beckn-llm-layer

This project implements an Adaptive RAG (Retrieval-Augmented Generation) system using LangChain and various other libraries.

Project Structure

adaptive_rag/
├── README.md
├── requirements.txt
├── main.py
├── config.py
├── environment.py
├── index.py
├── llms.py
├── graph.py
├── utils/
│   ├── document_processing.py
│   ├── routing.py
│   ├── grading.py
│   └── search.py

Installation

To install the required packages, run:

pip install -r requirements.txt

Configuration

Set your API keys in config.py. You can either directly set them in the file or export them as environment variables before running the project.

Example configuration in config.py:

import os

OPENAI_API_KEY = os.getenv("OPENAI_API_KEY", "<your-api-key>")
COHERE_API_KEY = os.getenv("COHERE_API_KEY", "<your-api-key>")
TAVILY_API_KEY = os.getenv("TAVILY_API_KEY", "<your-api-key>")
LANGCHAIN_TRACING_V2 = os.getenv("LANGCHAIN_TRACING_V2", "true")
LANGCHAIN_ENDPOINT = os.getenv("LANGCHAIN_ENDPOINT", "https://api.smith.langchain.com")
LANGCHAIN_API_KEY = os.getenv("LANGCHAIN_API_KEY", "<your-api-key>")

Running the Project

To run the project, use the following command:

python main.py

Overview of Files

• main.py: The entry point of the project. It sets the environment, builds the index, compiles the graph, and runs the workflow.
• config.py: Contains configuration settings for API keys and other environment variables.
• environment.py: Sets the necessary environment variables for the project.
• index.py: Handles the creation of the document index using LangChain.
• llms.py: Defines the language models and related configurations.
• graph.py: Builds and compiles the state graph for the Adaptive RAG workflow.
• utils/: A directory containing helper modules:
  • document_processing.py: Functions for document retrieval, generation, grading, and query transformation.
  • routing.py: Functions for routing questions and deciding the next steps in the workflow.
  • grading.py: Functions for grading the generated answers against the documents and questions.
  • search.py: Functions for performing web searches.

Example Usage

To run an example workflow, ensure your API keys are set, and then execute main.py. The project will process the input question through the Adaptive RAG workflow, retrieving relevant documents, generating answers, and grading the output.

License

This project is licensed under the MIT License. See the LICENSE file for more information.

Acknowledgments

This project utilizes the following libraries:

• LangChain
• Tiktoken
• Langchain-OpenAI
• Langchain-Cohere
• Langchainhub
• ChromaDB
• Langgraph
• Tavily-Python

Special thanks to the developers and contributors of these libraries for their invaluable tools and resources.