Update app.py
Browse files
app.py
CHANGED
|
@@ -1,203 +1,103 @@
|
|
| 1 |
-
import os
|
| 2 |
-
import json
|
| 3 |
-
import re
|
| 4 |
import gradio as gr
|
|
|
|
|
|
|
| 5 |
import requests
|
| 6 |
-
import
|
| 7 |
-
import urllib.parse
|
| 8 |
-
from tempfile import NamedTemporaryFile
|
| 9 |
-
from bs4 import BeautifulSoup
|
| 10 |
from typing import List
|
| 11 |
from pydantic import BaseModel, Field
|
| 12 |
-
|
| 13 |
-
from duckduckgo_search import DDGS
|
| 14 |
-
from langchain_community.vectorstores import FAISS
|
| 15 |
-
from langchain_community.document_loaders import PyPDFLoader
|
| 16 |
-
from langchain_community.embeddings import HuggingFaceEmbeddings
|
| 17 |
-
from langchain_community.llms import HuggingFaceHub
|
| 18 |
-
from langchain_core.documents import Document
|
| 19 |
-
from sentence_transformers import SentenceTransformer
|
| 20 |
-
from llama_parse import LlamaParse
|
| 21 |
-
import logging
|
| 22 |
-
|
| 23 |
-
# Set up logging
|
| 24 |
-
logging.basicConfig(level=logging.DEBUG, format='%(asctime)s - %(levelname)s - %(message)s')
|
| 25 |
|
| 26 |
# Environment variables and configurations
|
| 27 |
huggingface_token = os.environ.get("HUGGINGFACE_TOKEN")
|
| 28 |
-
llama_cloud_api_key = os.environ.get("LLAMA_CLOUD_API_KEY")
|
| 29 |
-
|
| 30 |
-
# Initialize SentenceTransformer and LlamaParse
|
| 31 |
-
sentence_model = SentenceTransformer('paraphrase-MiniLM-L6-v2')
|
| 32 |
-
llama_parser = LlamaParse(
|
| 33 |
-
api_key=llama_cloud_api_key,
|
| 34 |
-
result_type="markdown",
|
| 35 |
-
num_workers=4,
|
| 36 |
-
verbose=True,
|
| 37 |
-
language="en",
|
| 38 |
-
)
|
| 39 |
-
|
| 40 |
-
def load_document(file: NamedTemporaryFile, parser: str = "pypdf") -> List[Document]:
|
| 41 |
-
if parser == "pypdf":
|
| 42 |
-
loader = PyPDFLoader(file.name)
|
| 43 |
-
return loader.load_and_split()
|
| 44 |
-
elif parser == "llamaparse":
|
| 45 |
-
try:
|
| 46 |
-
documents = llama_parser.load_data(file.name)
|
| 47 |
-
return [Document(page_content=doc.text, metadata={"source": file.name}) for doc in documents]
|
| 48 |
-
except Exception as e:
|
| 49 |
-
print(f"Error using Llama Parse: {str(e)}")
|
| 50 |
-
print("Falling back to PyPDF parser")
|
| 51 |
-
loader = PyPDFLoader(file.name)
|
| 52 |
-
return loader.load_and_split()
|
| 53 |
-
else:
|
| 54 |
-
raise ValueError("Invalid parser specified. Use 'pypdf' or 'llamaparse'.")
|
| 55 |
-
|
| 56 |
-
def update_vectors(files, parser):
|
| 57 |
-
if not files:
|
| 58 |
-
return "Please upload at least one PDF file."
