This Directory contains all the work related to RAG, LLM, AI Agents, Generative AI and Agentic AI.

1- ai-platform: This directory contains the code for the RAG system, including the implementation of the vector database, embedding generation, and response generation. 2- fraud-detection: This directory contains code related to a fraud detection system, which may utilize the RAG system for retrieving relevant information to assist in fraud detection. 3- hybrid_fraud_rag-agent_async: This directory contains code for an asynchronous agent that combines the RAG system with other techniques for fraud detection. 4- RAG-agent: This directory contains code for an agent that utilizes the RAG system for various tasks, potentially including fraud detection or other applications. 5- langchain: This directory contains code related to the LangChain framework, which may be used for building and deploying language models and agents.

====================================================================================================================================== RAG (Retrieval-Augmented Generation)

Installing Weaviate (Open-Source Vector Database)
Create docker-compose.yml
	 docker compose up -d 
	 http://localhost:8080/v1/meta or in curl curl http://localhost:8080/v1/meta
Create .env file for having openai_API_KEY
Create requirements.txt for the dependencies

Install python client 
	pip install weaviate-client openai tiktoken
Create Schema (Vectorized Class) 
Insert Documents (Embeddings are automatically created using OpenAI)
Query + Generate Answer (RAG)
	Retrieves top similar chunks
	Sends them to OpenAI GPT
	Returns generated answer

Create venv 
Activate venv
python <projectname.py>

We have : 
    Local vector DB
    Chunk before storing for better retrieval quality
    Metadata Filtering
    OpenAI embeddings 
    OpenAI generation
    Semantic search + Hybrid Search (Vector + BM25)
    Generative responses
    
Using model:
    Embeddings:
        text-embedding-3-small
    Generation:
        gpt-4o-mini (fast + cheap)
        gpt-4o (better quality)

======================================================== Enterprise Level Structure

App/ │ ├── docker/ │ ├── docker-compose.yml # Infra (Redis, Weaviate, Kafka) │ └── requirements.txt │ ├── data/ │ └── fraud_cases.json # Historical fraud dataset │ ├── src/ │ ├── app.py # FastAPI entrypoint │ ├── configurations.py # Env configs (API keys, URLs) │ │ │ ├── orchestrator/ │ │ └── agent.py # Main routing logic (RAG vs API) │ │ │ ├── llm_core/ │ │ └── openai_client.py # OpenAI integration │ │ │ ├── fraud_rag/ │ │ ├── embeddings.py # Embedding generation │ │ ├── weaviate_client.py # Vector DB operations │ │ ├── rag_reasoner.py # Context + prompt builder │ │ │ ├── cache/ │ │ └── redis_cache.py # Caching layer │ │ │ ├── ingestion/ │ │ ├── kafka_consumer.py # Kafka ingestion service │ │ ├── kafka_producer.py # Send documents/events │ │ └── ingest_data.py # Initial data load │ │ │ ├── services_detect/ │ │ └── fraud_api.py # Mock payment/order APIs │ │ │ └── utils/ │ └── helpers.py # Common utilities └── tests/ │ └── test_api.py # test_api │ └── README.md