app.py

import gradio as gr
from huggingface_hub import InferenceClient
from torch import cuda, bfloat16
import torch
import transformers
from transformers import AutoTokenizer
from time import time
import chromadb
from chromadb.config import Settings
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.embeddings import HuggingFaceEmbeddings
from langchain.chains import RetrievalQA
from langchain.vectorstores import Chroma
from langchain.document_loaders import PyPDFLoader
import requests

"""
For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
"""
#client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
#notebook_login()

token_access = HF_API_TOKEN
headers = {"Authorization": f"Bearer {token_access}"}



model_id = 'mistralai/Mistral-7B-Instruct-v0.1'
model_config = transformers.AutoConfig.from_pretrained(
    model_id,
    max_new_tokens=200
)

model = transformers.AutoModelForCausalLM.from_pretrained(
    model_id,
    trust_remote_code=True,
    config=model_config,
    quantization_config=bnb_config,
    device_map='auto',
)
tokenizer = AutoTokenizer.from_pretrained(model_id)
query_pipeline = transformers.pipeline(
        "text-generation",
        model=model,
        tokenizer=tokenizer,
        torch_dtype=torch.float16,
        device_map="auto", max_new_tokens=200)


def respond(message, history, system_message, max_tokens, temperature, top_p):
    
    URL = "https://www.esmo.org/content/download/6594/114963/1/ES-Cancer-de-Mama-Guia-para-Pacientes.pdf"
    response = requests.get(URL)
    open("ES-Cancer-de-Mama-Guia-para-Pacientes.pdf", "wb").write(response.content)
    loader = PyPDFLoader("ES-Cancer-de-Mama-Guia-para-Pacientes.pdf")
    documents = loader.load()
    
    text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=20)
    all_splits = text_splitter.split_documents(documents)
    
    model_name = "sentence-transformers/paraphrase-multilingual-mpnet-base-v2"
    # model_kwargs = {"device": "cuda"}
    embeddings = HuggingFaceEmbeddings(model_name=model_name)#, model_kwargs=model_kwargs)
    
    vectordb = Chroma.from_documents(documents=all_splits, embedding=embeddings, persist_directory="chroma_db")

    pipeline=query_pipeline
    query = message
    docs = vectordb.similarity_search_with_score(query)
    context = []
    for doc, score in docs:
        if score < 7:
            doc_details = doc.to_json()['kwargs']
            context.append(doc_details['page_content'])
    if len(context) != 0:
        messages = [
            {"role": "user", "content": "Basándote en la siguiente información: " + "\n".join(context) + "\n Responde en castellano a la pregunta: " + query}]
        prompt = pipeline.tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
        outputs = pipeline(prompt, max_new_tokens=max_tokens, do_sample=True, temperature=temperature, top_k=50,
                           top_p=top_p)
        answer = outputs[0]["generated_text"]
        return answer[answer.rfind("[/INST]") + 8:], docs
    else:
        return "No tengo información para responder a esta pregunta", docs




"""
For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
"""
demo = gr.ChatInterface(
    respond,
    additional_inputs=[
        gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
        gr.Slider(
            minimum=0.1,
            maximum=1.0,
            value=0.95,
            step=0.05,
            label="Top-p (nucleus sampling)",
        ),
    ],
)


if __name__ == "__main__":
    demo.launch()