|
| 59 |
-
|
| 60 |
-
embed = HuggingFaceEmbeddings(model_name="sentence-transformers/all-mpnet-base-v2")
|
| 61 |
-
total_chunks = 0
|
| 62 |
-
|
| 63 |
-
all_data = []
|
| 64 |
-
for file in files:
|
| 65 |
-
data = load_document(file, parser)
|
| 66 |
-
all_data.extend(data)
|
| 67 |
-
total_chunks += len(data)
|
| 68 |
-
|
| 69 |
-
if os.path.exists("faiss_database"):
|
| 70 |
-
database = FAISS.load_local("faiss_database", embed, allow_dangerous_deserialization=True)
|
| 71 |
-
database.add_documents(all_data)
|
| 72 |
-
else:
|
| 73 |
-
database = FAISS.from_documents(all_data, embed)
|
| 74 |
-
|
| 75 |
-
database.save_local("faiss_database")
|
| 76 |
-
|
| 77 |
-
return f"Vector store updated successfully. Processed {total_chunks} chunks from {len(files)} files using {parser}."
|
| 78 |
-
|
| 79 |
-
def clear_cache():
|
| 80 |
-
if os.path.exists("faiss_database"):
|
| 81 |
-
os.remove("faiss_database")
|
| 82 |
-
return "Cache cleared successfully."
|
| 83 |
-
else:
|
| 84 |
-
return "No cache to clear."
|
| 85 |
-
|
| 86 |
-
def get_model(temperature, top_p, repetition_penalty):
|
| 87 |
-
return HuggingFaceHub(
|
| 88 |
-
repo_id="mistralai/Mistral-7B-Instruct-v0.3",
|
| 89 |
-
model_kwargs={
|
| 90 |
-
"temperature": temperature,
|
| 91 |
-
"top_p": top_p,
|
| 92 |
-
"repetition_penalty": repetition_penalty,
|
| 93 |
-
"max_length": 1000
|
| 94 |
-
},
|
| 95 |
-
huggingfacehub_api_token=huggingface_token
|
| 96 |
-
)
|
| 97 |
|
|
|
|
| 98 |
def duckduckgo_search(query):
|
| 99 |
-
logging.debug(f"Performing DuckDuckGo search for query: {query}")
|
| 100 |
with DDGS() as ddgs:
|
| 101 |
-
results =
|
| 102 |
-
logging.debug(f"Search returned {len(results)} results")
|
| 103 |
return results
|
| 104 |
|
| 105 |
-
|
| 106 |
-
|
| 107 |
-
|
| 108 |
-
|
| 109 |
-
|
| 110 |
-
embed = HuggingFaceEmbeddings(model_name="sentence-transformers/all-mpnet-base-v2")
|
| 111 |
|
| 112 |
-
|
| 113 |
-
|
| 114 |
-
|
| 115 |
-
retriever = database.as_retriever()
|
| 116 |
-
relevant_docs = retriever.get_relevant_documents(query)
|
| 117 |
-
context = "\n".join([f"Content: {doc.page_content}\nSource: {doc.metadata['source']}\n" for doc in relevant_docs])
|
| 118 |
-
else:
|
| 119 |
-
logging.debug("Using web search for context")
|
| 120 |
-
search_results = duckduckgo_search(query)
|
| 121 |
-
context = "\n".join(f"{result['title']}\n{result['body']}\nSource: {result['href']}\n"
|
| 122 |
-
for result in search_results if 'body' in result)
|
| 123 |
|
| 124 |
-
|
| 125 |
-
|
|
|
|
|
|
|
|
|
|
| 126 |
prompt = f"""<s>[INST] Using the following context:
|
| 127 |
{context}
|
| 128 |
Write a detailed and complete research document that fulfills the following user request: '{query}'
|
| 129 |
-
After the
|
| 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 |
-
|
| 155 |
-
|
| 156 |
-
|
| 157 |
-
|
| 158 |
-
|
