File size: 5,237 Bytes
86d1671 c167991 86d1671 c167991 86d1671 c167991 86d1671 c167991 86d1671 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 |
import gradio as gr
from langchain_community.vectorstores import FAISS
from langchain_community.document_loaders import PyPDFLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain_community.embeddings import HuggingFaceEmbeddings
from langchain.chains import ConversationalRetrievalChain
from langchain_community.llms import HuggingFaceEndpoint
from langchain.chains import ConversationChain
from langchain.memory import ConversationBufferMemory
import os
api_token = os.getenv("HF_TOKEN")
# List of LLMs
list_llm = ["meta-llama/Meta-Llama-3-8B-Instruct", "mistralai/Mistral-7B-Instruct-v0.2"]
list_llm_simple = [os.path.basename(llm) for llm in list_llm]
# Load and split PDF documents
def load_doc(list_file_path):
loaders = [PyPDFLoader(x) for x in list_file_path]
pages = []
for loader in loaders:
pages.extend(loader.load())
text_splitter = RecursiveCharacterTextSplitter(chunk_size=1024, chunk_overlap=64)
doc_splits = text_splitter.split_documents(pages)
return doc_splits
# Create vector database
def create_db(splits):
embeddings = HuggingFaceEmbeddings()
vectordb = FAISS.from_documents(splits, embeddings)
return vectordb
# Initialize LLM chain
def initialize_llmchain(llm_model, temperature, max_tokens, top_k, vector_db):
if llm_model == "meta-llama/Meta-Llama-3-8B-Instruct":
llm = HuggingFaceEndpoint(
repo_id=llm_model,
huggingfacehub_api_token=api_token,
temperature=temperature,
max_new_tokens=max_tokens,
top_k=top_k,
)
else:
llm = HuggingFaceEndpoint(
huggingfacehub_api_token=api_token,
repo_id=llm_model,
temperature=temperature,
max_new_tokens=max_tokens,
top_k=top_k,
)
memory = ConversationBufferMemory(
memory_key="chat_history",
output_key='answer',
return_messages=True
)
retriever = vector_db.as_retriever()
qa_chain = ConversationalRetrievalChain.from_llm(
llm,
retriever=retriever,
chain_type="stuff",
memory=memory,
return_source_documents=True,
verbose=False,
)
return qa_chain
# Function to handle chatbot responses
def respond(
message,
history: list[tuple[str, str]],
system_message,
max_tokens,
temperature,
top_p,
vector_db,
llm_model,
):
# Initialize LLM chain if not already initialized
if not hasattr(respond, 'qa_chain'):
respond.qa_chain = initialize_llmchain(llm_model, temperature, max_tokens, top_p, vector_db)
# Format chat history
formatted_chat_history = []
for user_message, bot_message in history:
formatted_chat_history.append(f"User: {user_message}")
formatted_chat_history.append(f"Assistant: {bot_message}")
formatted_chat_history.append(f"User: {message}")
# Generate response using QA chain
response = respond.qa_chain.invoke({"question": message, "chat_history": formatted_chat_history})
response_answer = response["answer"]
if response_answer.find("Helpful Answer:") != -1:
response_answer = response_answer.split("Helpful Answer:")[-1]
return response_answer
# CSS for styling the interface
css = """
body {
background-color: #06688E; /* Dark background */
color: white; /* Text color for better visibility */
}
.gr-button {
background-color: #42B3CE !important; /* White button color */
color: black !important; /* Black text for contrast */
border: none !important;
padding: 8px 16px !important;
border-radius: 5px !important;
}
.gr-button:hover {
background-color: #e0e0e0 !important; /* Slightly lighter button on hover */
}
.gr-slider-container {
color: white !important; /* Slider labels in white */
}
"""
# Initialize database and LLM chain
def initialize_database_and_llm(list_file_obj, llm_option, max_tokens, temperature, top_p):
list_file_path = [x.name for x in list_file_obj if x is not None]
doc_splits = load_doc(list_file_path)
vector_db = create_db(doc_splits)
llm_name = list_llm[llm_option]
return vector_db, llm_name
# Gradio interface
demo = gr.ChatInterface(
respond,
additional_inputs=[
gr.Files(file_count="multiple", file_types=["pdf"], label="Upload PDF documents", visible=False),
gr.Radio(list_llm_simple, label="Available LLMs", value=list_llm_simple, visible=False),
gr.Slider(minimum=128, maximum=9192, value=4096, step=128, label="Max new tokens", visible=False),
gr.Slider(minimum=0.01, maximum=1.0, value=0.5, step=0.1, label="Temperature", visible=False),
gr.Slider(minimum=1, maximum=10, value=3, step=1, label="Top-k", visible=False),
],
css=css,
title="RAG PDF Chatbot",
description="Query your PDF documents using a Retrieval Augmented Generation (RAG) chatbot.",
)
# Preprocessing events
demo.preprocess(
initialize_database_and_llm,
inputs=["document", "llm_btn", "slider_maxtokens", "slider_temperature", "slider_topk"],
outputs=["vector_db", "llm_model"],
api_name="initialize",
)
if __name__ == "__main__":
demo.launch(share=True)
|