RAG-Chatbot / app.py
Ultronprime's picture
Update app.py
d93c60d verified
raw
history blame
17 kB
import gradio as gr
import os
from pathlib import Path
import json
import csv
import pandas as pd
from tqdm import tqdm
api_token = os.getenv("HF_TOKEN")
from langchain_community.vectorstores import FAISS
from langchain_community.document_loaders import PyPDFLoader, TextLoader, CSVLoader, JSONLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain_community.vectorstores import Chroma
from langchain.chains import ConversationalRetrievalChain
from langchain_community.embeddings import HuggingFaceEmbeddings
from langchain_community.llms import HuggingFacePipeline
from langchain.chains import ConversationChain
from langchain.memory import ConversationBufferMemory
from langchain_community.llms import HuggingFaceEndpoint
import torch
# import spaces
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 documents of various types
def load_doc(list_file_path, progress=gr.Progress()):
doc_splits = []
progress(0, desc="Preparing to load documents")
total_files = len(list_file_path)
for i, file_path in enumerate(list_file_path):
progress((i/total_files) * 0.5, desc=f"Loading {Path(file_path).name}")
file_ext = Path(file_path).suffix.lower()
try:
# PDF documents
if file_ext == '.pdf':
loader = PyPDFLoader(file_path)
pages = loader.load()
doc_splits.extend(split_documents(pages))
# Text-based documents
elif file_ext in ['.txt', '.md', '.py', '.js', '.html', '.css']:
loader = TextLoader(file_path)
documents = loader.load()
doc_splits.extend(split_documents(documents))
# CSV files
elif file_ext == '.csv':
loader = CSVLoader(file_path)
documents = loader.load()
doc_splits.extend(split_documents(documents))
# JSON files
elif file_ext in ['.json', '.jsonl']:
# For JSON, we need to determine if it's JSON or JSONL
with open(file_path, 'r') as f:
content = f.read().strip()
if content.startswith('[') or content.startswith('{'):
# Regular JSON
loader = JSONLoader(
file_path=file_path,
jq_schema='.',
text_content=False
)
documents = loader.load()
doc_splits.extend(split_documents(documents))
else:
# JSONL - process line by line
documents = []
with open(file_path, 'r') as f:
for line in f:
if line.strip():
try:
json_obj = json.loads(line)
text = json.dumps(json_obj)
documents.append(text)
except json.JSONDecodeError:
continue
text_splitter = RecursiveCharacterTextSplitter(
chunk_size=1024,
chunk_overlap=64
)
doc_splits.extend(text_splitter.create_documents(documents))
except Exception as e:
print(f"Error processing {file_path}: {str(e)}")
continue
progress(0.5 + (i/total_files) * 0.5, desc=f"Processed {Path(file_path).name}")
return doc_splits
# Helper function to split documents
def split_documents(documents):
text_splitter = RecursiveCharacterTextSplitter(
chunk_size=1024,
chunk_overlap=64
)
return text_splitter.split_documents(documents)
# Create vector database
def create_db(splits, progress=gr.Progress()):
progress(0, desc="Creating vector database")
embeddings = HuggingFaceEmbeddings()
# Create vectors with progress bar
total_chunks = len(splits)
vectordb = FAISS.from_documents(
documents=splits,
embedding=embeddings
)
progress(1.0, desc="Vector database creation complete")
return vectordb
# Initialize langchain LLM chain
def initialize_llmchain(llm_model, temperature, max_tokens, top_k, vector_db, progress=gr.Progress()):
progress(0, desc=f"Initializing {llm_model}")
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,
)
progress(0.5, desc="Setting up memory and retriever")
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,
)
progress(1.0, desc="LLM chain initialized")
return qa_chain
# Initialize database
def initialize_database(list_file_obj, progress=gr.Progress()):
# Create a list of documents (when valid)
list_file_path = [x.name for x in list_file_obj if x is not None]
if not list_file_path:
return None, "No valid files uploaded. Please upload at least one file."