| 159 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 160 |
|
| 161 |
-
def chatbot_interface(message, history
|
| 162 |
-
|
| 163 |
-
main_content, sources = get_response_with_search(message, temperature, top_p, repetition_penalty, use_pdf)
|
| 164 |
formatted_response = f"{main_content}\n\nSources:\n{sources}"
|
| 165 |
-
logging.debug(f"Formatted response: {formatted_response[:100]}...") # Log first 100 characters
|
| 166 |
return formatted_response
|
| 167 |
|
| 168 |
-
# Gradio interface
|
| 169 |
-
|
| 170 |
-
|
| 171 |
-
|
| 172 |
-
|
| 173 |
-
|
| 174 |
-
|
| 175 |
-
|
| 176 |
-
|
| 177 |
-
|
| 178 |
-
|
| 179 |
-
|
| 180 |
-
|
| 181 |
-
|
| 182 |
-
|
| 183 |
-
msg = gr.Textbox(label="Ask a question")
|
| 184 |
-
submit_button = gr.Button("Submit")
|
| 185 |
-
with gr.Column(scale=1):
|
| 186 |
-
temperature = gr.Slider(label="Temperature", minimum=0.0, maximum=1.0, value=0.7, step=0.1)
|
| 187 |
-
top_p = gr.Slider(label="Top P", minimum=0.0, maximum=1.0, value=0.95, step=0.05)
|
| 188 |
-
repetition_penalty = gr.Slider(label="Repetition Penalty", minimum=1.0, maximum=2.0, value=1.1, step=0.1)
|
| 189 |
-
use_pdf = gr.Checkbox(label="Use PDF Documents", value=False)
|
| 190 |
-
|
| 191 |
-
def respond(message, chat_history, temperature, top_p, repetition_penalty, use_pdf):
|
| 192 |
-
bot_message = chatbot_interface(message, chat_history, temperature, top_p, repetition_penalty, use_pdf)
|
| 193 |
-
chat_history.append((message, bot_message))
|
| 194 |
-
return "", chat_history
|
| 195 |
-
|
| 196 |
-
submit_button.click(respond, inputs=[msg, chatbot, temperature, top_p, repetition_penalty, use_pdf], outputs=[msg, chatbot])
|
| 197 |
-
|
| 198 |
-
clear_button = gr.Button("Clear Cache")
|
| 199 |
-
clear_output = gr.Textbox(label="Cache Status")
|
| 200 |
-
clear_button.click(clear_cache, inputs=[], outputs=clear_output)
|
| 201 |
|
| 202 |
if __name__ == "__main__":
|
| 203 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
import gradio as gr
|
| 2 |
+
from huggingface_hub import InferenceApi
|
| 3 |
+
from duckduckgo_search import DDGS
|
| 4 |
import requests
|
| 5 |
+
import json
|
|
|
|
|
|
|
|
|
|
| 6 |
from typing import List
|
| 7 |
from pydantic import BaseModel, Field
|
| 8 |
+
import os
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 9 |
|
| 10 |
# Environment variables and configurations
|
| 11 |
huggingface_token = os.environ.get("HUGGINGFACE_TOKEN")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 12 |
|
| 13 |
+
# Function to perform a DuckDuckGo search
|
| 14 |
def duckduckgo_search(query):
|
|
|
|
| 15 |
with DDGS() as ddgs:
|
| 16 |
+
results = ddgs.text(query, max_results=5)
|
|
|
|
| 17 |
return results
|
| 18 |
|
| 19 |
+
class CitingSources(BaseModel):
|
| 20 |
+
sources: List[str] = Field(
|
| 21 |
+
...,
|
| 22 |
+
description="List of sources to cite. Should be an URL of the source."
