app.py

# https://python.langchain.com/docs/tutorials/rag/
import gradio as gr
from langchain import hub
from langchain_chroma import Chroma
from langchain_core.output_parsers import StrOutputParser
from langchain_core.runnables import RunnablePassthrough
from langchain_mistralai import MistralAIEmbeddings
from langchain_community.embeddings import HuggingFaceInstructEmbeddings
from langchain_text_splitters import RecursiveCharacterTextSplitter
from langchain_mistralai import ChatMistralAI
from langchain_community.document_loaders import PyPDFLoader
import requests
from pathlib import Path
from langchain_community.document_loaders import WebBaseLoader
import bs4
from langchain_core.rate_limiters import InMemoryRateLimiter
from urllib.parse import urljoin

rate_limiter = InMemoryRateLimiter(
    requests_per_second=0.1,  # <-- MistralAI free. We can only make a request once every second
    check_every_n_seconds=0.01,  # Wake up every 100 ms to check whether allowed to make a request,
    max_bucket_size=10,  # Controls the maximum burst size.
)

# # Get data from url
# url = 'https://camels.readthedocs.io/_/downloads/en/latest/pdf/'
# r = requests.get(url, stream=True)
# document_path = Path('data.pdf')

# document_path.write_bytes(r.content)
# # document_path = "camels-readthedocs-io-en-latest.pdf"
# loader = PyPDFLoader(document_path)
# docs = loader.load()

# # Load, chunk and index the contents of the blog.
# url = "https://lilianweng.github.io/posts/2023-06-23-agent/"
# loader = WebBaseLoader(
#     web_paths=(url,),
#     bs_kwargs=dict(
#         parse_only=bs4.SoupStrainer(
#             class_=("post-content", "post-title", "post-header")
#         )
#     ),
# )
# loader = WebBaseLoader(url)
# docs = loader.load()

# def get_subpages(base_url):
#     visited_urls = []
#     urls_to_visit = [base_url]

#     while urls_to_visit:
#         url = urls_to_visit.pop(0)
#         if url in visited_urls:
#             continue
        
#         visited_urls.append(url)
#         response = requests.get(url)
#         soup = bs4.BeautifulSoup(response.content, "html.parser")

#         for link in soup.find_all("a", href=True):
#             full_url = urljoin(base_url, link['href'])
#             if base_url in full_url and not full_url.endswith(".html") and full_url not in visited_urls:
#                 urls_to_visit.append(full_url)
#     visited_urls = visited_urls[1:]

#     return visited_urls

# base_url = "https://camels.readthedocs.io/en/latest/"
# # base_url = "https://carla.readthedocs.io/en/latest/"
# # urls = get_subpages(base_url)

urlsfile = open("urls.txt")
urls = urlsfile.readlines()
urls = [url.replace("\n","") for url in urls]
urlsfile.close()

# Load, chunk and index the contents of the blog.
loader = WebBaseLoader(urls)
docs = loader.load()

def format_docs(docs):
    return "\n\n".join(doc.page_content for doc in docs)

def RAG(llm, docs, embeddings):

    # Split text
    text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=200)
    splits = text_splitter.split_documents(docs)

    # Create vector store
    vectorstore = Chroma.from_documents(documents=splits, embedding=embeddings)

    # Retrieve and generate using the relevant snippets of the documents
    retriever = vectorstore.as_retriever()

    # Prompt basis example for RAG systems
    prompt = hub.pull("rlm/rag-prompt")

    # Create the chain
    rag_chain = (
        {"context": retriever | format_docs, "question": RunnablePassthrough()}
        | prompt
        | llm
        | StrOutputParser()
    )

    return rag_chain

# LLM model
llm = ChatMistralAI(model="mistral-large-latest", rate_limiter=rate_limiter)

# Embeddings
embed_model = "sentence-transformers/multi-qa-distilbert-cos-v1"
# embed_model = "nvidia/NV-Embed-v2"
embeddings = HuggingFaceInstructEmbeddings(model_name=embed_model)
# embeddings = MistralAIEmbeddings()

# RAG chain
rag_chain = RAG(llm, docs, embeddings)

def handle_prompt(message, history):
    try:
        # Stream output
        out=""
        for chunk in rag_chain.stream(message):
            out += chunk
            yield out
    except:
        raise gr.Error("Requests rate limit exceeded")

greetingsmessage = "Hi, I'm the CAMELS DocBot, I'm here to assist you with any question related to the CAMELS simulations documentation"
example_questions = [
                    "How can i read a halo file?",
                    "Which simulation suites are included in CAMELS?",
                    "Which are the largest volumes in CAMELS simulations?",
                    "How can I get the power spectrum of a simulation?"
                     ]

demo = gr.ChatInterface(handle_prompt, type="messages", title="CAMELS DocBot", examples=example_questions, theme=gr.themes.Soft(), description=greetingsmessage)#, chatbot=chatbot)

demo.launch()