ReCARE AI Assistant - "First-Class" RAG System

A production-ready Retrieval-Augmented Generation (RAG) system built to answer questions accurately based on internal documents and equipment manuals.

🌟 Key Features

• Hybrid Native Search: Combines semantic searches (Dense Vectors) with exact keyword searches (BM25 Sparse Vectors) using Qdrant.
• Contextual Document Compression: Employs a HuggingFace Cross-Encoder (BGE-Reranker) to evaluate and prune irrelevant documents before passing context to the LLM, reducing hallucinations and inference time.
• LLM-Ops / Observability: Integrated with LangSmith for real-time monitoring of traces, token usage, latency, and costs.
• Multi-Interface:
  • 🤖 Telegram Bot: Asynchronous and fast conversational interface.
  • 🌐 Streamlit Web App: Clean, professional web UI with streaming responses.
• Containerized: Ready for zero-cost cloud deployments (Render, Koyeb, Streamlit Cloud) via Dockerfile.

🛠️ Technology Stack

• Vector Database: Qdrant (Local / Cloud)
• Primary LLM: Google Gemini 2.0 Flash (gemini-2.0-flash)
• Embedding Model: Google Generative AI Embeddings
• Reranker: cross-encoder/ms-marco-MiniLM-L-6-v2 (Optimized for fast CPU inference)
• Framework: LangChain

🚀 Getting Started

1. Requirements

• Python 3.11+
• Git

2. Installation

Clone the repository and install the dependencies:

git clone https://github.com/YOUR_GITHUB_USERNAME/YOUR_REPOSITORY_NAME.git
cd YOUR_REPOSITORY_NAME

# Create and activate a virtual environment
python -m venv venv
source venv/bin/activate  # On Windows use: venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

3. Environment Variables

Create a .env file in the root of the project with your API keys:

# Google AI Studio (Required for LLM and Embeddings)
GOOGLE_API_KEY=your_google_api_key_here

# Telegram Bot (Required only if running telegram_bot.py)
TELEGRAM_BOT_TOKEN=your_telegram_token_here

# LangSmith Observability (Optional but recommended)
LANGCHAIN_TRACING_V2=true
LANGCHAIN_ENDPOINT=https://api.smith.langchain.com
LANGCHAIN_API_KEY=your_langsmith_api_key_here
LANGCHAIN_PROJECT=ReCARE-RAG

# Reranker Toggle (Optional)
# Set to 'false' if deploying to low-RAM free-tier servers (e.g. Render Free)
USE_RERANKER=true

⚠️ Security Note: Never commit your .env file to GitHub. It is already safely included in the .gitignore.

4. Build the Knowledge Base

1. Place your .pdf, .txt, or .docx training files into the documentos_fonte/ directory.
2. Run the ingestion script to vectorize the documents into the local Qdrant database:

python ingest.py

Note: The first run will download embedding tools and the Reranker model locally.

5. Run the Application

You can run either the web interface or the Telegram bot:

Web UI (Streamlit):

streamlit run streamlit_app.py

Telegram Assistant:

python telegram_bot.py

☁️ Cloud Deployment (Zero Cost)

This project includes a Dockerfile pre-configured to run the Streamlit application in containerized environments.

1. Create a free Qdrant Cloud cluster.
2. Update config.py to point to your new cloud URL and API Key instead of the local path.
3. Push this repository to GitHub.
4. Connect the repository to Render.com (Web Service free tier) or Streamlit Community Cloud.
5. Add your .env secrets directly into the Cloud Provider's environment variables dashboard.

🤝 Contributing

Contributions are welcome! Please feel free to submit a Pull Request.