|
| 23 |
+
)
|
|
|
|
| 24 |
|
| 25 |
+
def get_response_with_search(query):
|
| 26 |
+
# Perform the web search
|
| 27 |
+
search_results = duckduckgo_search(query)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 28 |
|
| 29 |
+
# Use the search results as context for the model
|
| 30 |
+
context = "\n".join(f"{result['title']}\n{result['body']}\nSource: {result['href']}\n"
|
| 31 |
+
for result in search_results if 'body' in result)
|
| 32 |
+
|
| 33 |
+
# Prompt formatted for Mistral-7B-Instruct
|
| 34 |
prompt = f"""<s>[INST] Using the following context:
|
| 35 |
{context}
|
| 36 |
Write a detailed and complete research document that fulfills the following user request: '{query}'
|
| 37 |
+
After writing the document, please provide a list of sources used in your response. [/INST]"""
|
| 38 |
+
|
| 39 |
+
# API endpoint for Mistral-7B-Instruct-v0.3
|
| 40 |
+
API_URL = "https://api-inference.huggingface.co/models/mistralai/Mistral-7B-Instruct-v0.3"
|
| 41 |
+
|
| 42 |
+
# Headers
|
| 43 |
+
headers = {"Authorization": f"Bearer {huggingface_token}"}
|
| 44 |
+
|
| 45 |
+
# Payload
|
| 46 |
+
payload = {
|
| 47 |
+
"inputs": prompt,
|
| 48 |
+
"parameters": {
|
| 49 |
+
"max_new_tokens": 1000,
|
| 50 |
+
"temperature": 0.7,
|
| 51 |
+
"top_p": 0.95,
|
| 52 |
+
"top_k": 40,
|
| 53 |
+
"repetition_penalty": 1.1
|
| 54 |
+
}
|
| 55 |
+
}
|
| 56 |
+
|
| 57 |
+
# Make the API call
|
| 58 |
+
response = requests.post(API_URL, headers=headers, json=payload)
|
| 59 |
+
|
| 60 |
+
if response.status_code == 200:
|
| 61 |
+
result = response.json()
|
| 62 |
+
if isinstance(result, list) and len(result) > 0:
|
| 63 |
+
generated_text = result[0].get('generated_text', 'No text generated')
|
| 64 |
+
|
| 65 |
+
# Remove the instruction part
|
| 66 |
+
content_start = generated_text.find("[/INST]")
|
| 67 |
+
if content_start != -1:
|
| 68 |
+
generated_text = generated_text[content_start + 7:].strip()
|
| 69 |
+
|
| 70 |
+
# Split the response into main content and sources
|
| 71 |
+
parts = generated_text.split("Sources:", 1)
|
| 72 |
+
main_content = parts[0].strip()
|
| 73 |
+
sources = parts[1].strip() if len(parts) > 1 else ""
|
| 74 |
+
|
| 75 |
+
return main_content, sources
|
| 76 |
+
else:
|
| 77 |
+
return f"Unexpected response format: {result}", ""
|
| 78 |
+
else:
|
| 79 |
+
return f"Error: API returned status code {response.status_code}", ""
|
| 80 |
|
| 81 |
+
def chatbot_interface(message, history):
|
| 82 |
+
main_content, sources = get_response_with_search(message)
|
|
|
|
| 83 |
formatted_response = f"{main_content}\n\nSources:\n{sources}"
|
|
|
|
| 84 |
return formatted_response
|
| 85 |
|
| 86 |
+
# Gradio chatbot interface
|
| 87 |
+
iface = gr.ChatInterface(
|
| 88 |
+
fn=chatbot_interface,
|
| 89 |
+
title="AI-powered Web Search Assistant",
|
| 90 |
+
description="Ask questions, and I'll search the web and provide answers using the Mistral-7B-Instruct model.",
|
| 91 |
+
examples=[
|
| 92 |
+
["What are the latest developments in AI?"],
|
| 93 |
+
["Tell me about recent updates on GitHub"],
|
| 94 |
+
["What are the best hotels in Galapagos, Ecuador?"],
|
| 95 |
+
["Summarize recent advancements in Python programming"],
|
| 96 |
+
],
|
| 97 |
+
retry_btn="Retry",
|
| 98 |
+
undo_btn="Undo",
|
| 99 |
+
clear_btn="Clear",
|
| 100 |
+
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 101 |
|
| 102 |
if __name__ == "__main__":
|
| 103 |
+
iface.launch(share=True)
|