on1onmangoes's picture
Update app.py
4204dae verified
raw
history blame
9.1 kB
import gradio as gr
from gradio_client import Client, handle_file
import os
# Define your Hugging Face token (make sure to set it as an environment variable)
HF_TOKEN = os.getenv("HF_TOKEN") # Replace with your actual token if not using an environment variable
# Initialize the Gradio Client for the specified API
client = Client("on1onmangoes/CNIHUB101324v10", hf_token=HF_TOKEN)
# Update the conversation history within the function.
# Return the updated history along with any other required outputs.
client_name = ['rosariarossi','bianchifiordaliso','lorenzoverdi','lucia', 'quarto4', 'quinto5', 'secondo6', 'sesto6', 'settimo7','ottavo8','nono9']
def stream_chat_with_rag(
message: str,
history: list,
client_name: str,
system_prompt: str,
num_retrieved_docs: int = 10,
num_docs_final: int = 9,
temperature: float = 0,
max_new_tokens: int = 1024,
top_p: float = 1.0,
top_k: int = 20,
penalty: float = 1.2,
):
print(f"Message: {message}")
print(f"History: {history}")
# Build the conversation prompt including system prompt and history
conversation = f"{system_prompt}\n\nFor Client: {client_name}\n"
# Add previous conversation history
for user_input, assistant_response in history:
conversation += f"User: {user_input}\nAssistant: {assistant_response}\n"
# Add the current user message
conversation += f"User: {message}\nAssistant:"
# Call the API with the user's message
question = message
answer = client.predict(question=question, api_name="/answer_with_rag")
# Debugging: Print the raw response
print("Raw answer from API:")
print(answer)
# Format the assistant's answer and the relevant documents separately
formatted_answer = format_answer_string(answer)
# Update the conversation history with the new message and answer
history.append((message, formatted_answer))
# Return the formatted answer
return formatted_answer
def format_answer_string(answer: str):
"""
This function extracts and formats the assistant's response before document metadata.
Anything after the marker `[(` (where documents are listed) is ignored.
"""
# Step 1: Split the response at the start of the document metadata
split_marker = "[("
if split_marker in answer:
# Everything before the marker is the relevant answer
answer_before_docs = answer.split(split_marker)[0]
else:
# If no documents metadata, return the entire answer
answer_before_docs = answer
# Step 2: Clean up formatting by replacing escaped newline characters
formatted_answer = answer_before_docs.replace("\\n", "\n").strip()
# Step 3: Remove potential starting and ending artifacts like (' and ,) if present
if formatted_answer.startswith("(\"") and formatted_answer.endswith("\","):
formatted_answer = formatted_answer[2:-2].strip()
# Optional: Add a prefix for clarity
formatted_answer = "Co-Pilot: " + formatted_answer
return formatted_answer
def format_relevant_documents(relevant_docs: list):
"""
This function formats the relevant document metadata and content for readable output.
It extracts the heading, page number, and a snippet of the content from each document.
"""
formatted_docs = "Relevant Documents:\n\n"
for idx, (doc, score) in enumerate(relevant_docs):
# Extract the relevant metadata
heading = doc.metadata.get('heading', 'Unnamed Document')
page_number = int(doc.metadata.get('page_number', -1))
source = doc.metadata.get('source', 'Unknown Source')
confidence = round(score, 4) # Rounding the score for cleaner output
# Add the formatted details to the output string
formatted_docs += f"Document {idx + 1}:\n"
formatted_docs += f" - Heading: {heading}\n"
formatted_docs += f" - Page Number: {page_number}\n"
formatted_docs += f" - Source: {source}\n"
formatted_docs += f" - Confidence Score: {confidence}\n"
# Optionally include a snippet from the content
content_snippet = doc.page_content[:200] # Take the first 200 characters for preview
formatted_docs += f" - Content Snippet: {content_snippet}...\n\n"
return formatted_docs.strip()
# Function to handle PDF processing API call
def process_pdf(pdf_file, client_name):
return client.predict(
pdf_file=handle_file(pdf_file),
client_name=client_name, # Hardcoded client name
api_name="/process_pdf2"
)[1] # Return only the result string
# Function to handle search API call
def search_api(query):
return client.predict(query=query, api_name="/search_with_confidence")
# Function to handle RAG API call
def rag_api(question):
return client.predict(question=question, api_name="/answer_with_rag")
# CSS for custom styling
CSS = """
# chat-container {
height: 100vh;
}
"""
# Title for the application
TITLE = "<h1 style='text-align:center;'>CNI RAG QA v0</h1>"
# Create the Gradio Blocks interface
with gr.Blocks(css=CSS) as demo:
gr.HTML(TITLE)
with gr.Tab("Chat"):
chatbot = gr.Chatbot() # Create a chatbot interface
chat_interface = gr.ChatInterface(
fn=stream_chat_with_rag,
chatbot=chatbot,
additional_inputs_accordion=gr.Accordion(label="⚙️ Parameters", open=False, render=False),
additional_inputs=[
gr.Dropdown(client_name,value="rosariarossi",label="Select Client", render=False,allow_custom_value=True),
gr.Textbox(
value="You are an expert assistant",
label="System Prompt",
render=False,
),
gr.Slider(
minimum=1,
maximum=10,
step=1,
value=10,
label="Number of Initial Documents to Retrieve",
render=False,
),
gr.Slider(
minimum=1,
maximum=10,
step=1,
value=9,
label="Number of Final Documents to Retrieve",
render=False,
),
gr.Slider(
minimum=0.2,
maximum=1,
step=0.1,
value=0,
label="Temperature",
render=False,
),
gr.Slider(
minimum=128,
maximum=8192,
step=1,
value=1024,
label="Max new tokens",
render=False,
),
gr.Slider(
minimum=0.0,
maximum=1.0,
step=0.1,
value=1.0,
label="Top P",
render=False,
),
gr.Slider(
minimum=1,
maximum=20,
step=1,
value=20,
label="Top K",
render=False,
),
gr.Slider(
minimum=0.0,
maximum=2.0,
step=0.1,
value=1.2,
label="Repetition Penalty",
render=False,
),
],
)
# with gr.Tab("Process PDF"):
# pdf_input = gr.File(label="Upload PDF File")
# pdf_output = gr.Textbox(label="PDF Result", interactive=False)
# pdf_button = gr.Button("Process PDF")
# pdf_button.click(
# process_pdf,
# inputs=[pdf_input],
# outputs=pdf_output
# )
with gr.Tab("Process PDF"):
pdf_input = gr.File(label="Upload PDF File")
select_client_dropdown = gr.Dropdown(client_name, value="rosariarossi", label="Select or Type Client", allow_custom_value=True)
pdf_output = gr.Textbox(label="PDF Result", interactive=False)
pdf_button = gr.Button("Process PDF")
pdf_button.click(
process_pdf,
inputs=[pdf_input, select_client_dropdown], # Pass both PDF and client name as inputs
outputs=pdf_output
)
with gr.Tab("Answer with RAG"):
question_input = gr.Textbox(label="Enter Question for RAG")
answer_with_rag_select_client_dropdown = gr.Dropdown(client_name, value="rosariarossi", label="Select or Type Client", allow_custom_value=True)
rag_output = gr.Textbox(label="RAG Answer Result", interactive=False)
rag_button = gr.Button("Get Answer")
rag_button.click(
rag_api,
inputs=[question_input,answer_with_rag_select_client_dropdown ],
outputs=rag_output
)
# Launch the app
if __name__ == "__main__":
demo.launch()