File size: 3,886 Bytes
0799221 a309d4e 0799221 1098b47 0799221 350ee28 0799221 a309d4e 0799221 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 |
import pinecone
from langchain.vectorstores import Pinecone
from langchain.embeddings.openai import OpenAIEmbeddings
from langchain.text_splitter import CharacterTextSplitter
from langchain.chat_models import ChatOpenAI
from langchain.chains import ConversationalRetrievalChain
import os
from langchain.docstore.document import Document
from langchain.document_loaders import PyPDFLoader
import gradio as gr
import openai
import backoff
openai.api_key = os.environ["OPENAI_API_KEY"]
# initialize pinecone
pinecone.init(api_key=os.environ['PINECONE_API_KEY'],
environment='us-east1-gcp')
# delete the vectors in the index if there are any
index = pinecone.Index('asesura')
delete_response = index.delete(delete_all=True)
# load the documents
loader = PyPDFLoader("SaludClasico2023.pdf")
pages = loader.load_and_split()
n_pages = [page.page_content.replace("CÓDIGO CLAUSULADO - 01/01/2023 - 1411 - P - 35 - F-14-11-0090-214- D00I. CÓDIGO NOTA TÉCNICA – 01/01/2023 - 1411 - NT-P - 35 - N-14-11-0090-031", "") for page in pages]
print(len(n_pages))
# If a page has less than 100 characters, append it to the next one
for i in range(len(n_pages)-1):
if len(n_pages[i]) < 100:
n_pages[i+1] = n_pages[i] + n_pages[i+1]
n_pages[i] = ""
n_pages = [page for page in n_pages if len(page) > 100]
pages_summarized = []
#summarize each page
for i, chunk in enumerate(n_pages):
messages = [
{"role": "system", "content": "Summarize this text in the same language as the user's input."}]
messages.append({"role": "user", "content": chunk})
response = openai.ChatCompletion.create(
model="gpt-4",
messages=messages,
temperature=.2,
top_p=1,
frequency_penalty=0,
presence_penalty=0
)
pages_summarized.append(
response["choices"][0]["message"]['content'].strip())
pages = [Document(page_content = page_summarized, metadata={"page": i}) for i, page_summarized in enumerate(pages_summarized)]
# split the documents into chunks
text_splitter = CharacterTextSplitter(chunk_size=1000, chunk_overlap=100)
documents = text_splitter.split_documents(pages)
embeddings = OpenAIEmbeddings()
# initialize pinecone
pinecone.init(
api_key="15f4e36a-00e9-46ad-8dcb-f01e5b2f7568", # find at app.pinecone.io
environment="us-east1-gcp" # next to api key in console
)
index_name = "asesura"
# create the new vectors
docsearch = Pinecone.from_documents(
documents, embeddings, index_name=index_name)
# create the conversational chain
qa = ConversationalRetrievalChain.from_llm(ChatOpenAI(
temperature=0), docsearch.as_retriever(), return_source_documents=True)
@backoff.on_exception(backoff.expo, openai.error.RateLimitError)
@backoff.on_exception(backoff.expo, openai.error.APIConnectionError)
def predict(query):
response = openai.ChatCompletion.create(
model="gpt-4",
messages=[{"role": "system", "content": "Change the user's question so it is correctly made, with the correct punctuation, grammar, and spelling. The question is being made to an insurance policy called 'Sura Seguro de Salud Clásico' by a company named Sura. Return the question in the same language as the user."},
{"role": "user", "content": query}],
temperature=0.1)
refined_q = response.choices[0].message.content
print(refined_q)
response = qa(
{"question": refined_q, "chat_history": []})['answer']
return response
pregunta = gr.Textbox(label="Pregunta", placeholder="Escribe tu pregunta")
respuesta = gr.Textbox(
label="Respuesta")
gr.Interface(predict, pregunta, respuesta, title="Asesura",
description="Bienvenido a tu asesor personal de seguros Sura. Pregunta lo que necesites saber sobre el Seguro de Salud Clásico de Sura 2023").launch()
|