README.md

metadata

title: SHL Assessment Recommender
emoji: 🚀
colorFrom: blue
colorTo: indigo
sdk: docker
pinned: false
license: mit
app_port: 7860

SHL Assessment Recommender

Overview

This project implements a RAG-based recommendation system for SHL assessments. It uses:

• Embeddings: sentence-transformers/all-MiniLM-L6-v2
• Vector Store: FAISS
• Reranking: Google Gemini 2.5 Flash (via LangChain)
• API: FastAPI

Setup

1. Install Dependencies:

   python3 -m venv venv
   source venv/bin/activate
   pip install -r requirements.txt
   

   (Note: Dependencies include sentence-transformers, faiss-cpu, fastapi, uvicorn, langchain, langchain-google-genai, python-dotenv)

2. Environment Variables: Create a .env file in the root directory:

   GOOGLE_API_KEY=your_api_key_here
   

3. Ingest Data:

   python shl_recommender/src/ingest.py
   

Running the API

To start the server (runs on port 8001):

# From the project root
source venv/bin/activate
python shl_recommender/src/app.py

Evaluation

To calculate Recall@10 on the training set:

python shl_recommender/src/metrics.py

API Usage

Endpoint: POST /recommend Body:

{
  "query": "Looking for a Java developer with good communication skills"
}

Example:

curl -X POST http://localhost:8001/recommend \
     -H "Content-Type: application/json" \
     -d '{"query": "Java developer"}'