Interviewee - Helen Reece (LangGraph + RAG + React)

A production-grade, local-first interactive job screening interview platform. The application uses an asynchronous LangGraph state workflow loop coupled with a ChromaDB vector store context retrieval component to evaluate candidate resumes using local Ollama LLM instances.

---

📁 System Architecture & Directory Layout

interviewee/
├── backend/                       # Python Core API Ecosystem
│   ├── api/
│   │   └── main.py                # FastAPI endpoints & CORS definitions
│   ├── core/
│   │   ├── agent/
│   │   │   ├── graph.py           # LangGraph State Graph & node routers
│   │   │   └── trigger.py         # Native terminal conversation loop
|   |   ├── ingestion/
│   │   │   ├── parser.py          # Multi-layout PDF & Word Extraction service
│   │   │   └── chroma_service.py  # Chroma API Vector store sliding window ingestion 
│   │   └── config.py              # Pydantic Settings Path Resolver
│   ├── tests/
│   │   │   ├── test_agent.py      # Node architecture tests
│   │   │   ├── test_ingestion.py  # Ingestion pipeline tests
│   │   │   └── test_api.py        # API routing mock test engines
│   ├── pyproject.toml             # Local dynamic editable packaging manifest
│   └── .env                       # Machine environment parameters
├── datastore/
│   ├── uploads/                   # Bucket for storing the uploaded resumes
│   └── chroma_db/                 # Persistent storage vector database files
└── web/                           # Front-End Web Client Engine
    ├── src/
    │   ├── App.tsx                # Single-pane Interactive Chat UI Matrix
    │   ├── main.tsx               # Bootstrap client mounter
    │   └── index.css              # Structural baseline styles
    ├── index.html                 # Browser layout structural template
    ├── package.json               # Node Package configuration scripts
    ├── tsconfig.json              # TypeScript compilation rules
    └── vite.config.ts             # Vite bundler parameters

---

🛠️ Environmental Settings (backend/.env)

Configure the project workspace boundaries by creating a .env configuration template file inside the backend/ path directory:

CHROMA_DB_PATH=../datastore/chroma_db
CHROMA_COLLECTION_NAME=resumes
OLLAMA_URL=http://localhost:11434
OLLAMA_MODEL=llama3.2

---

🚀 Execution & Quick-Start Walkthrough

Prerequisites

Ensure your local Ollama server runtime instance is running and has the model pre-pulled on your system architecture:

ollama run llama3.2

1. Fire up the Backend API Server

Navigate to the backend/ folder directory, install editable packagers, and kickstart the uvicorn workspace server:

cd backend
pip install -e .
pip install uvicorn pytest pydantic-settings
uvicorn api.main:app --reload

• Interactive Document API Portal: Once up, test raw endpoint integrations through the native Swagger documentation dashboard at http://127.0.0.

2. Boot Up the React App Web Client

Open a separate, dedicated operational terminal matrix, move into the client folder path, install node dependencies, and run the developer bundle:

cd web
npm install
npm run dev

• Web UI Interface Portal: Access the client interface panel directly by pointing your local desktop browser to http://localhost:5173.

---

🧪 Automated Testing Procedures

Our test suites run with standard Python isolation paradigms using pytest to prevent disk collisions. Execute the test modules right from the backend/ repository folder level:

• Switch to tests:

  cd backend/tests

• Run LangGraph workflow step logic states and router routing validations:

  pytest tests/test_agent.py

• Run full mock FastAPI client integration tests:

  pytest tests/test_api.py

• Run internal resume parsing string text extractions and chunking checks:

  pytest tests/test_ingestion.py

• Run all tests:

  pytest