🛠️ Tech Stack

• Python 3.12
• FastAPI
• Groq API (Llama 3.3 70B)
• Pydantic
• Python AST
• Uvicorn
• Render (deployment)

---

📊 Quality Scoring System

Documentation is evaluated across 6 weighted dimensions:

Category	Weight	What It Evaluates
Code Quality	20%	Docstrings, type hints in source
Completeness	20%	All functions/classes covered
Structure	15%	README sections present
Examples	15%	Practical usage examples
Clarity	15%	Readability and explanation
Coverage	15%	Params, returns, behavior documented

Grades:
A (90–100) • B (80–89) • C (70–79) • D (60–69) • F (<60)

---

🧪 Example Usage

API Request

curl -X POST "https://devdocs-ai-jjof.onrender.com/analyze" \
  -H "Content-Type: application/json" \
  -d '{"code": "def add(a: int, b: int) -> int:\n    return a + b"}'

Python Client

import requests

response = requests.post(
    "https://devdocs-ai-jjof.onrender.com/analyze",
    json={"code": "def greet(name): return f'Hello {name}'"}
)

print(response.json())

---

📡 API Endpoints

• POST /analyze – Analyze raw Python code
• POST /analyze/file – Upload and analyze .py file
• GET /health – Health check endpoint

Validation Rules:

• .py files only
• Max size: 1MB
• UTF-8 encoding
• Proper HTTP errors (400, 413, 500)

---

🔗 Links

• Live API: https://devdocs-ai-jjof.onrender.com

---

Built as part of my applied AI engineering journey 🚀