# Load document and create splits
doc_splits = load_doc(list_file_path, progress)
if not doc_splits:
return None, "Could not extract any text from the uploaded files."
# Create or load vector database
vector_db = create_db(doc_splits, progress)
# Count documents by type
file_types = {}
for path in list_file_path:
ext = Path(path).suffix.lower()
file_types[ext] = file_types.get(ext, 0) + 1
file_type_summary = ", ".join([f"{count} {ext}" for ext, count in file_types.items()])
return vector_db, f"Database created with {len(doc_splits)} chunks from {len(list_file_path)} files ({file_type_summary})!"
# Initialize LLM
def initialize_LLM(llm_option, llm_temperature, max_tokens, top_k, vector_db, progress=gr.Progress()):
if vector_db is None:
return None, "Please create a vector database first!"
llm_name = list_llm[llm_option]
print("llm_name: ", llm_name)
qa_chain = initialize_llmchain(llm_name, llm_temperature, max_tokens, top_k, vector_db, progress)
return qa_chain, f"QA chain initialized with {llm_name}. Chatbot is ready!"
def format_chat_history(message, chat_history):
formatted_chat_history = []
for user_message, bot_message in chat_history:
formatted_chat_history.append(f"User: {user_message}")
formatted_chat_history.append(f"Assistant: {bot_message}")
return formatted_chat_history
def conversation(qa_chain, message, history):
if qa_chain is None:
return None, gr.update(value=""), history, "Please initialize the chatbot first!", 0, "", 0, "", 0
formatted_chat_history = format_chat_history(message, history)
# Generate response using QA chain
response = 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]
response_sources = response["source_documents"]
# Handle source documents
source_contents = ["", "", ""]
source_pages = [0, 0, 0]
for i, source in enumerate(response_sources[:3]):
source_contents[i] = source.page_content.strip()
# Check if the metadata contains a page number
if "page" in source.metadata:
source_pages[i] = source.metadata["page"] + 1
elif "source" in source.metadata:
source_pages[i] = 1
source_contents[i] = f"From: {source.metadata['source']}\n{source_contents[i]}"
# Append user message and response to chat history
new_history = history + [(message, response_answer)]
return qa_chain, gr.update(value=""), new_history, source_contents[0], source_pages[0], source_contents[1], source_pages[1], source_contents[2], source_pages[2]
def get_file_icon(file_path):
"""Return an appropriate emoji icon based on file extension"""
ext = Path(file_path).suffix.lower()
icons = {
'.pdf': '📄',
'.txt': '📝',
'.md': '📋',
'.py': '🐍',
'.js': '⚙️',
'.json': '📊',
'.jsonl': '📊',
'.csv': '📈',
'.html': '🌐',
'.css': '🎨',
}
return icons.get(ext, '📁')
def display_file_list(file_obj):
if not file_obj:
return "No files uploaded yet"
file_list = [f"{get_file_icon(x.name)} {Path(x.name).name}" for x in file_obj if x is not None]
return "\n".join(file_list)
def demo():
with gr.Blocks(theme=gr.themes.Soft(primary_hue="indigo", secondary_hue="blue", neutral_hue="sky")) as demo:
vector_db = gr.State()
qa_chain = gr.State()
gr.HTML("<center><h1>📚 Enhanced RAG Chatbot</h1></center>")
gr.Markdown("""<b>Query your documents!</b> This enhanced AI agent performs retrieval augmented generation (RAG) on various document types
including PDFs, text files, markdown, code files, and structured data (CSV, JSON, JSONL). <b>Please do not upload confidential documents.</b>
""")
with gr.Row():
with gr.Column(scale=86):
gr.Markdown("<b>Step 1 - Upload Documents and Initialize RAG Pipeline</b>")
