AI Software Engineer Assistant

An intelligent AI-powered coding workspace that understands repositories, explains code, answers architectural questions, generates tests, detects bugs, and assists developers with deep codebase intelligence.

Think: GitHub Copilot + Codebase Chat + AI Engineering Workspace.

🚀 Features

Core Features

Repository ingestion and indexing
AI-powered codebase chat
Semantic code search
AST-aware code chunking
Streaming AI responses
Repository-aware question answering
Authentication and user workspaces
Persistent chat history

Advanced Features

Unit test generation
Bug detection and fix suggestions
Architecture diagram generation
Dependency graph visualization
Multi-language repository support
Intelligent code explanations
PR review assistance
Symbol-aware retrieval

🧠 Why This Project Matters

This project demonstrates:

AI Engineering
Retrieval-Augmented Generation (RAG)
Vector Databases
Backend Architecture
Developer Tooling
Distributed Systems Thinking
Modern Full-Stack Engineering
Production AI Systems

It is designed as a flagship portfolio project for software engineering and AI engineering roles.

🏗️ System Architecture

                        ┌─────────────────┐
                        │   Next.js UI    │
                        │ React Frontend  │
                        └────────┬────────┘
                                 │
                           HTTPS/API
                                 │
                    ┌────────────▼────────────┐
                    │       FastAPI API       │
                    │ AI Orchestration Layer  │
                    └───────┬───────┬─────────┘
                            │       │
               ┌────────────┘       └────────────┐
               │                                 │
       ┌───────▼────────┐              ┌────────▼────────┐
       │ PostgreSQL     │              │ Redis Cache     │
       │ chats/users    │              │ sessions/cache  │
       └────────────────┘              └─────────────────┘

               ┌────────────────────────────────┐
               │ Vector Database (Qdrant)       │
               │ embeddings + semantic search   │
               └────────────────────────────────┘

                            │
                   ┌────────▼────────┐
                   │ Repo Ingestion  │
                   │ Git + Parsers   │
                   └────────┬────────┘
                            │
                 ┌──────────▼──────────┐
                 │ AST + Embedding     │
                 │ Processing Pipeline │
                 └─────────────────────┘

⚙️ Tech Stack

Backend

FastAPI
Python
PostgreSQL
Redis
Celery / Background Workers
Docker

AI & Retrieval

OpenAI / Local LLMs
Qdrant / Pinecone
Embeddings
RAG Pipelines
AST Parsing
Tree-sitter

Frontend

Next.js
React
TailwindCSS
TypeScript

Deployment

Docker
Vercel
Railway / Render / Fly.io

📁 Project Structure

ai-software-engineer-assistant/
│
├── backend/
│   ├── api/
│   ├── ingestion/
│   ├── parsers/
│   ├── embeddings/
│   ├── rag/
│   ├── services/
│   ├── workers/
│   ├── models/
│   └── main.py
│
├── frontend/
│   ├── app/
│   ├── components/
│   ├── hooks/
│   ├── lib/
│   └── styles/
│
├── docker/
├── infra/
├── scripts/
├── docs/
└── README.md

🔥 MVP Goals

Version 1

Upload/index repositories
Ask questions about code
Semantic search
Streaming chat responses
File-aware citations
Authentication
Persistent workspaces

🧩 Future Roadmap

Version 2

Test generation
AI code fixes
PR review assistant
Architecture graph generation

Version 3

Multi-repository understanding
Autonomous coding agents
Team collaboration
IDE integration

🧠 How It Works

1. Repository Ingestion

The system clones repositories and parses files while ignoring unnecessary directories such as:

node_modules
build/
dist/
binaries
.git/

2. Intelligent Chunking

Instead of splitting files arbitrarily, the assistant uses:

AST parsing
Function-level chunking
Class-level chunking
Symbol-aware extraction

This dramatically improves retrieval quality.

3. Embedding Pipeline

Each chunk is embedded using modern embedding models and stored inside a vector database.

Metadata stored:

file path
language
symbols
repository
function names

4. Retrieval-Augmented Generation (RAG)

When a user asks a question:

Query embedding is generated
Relevant chunks are retrieved
Context is reranked
LLM generates repository-aware answers
Streaming response is returned to frontend

🔒 Authentication

Supports:

JWT Authentication
OAuth (GitHub planned)
Workspace isolation
Session management

⚡ Streaming Responses

The platform uses:

FastAPI StreamingResponse
Server-Sent Events (SSE)

to provide real-time AI responses similar to ChatGPT and Cursor.

📊 Planned AI Capabilities

Explain code
Detect bugs
Generate unit tests
Refactor suggestions
Security analysis
Architecture visualization
Dependency analysis
AI pair programming

🐳 Running Locally

Clone Repository

git clone https://github.com/your-username/ai-software-engineer-assistant.git
cd ai-software-engineer-assistant

Backend Setup

cd backend

python -m venv venv
source venv/bin/activate

pip install -r requirements.txt

uvicorn main:app --reload

Frontend Setup

cd frontend

npm install
npm run dev

Docker

docker-compose up --build

🧪 Example Questions

"How does authentication work?"
"Explain the repository architecture."
"Generate tests for the payment service."
"Find potential bugs in the auth module."
"Which services depend on Redis?"
"Show me API flow for user login."

📈 Engineering Highlights

This project showcases:

Advanced RAG systems
AI infrastructure engineering
Semantic retrieval systems
Backend API design
Full-stack development
Streaming systems
Distributed architecture
Developer tooling engineering

🎯 Ideal Use Cases

AI pair programming
Large repository understanding
Developer onboarding
Technical documentation
Code review assistance
Engineering productivity

🛠️ Planned Integrations

GitHub
GitLab
Bitbucket
VSCode Extension
Slack
CI/CD pipelines

📜 License

MIT License

👨‍💻 Author

Built as a flagship AI engineering project focused on modern developer tooling and intelligent code understanding systems.

Name		Name	Last commit message	Last commit date
Latest commit History 21 Commits
.github/workflows		.github/workflows
backend		backend
frontend		frontend
.gitignore		.gitignore
README.md		README.md
docker-compose.yml		docker-compose.yml
security_audit.md		security_audit.md

Folders and files

Latest commit

History

Repository files navigation