A comprehensive FastAPI-based application for HR document processing, intelligent question answering, web scraping, and automated email generation using RAG (Retrieval-Augmented Generation) technology.

• Document Processing: Upload and process various document types (PDF, TXT, Markdown)
• RAG Question Answering: Intelligent question answering based on your document corpus
• Web Scraping & Summarization: Scrape company websites and generate detailed summaries
• HR Email Generation: Automated generation of professional HR emails for various scenarios
• Vector Database Integration: Semantic search capabilities using Qdrant
• Multiple LLM Providers: Support for OpenAI, Cohere, and HuggingFace models
• Folder Upload: Batch upload of documents while preserving folder structure

• Python 3.10 or later
• MongoDB (via Docker)
• Qdrant Vector Database (via Docker)

1. Download and install MiniConda from the official documentation

2. Create a new environment:

conda create -n hr-toolkit python=3.10

3. Activate the environment:

conda activate hr-toolkit

pip install -r requirements.txt

cp .env.example .env

Edit the .env file with your configuration:

• Set your API keys (OpenAI/OpenRouter, Cohere, HuggingFace)
• Configure MongoDB connection settings
• Set LLM model preferences
• Adjust file upload and processing parameters

cd docker
cp .env.example .env

Update the Docker .env file with your credentials.

cd docker
docker compose up -d

This will start:

• MongoDB instance on port 27007
• Qdrant vector database

Start the FastAPI server:

uvicorn main:app --reload --host 0.0.0.0 --port 5000

The API will be available at http://localhost:5000

API documentation will be available at http://localhost:5000/docs

• GET /api/v1/ - Welcome endpoint with app information

• POST /api/v1/data/upload/{project_id} - Upload a single file
• POST /api/v1/data/upload-folder/{project_id} - Upload multiple files
• POST /api/v1/data/process/{project_id} - Process uploaded files into chunks

• POST /api/v1/nlp/index/push/{project_id} - Index processed chunks into vector database
• GET /api/v1/nlp/index/info/{project_id} - Get vector database collection information
• POST /api/v1/nlp/index/search/{project_id} - Semantic search in document collection
• POST /api/v1/nlp/index/answer/{project_id} - Get RAG-based answers to questions

• POST /api/v1/web-scraping/summarize - Scrape and summarize a website

• POST /api/v1/hr-email/generate - Generate professional HR emails

The application supports multiple LLM providers:

• OpenAI/OpenRouter: For generation (configured via OPENROUTER_API_KEY)
• Cohere: Alternative generation provider
• HuggingFace: For embeddings (sentence-transformers)

Qdrant is used for vector storage with configurable:

• Distance method (cosine, euclidean, dot)
• Embedding size
• Collection management

Supported formats:

• Text files (.txt)
• PDF documents (.pdf)
• Markdown files (.md)

Configurable parameters:

• Maximum file size
• Chunk size for processing
• Chunk overlap for context preservation