Use Ollama with a local LLM to query websites

ollama-web-qa.py

import argparse
import requests
from langchain.llms import Ollama
from langchain.document_loaders import WebBaseLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.embeddings import GPT4AllEmbeddings
from langchain.vectorstores import Chroma
from langchain.chains import RetrievalQA

def main(args):
    """
    Main function to execute the retrieval QA chain.

    Args:
        args.model (str): The model to be used by Ollama.
        args.website (str): URL of the website to retrieve data from.
        args.question (str): The query or question to be asked about the website content.
        args.base_url (str): The base URL for the Ollama.
        args.chunk_size (int): the number of tokens to chunk the document up into
        args.chunk_overlap (int): the number of tokens to overlap between chunks
        args.num_docs (int): the number of matching document chunks to retrive
        args.stop_matches (int): The number the of top matching document chunks to retrieve
        args.system (str): The LLM system message.
        args.temp (float): the LLM temperature setting

    Initialize Ollama with the specified model and base_url,
    load data from the given website, splits the text for efficient processing,
    creates a vector store from the document splits, and finally perform a
    retrieval-based question answering using the specified question.
    """
    check_server_availability(args.base_url)
    ollama = Ollama(
                base_url=args.base_url,
                model=args.model,
                system=args.system,
                temperature=args.temp,
                num_ctx=2048
	)

    loader = WebBaseLoader(args.website)
    data = loader.load()

    text_splitter = RecursiveCharacterTextSplitter(chunk_size=args.chunk_size, chunk_overlap=args.chunk_overlap)
    all_splits = text_splitter.split_documents(data)

    vectorstore = Chroma.from_documents(documents=all_splits, embedding=GPT4AllEmbeddings())
    qachain = RetrievalQA.from_chain_type(ollama, retriever=vectorstore.as_retriever(search_kwargs={"k": args.top_matches}))

    print("\n### model response:")
    print(qachain({"query": args.question}))


def check_server_availability(base_url):
    """
    Check if the Ollama server is running at the specified base URL.
    """
    try:
        response = requests.get(base_url)
        if response.status_code == 200:
            print(f"Successfully connected to the Ollama server at {base_url}.")
        else:
            print(f"Failed to connect to the Ollama server at {base_url}. Exiting.")
            sys.exit(1)
    except requests.ConnectionError:
        print(f"Could not connect to the Ollama server at {base_url}. Exiting.")
        sys.exit(1)

if __name__ == "__main__":
    parser = argparse.ArgumentParser(description="Run a retrieval QA chain on a specified website.")
    parser.add_argument("website", help="The website URL to retrieve data from")
    parser.add_argument("question", help="The question to ask about the website's content")
    parser.add_argument("--model", default="zephyr:latest", help="The model to use (default: zephyr:latest)")
    parser.add_argument("--base_url", default="http://localhost:11434", help="The base URL for the Ollama (default: http://localhost:11434)")
    parser.add_argument("--chunk_size", type=int, default=200, help="The document token chunk size (default: 200)")
    parser.add_argument("--chunk_overlap", type=int, default=50, help="The amount of chunk overlap (default: 100)")
    parser.add_argument("--top_matches", type=int, default=4, help="The number the of top matching document chunks to retrieve (default: 4)")
    parser.add_argument("--system", default="You are a helpful assistant.", help="The system message provided to the LLM")
    parser.add_argument("--temp", type=float, default=0.0, help="The model temperature setting (default: 0.0")

    args = parser.parse_args()

    main(args)

Thanks for the script!

Interesting choice to use both GPT4All and Ollama. Any particular reason why this cannot be done with just one or the other?