Update main.py

main.py

@@ -3,101 +3,41 @@ import pandas as pd
 
 import io
 from flask import Flask, request
-from twilio.twiml.messaging_response import MessagingResponse
-from langchain.llms import GooglePalm
-import pandas as pd
-#from yolopandas import pd
-import os
-from langchain.embeddings import GooglePalmEmbeddings
-# a class to create a question answering system based on information retrieval
-from langchain.chains import RetrievalQA
-# a class for splitting text into fixed-sized chunks with an optional overlay
-from langchain.text_splitter import RecursiveCharacterTextSplitter
-# a class to create a vector index using FAISS, a library for approximate nearest neighbor search
-from langchain.vectorstores import FAISS
-# a class for loading PDF documents from a directory
-from langchain.document_loaders import PyPDFDirectoryLoader
 from langchain.chains.question_answering import load_qa_chain
-from langchain.chains import ConversationalRetrievalChain
-from langchain.schema.vectorstore import VectorStoreRetriever
-import google.generativeai
-
+from langchain import PromptTemplate, LLMChain
+from langchain_experimental.agents.agent_toolkits import create_pandas_dataframe_agent
+from langchain.chat_models import ChatOpenAI
+from langchain.agents.agent_types import AgentType
+from langchain.chat_models import ChatOpenAI
+import pandas as pd
+from langchain.llms import OpenAI
 from dotenv import load_dotenv
+import google.generativeai as palm
+from langchain.llms import GooglePalm
+import json
 
-load_dotenv()
-
-
-
-
-def get_pdf_text(pdf_docs):
-    text=""
-    for pdf in pdf_docs:
-        pdf_reader= PdfReader(pdf)
-        for page in pdf_reader.pages:
-            text+= page.extract_text()
-    return  text
-
-# load PDF files from a directory
-loader = PyPDFDirectoryLoader("documents/")
-data = loader.load()
-
-# print the loaded data, which is a list of tuples (file name, text extracted from the PDF)
-#print(data)
-
-# split the extracted data into text chunks using the text_splitter, which splits the text based on the specified number of characters and overlap
-text_splitter = RecursiveCharacterTextSplitter(chunk_size=10000, chunk_overlap=20)
-
-text_chunks = text_splitter.split_documents(data)
-
-# print the number of chunks obtained
-#print(len(text_chunks))
-
-embeddings = GooglePalmEmbeddings(google_api_key=os.environ['PALM'])
 
-# create embeddings for each text chunk using the FAISS class, which creates a vector index using FAISS and allows efficient searches between vectors
-vector_store = FAISS.from_documents(text_chunks, embedding=embeddings)
+from dotenv import load_dotenv
 
-#print(type(vector_store))
+load_dotenv()
 
 
+app = Flask(__name__)
 
-def ask_pdfs(user_question):
+@app.route("/predict", methods=["POST"])
+def bot(json_table, user_question):
     load_dotenv()
-
-
-
+    #
+    df = pd.DataFrame(json_table)
+    #df['Profit'] = df['Profit'].apply(lambda x: "R{:.1f}".format((x)))
+    #df['Revenue'] = df['Revenue'].apply(lambda x: "R{:.1f}".format((x)))
+    #llm = ChatOpenAI(model_name='gpt-3.5-turbo-0613', temperature=0, openai_api_key=os.getenv('OPENAI_API_KEY'))
     llm = GooglePalm(temperature=0, google_api_key=os.environ['PALM'])
-
-    # Create a question answering system based on information retrieval using the RetrievalQA class, which takes as input a neural language model, a chain type and a retriever (an object that allows you to retrieve the most relevant chunks of text for a query)
-    retriever = VectorStoreRetriever(vectorstore=vector_store)
-    qa = RetrievalQA.from_llm(llm=llm, retriever=retriever)
-    response =qa.run(user_question)
-    #print("Response:",response)
-
+    #agent = create_pandas_dataframe_agent(llm, df, verbose=True, agent_type=AgentType.OPENAI_FUNCTIONS)
+    agent = create_pandas_dataframe_agent(llm, df, agent="structured_chat-zero-shot-react-description", verbose=True)
+    response = agent.run(user_question)
     return response
 
-app = Flask(__name__)
-
-@app.route("/", methods=["POST"])
-def whatsapp():
-
-    # user input
-    user_msg = request.values.get('Body', '').lower()
-
-    # creating object of MessagingResponse
-    response = MessagingResponse()
-
-    print(user_msg)
-
-    # User Query
-    q = user_msg
-
-    answer = ask_pdfs(q)
-
-    response.message(answer)
-
-    return str(response)
-
 
 if __name__ == "__main__":
     app.run(debug=True,host="0.0.0.0", port=7860)