appchat.py

# -*- coding: utf-8 -*-
"""chatbot_with_memory (1).ipynb

Automatically generated by Colaboratory.

Original file is located at
    https://colab.research.google.com/drive/1sIEqI5-wciuiYOdlEYwBkTPUIlvMEzkF
"""

!pip install chromadb==0.4.6
!pip install pydantic==1.10
!pip install sentence-transformers

!pip install huggingface_hub

!pip install transformers

from langchain.document_loaders import TextLoader  #for textfiles
from langchain.text_splitter import CharacterTextSplitter #text splitter
from langchain.embeddings import HuggingFaceEmbeddings #for using HugginFace models
from langchain.vectorstores import FAISS
from langchain.chains.question_answering import load_qa_chain
from langchain.chains.question_answering import load_qa_chain
from langchain import HuggingFaceHub
from langchain.document_loaders import UnstructuredPDFLoader  #load pdf
from langchain.indexes import VectorstoreIndexCreator #vectorize db index with chromadb
from langchain.chains import RetrievalQA
from langchain.document_loaders import UnstructuredURLLoader  #load urls into docoument-loader
from langchain.chains.question_answering import load_qa_chain
from langchain import HuggingFaceHub
import os
huggingfacehub_api_token = os.environ.get("HUGGINGFACEHUB_API_TOKEN")


pip install pypdf

from langchain.document_loaders import PyPDFLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter

#import csvfrom langchain.document_loaders import PyPDFLoader
# Load the PDF file from current working directory
loader = PyPDFLoader("/content/Document sans titre (5).pdf")
# Split the PDF into Pages
pages = loader.load_and_split()

#import from langchain.text_splitter import RecursiveCharacterTextSplitter
# Define chunk size, overlap and separators
text_splitter = RecursiveCharacterTextSplitter(
    chunk_size= 128,
    chunk_overlap=64,
    separators=['\n\n', '\n', '(?=>\. )', ' ', '']
)
docs  = text_splitter.split_documents(pages)

from langchain.embeddings import HuggingFaceEmbeddings
embeddings = HuggingFaceEmbeddings()

pip install faiss-gpu

#Create the vectorized db
# Vectorstore: https://python.langchain.com/en/latest/modules/indexes/vectorstores.html
from langchain.vectorstores import FAISS
db = FAISS.from_documents(docs, embeddings)

llm=HuggingFaceHub(repo_id="google/flan-t5-xxl", model_kwargs={"temperature":1, "max_length":1000000, "max_new_tokens": 500})
chain = load_qa_chain(llm, chain_type="stuff")

#QUERYING
query = "quelles sont les villes les facultees de medcine  ?"
docs = db.similarity_search(query)
chain.run(input_documents=docs, question=query)

from langchain.chains import RetrievalQA
qa = RetrievalQA.from_chain_type(llm=llm, chain_type="stuff",
retriever=db.as_retriever(search_kwargs={"k": 3}))

query = "donner moi plus  des information sur les facultees de medcine?"
qa.run(query)

query = "What is the meaning of Descriptive Data Analysis?"
qa.run(query)#import csv

repo_id = 'google/flan-t5-xxl'  # has 3B parameters: https://huggingface.co/lmsys/fastchat-t5-3b-v1.0
llm = HuggingFaceHub(huggingfacehub_api_token=os.environ["HUGGINGFACEHUB_API_TOKEN"],
                     repo_id=repo_id,
                     model_kwargs={'temperature':0.5, 'max_length':256})

query1 = "Bonjour, je suis zaynab ,j'ai des questions a vous "
query2 = "j'habite a marrakech. tu sait son pays?"
query3 = "quel est mon prenom?"
query4 = "ou j'habite"

pip install langchain --upgrade

from langchain import HuggingFaceHub
from langchain.chains import ConversationChain

"""### Conversation Buffer memory"""

from langchain.chains.conversation.memory import ConversationBufferMemory
 # Adjust the import path accordingly
memory = ConversationBufferMemory()
conversation_buf = ConversationChain(
    llm=llm,
    memory=memory)

print("input: ",query1)
conversation_buf.predict(input=query1)

print("input: ",query2)
conversation_buf.predict(input=query2)

memory.load_memory_variables({})

print("input: ",query3)
conversation_buf.predict(input=query3)

print("input: ",query4)
conversation_buf.predict(input=query4)

print(memory.buffer)

