Project Description: Legit-RAG

What is the project about?

Legit-RAG is a modular Retrieval-Augmented Generation (RAG) system. It's designed to answer questions by retrieving relevant information from a knowledge base and then generating a comprehensive answer.

What problem does it solve?

It addresses the limitations of large language models (LLMs) in providing accurate and up-to-date answers, especially for domain-specific or knowledge-intensive queries. By grounding the LLM's responses in retrieved, factual information, it reduces hallucinations and improves the reliability of answers. It also allows to build a question answering system on top of a custom knowledge base.

What are the features of the project?

• Modular Design: Built with independent, interchangeable components for each step of the RAG process.
• 5-Step RAG Workflow:
  1. Query Routing: Intelligently routes queries (ANSWER, CLARIFY, REJECT).
  2. Query Reformulation: Refines queries for better retrieval.
  3. Context Retrieval: Performs hybrid search (semantic + keyword).
  4. Completion Check: Evaluates if the context is sufficient.
  5. Answer Generation: Creates the final response with citations.
• Extensibility: Easily add new LLM providers, vector databases, document types, and search strategies.
• Hybrid Search: Combines semantic and keyword-based search for robust retrieval.
• Configurability: Customize LLM models, vector DB settings, completion thresholds, and more.
• API Endpoints: Provides REST API for adding documents and querying the system.
• Logging System: Event logging and visualization.

What are the technologies used in the project?

• FastAPI: Web framework for building the API.
• Qdrant: Vector database for storing and searching document embeddings.
• OpenAI: Large language models (LLMs) for various tasks (routing, reformulation, generation).
• Docker & Docker Compose: Containerization for easy deployment and environment management.
• Python 3.10+: Programming language.

What are the benefits of the project?

• Improved Accuracy: Provides more reliable and factual answers compared to standalone LLMs.
• Reduced Hallucinations: Grounding in retrieved context minimizes the risk of LLMs generating incorrect information.
• Customizability: Adaptable to different knowledge domains and use cases.
• Scalability: Modular design allows for scaling individual components as needed.
• Maintainability: Clear separation of concerns makes the system easier to understand and maintain.
• Extensibility: Easy integration of new technologies and features.

What are the use cases of the project?

• Question Answering Systems: Build chatbots or search interfaces that provide accurate answers from a specific knowledge base.
• Domain-Specific Search: Create search engines tailored to particular industries or fields.
• Content Generation: Assist in generating content by providing relevant context and information.
• Research Tools: Help researchers quickly find and synthesize information from large document collections.
• Customer Support: Automate responses to frequently asked questions using a company's internal documentation.
• Knowledge Management: Organize and access information within an organization more effectively.