with gr.Row():
with gr.Column(scale=7):
document = gr.Files(
height=300,
file_count="multiple",
file_types=[".pdf", ".txt", ".md", ".py", ".js", ".json", ".jsonl", ".csv", ".html", ".css"],
interactive=True,
label="Upload Documents"
)
with gr.Column(scale=3):
file_list = gr.Textbox(
label="Uploaded Files",
value="No files uploaded yet",
interactive=False,
lines=12
)
document.upload(
display_file_list,
inputs=[document],
outputs=[file_list]
)
with gr.Row():
db_btn = gr.Button("Create Vector Database", variant="primary")
with gr.Row():
db_progress = gr.Textbox(
value="Not initialized",
show_label=False,
container=True
)
gr.Markdown("<b>Step 2 - Select LLM and Parameters</b>")
with gr.Row():
llm_btn = gr.Radio(
list_llm_simple,
label="Available LLMs",
value=list_llm_simple[0],
type="index"
)
with gr.Row():
with gr.Accordion("LLM Parameters", open=False):
with gr.Row():
slider_temperature = gr.Slider(
minimum=0.01,
maximum=1.0,
value=0.5,
step=0.1,
label="Temperature",
info="Controls randomness in generation",
interactive=True
)
with gr.Row():
slider_maxtokens = gr.Slider(
minimum=128,
maximum=9192,
value=4096,
step=128,
label="Max New Tokens",
info="Maximum tokens to generate",
interactive=True
)
with gr.Row():
slider_topk = gr.Slider(
minimum=1,
maximum=10,
value=3,
step=1,
label="Top-k",
info="Number of tokens to consider",
interactive=True
)
with gr.Row():
qachain_btn = gr.Button("Initialize Chatbot", variant="primary")
with gr.Row():
llm_progress = gr.Textbox(
value="Not initialized",
show_label=False,
container=True
)
with gr.Column(scale=200):
gr.Markdown("<b>Step 3 - Chat with Your Documents</b>")
chatbot = gr.Chatbot(height=505)
with gr.Accordion("Relevant Context from Documents", open=False):
with gr.Row():
doc_source1 = gr.Textbox(label="Reference 1", lines=2, container=True, scale=20)
source1_page = gr.Number(label="Page", scale=1)
with gr.Row():
doc_source2 = gr.Textbox(label="Reference 2", lines=2, container=True, scale=20)
source2_page = gr.Number(label="Page", scale=1)
with gr.Row():
doc_source3 = gr.Textbox(label="Reference 3", lines=2, container=True, scale=20)
source3_page = gr.Number(label="Page", scale=1)
with gr.Row():
msg = gr.Textbox(
placeholder="Ask a question about your documents...",
container=True,
lines=2
)
with gr.Row():
submit_btn = gr.Button("Submit", variant="primary")
clear_btn = gr.ClearButton([msg, chatbot], value="Clear")
# Preprocessing events
db_btn.click(
initialize_database,
inputs=[document],
outputs=[vector_db, db_progress]
)
qachain_btn.click(
initialize_LLM,
inputs=[llm_btn, slider_temperature, slider_maxtokens, slider_topk, vector_db],
outputs=[qa_chain, llm_progress]
).then(
lambda:[None,"",0,"",0,"",0],
inputs=None,
outputs=[chatbot, doc_source1, source1_page, doc_source2, source2_page, doc_source3, source3_page],
queue=False
)
# Chatbot events
msg.submit(
conversation,
inputs=[qa_chain, msg, chatbot],
outputs=[qa_chain, msg, chatbot, doc_source1, source1_page, doc_source2, source2_page, doc_source3, source3_page],
queue=False
)
submit_btn.click(
conversation,
inputs=[qa_chain, msg, chatbot],
outputs=[qa_chain, msg, chatbot, doc_source1, source1_page, doc_source2, source2_page, doc_source3, source3_page],
queue=False
)
clear_btn.click(
lambda:[None,"",0,"",0,"",0],
inputs=None,
outputs=[chatbot, doc_source1, source1_page, doc_source2, source2_page, doc_source3, source3_page],
queue=False
)
demo.queue().launch(debug=True)
if __name__ == "__main__":
demo()