"""### Conversation Buffer Window Memory"""

from langchain.memory import ConversationBufferWindowMemory

memory2 = ConversationBufferWindowMemory(k=2)
conversation_buf2 = ConversationChain(
    llm=llm,
    memory=memory2
)

print("input: ",query1)
conversation_buf2.predict(input=query1)

print("input: ",query2)
conversation_buf2.predict(input=query2)

print("input: ",query3)
conversation_buf2.predict(input=query3)

print(memory2.buffer)

"""### Conversation Summary Memory"""

from langchain.memory import ConversationSummaryBufferMemory

memory3 = ConversationSummaryBufferMemory(llm=llm, max_token_limit=80)
conversation_buf3 = ConversationChain(
    llm=llm,
    memory=memory3
)

print("input: ",query1)
conversation_buf3.predict(input=query1)

print("input: ",query2)
conversation_buf3.predict(input=query2)

print("input: ",query3)
conversation_buf3.predict(input=query3)

memory3.load_memory_variables({})

"""### Chat PDF with Memory

Updated version of Pydantic package (dependency of chromadb) has changed leaving chromadb, incompatible: here are the possible solutions: [import error chromadb](https://github.com/langchain-ai/langchain/issues/1957)  || Install specific versions of chromadb and pydantic while the bug is resolved

![image.png](data:image/png;base64,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
"""

!pip install pypdf

import langchain
import chromadb

import os
import getpass

from langchain.document_loaders import PyPDFLoader  #document loader: https://python.langchain.com/docs/modules/data_connection/document_loaders
from langchain.text_splitter import RecursiveCharacterTextSplitter  #document transformer: text splitter for chunking
from langchain.embeddings import HuggingFaceEmbeddings
from langchain import PromptTemplate
from langchain.vectorstores import Chroma #vector store
from langchain import HuggingFaceHub  #model hub
from langchain.chains import RetrievalQA

from langchain.memory import ConversationBufferMemory

#loading the API key
import getpass
import os
os.environ['HUGGING_FACE_HUB_API_KEY'] = getpass.getpass('Hugging face api key:')

path = input("Enter PDF file path: ")#"C:/Users/Sourav/Downloads/pdf"
loader = PyPDFLoader(path)
pages = loader.load()

#number of pages
len(pages)

splitter = RecursiveCharacterTextSplitter(chunk_size=256, chunk_overlap=10)
docs = splitter.split_documents(pages)

tokens = docs
num_tokens = len(tokens)
print("Nombre de jetons :", num_tokens)

for token in tokens:
    print(token)

embeddings = HuggingFaceEmbeddings()
doc_search = Chroma.from_documents(docs, embeddings)

print(doc_search)

query = "Quelle sont les Facultees existent ?"
similar_docs = doc_search.similarity_search(query, k=3)

print(similar_docs)

query = "donner moi des information ecole nationale d'Industrie Minérale ?"
similar_docs = doc_search.similarity_search(query, k=10)

repo_id = 'google/flan-t5-xxl'  # has 3B parameters: https://huggingface.co/lmsys/fastchat-t5-3b-v1.0
llm = HuggingFaceHub(huggingfacehub_api_token=os.environ['HUGGING_FACE_HUB_API_KEY'],
                     repo_id=repo_id,
                     model_kwargs={'temperature':1, 'max_length':10000000000, "max_tokens":1000000000})

template = """
Use the following context (delimited by <ctx></ctx>) and the chat history (delimited by <hs></hs>) to answer the question:
------
<ctx>
{context}
</ctx>
------
<hs>
{history}
</hs>
------
{question}
Answer:
"""
prompt = PromptTemplate(
    input_variables=["history", "context", "question"],
    template=template,
)

memory = ConversationBufferMemory(
    memory_key="history",
    input_key="question"
)

retrieval_chain = RetrievalQA.from_chain_type(llm,
                                              chain_type='stuff',
                                              retriever=doc_search.as_retriever(),
                                              chain_type_kwargs={
                                                  "prompt": prompt,
                                                  "memory": memory
                                              })

query = " donner moi les villes de ces facultees de medcine? "
retrieval_chain.run(query)

query = "donner moi des information sur  Facultees de medcine  ?"
retrieval_chain.run(query)

memory.load_memory_variables({})