Rahatara commited on
Commit
3c3edd2
1 Parent(s): ab84a9d

Create app.py

Browse files
Files changed (1) hide show
  1. app.py +124 -0
app.py ADDED
@@ -0,0 +1,124 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import os
2
+ import gradio as gr
3
+ import fitz # PyMuPDF
4
+ from sentence_transformers import SentenceTransformer
5
+ import numpy as np
6
+ import faiss
7
+ from typing import List, Tuple, Dict
8
+ from google.generativeai import GenerativeModel, configure, types
9
+
10
+ # Set up the Google API for the Gemini model
11
+ GOOGLE_API_KEY = os.environ.get("GOOGLE_API_KEY")
12
+ configure(api_key=GOOGLE_API_KEY)
13
+
14
+ # Placeholder for the app's state
15
+ class MyApp:
16
+ def __init__(self) -> None:
17
+ self.documents = []
18
+ self.embeddings = None
19
+ self.index = None
20
+ self.model = SentenceTransformer('all-MiniLM-L6-v2')
21
+
22
+ def load_pdfs(self, files: List[gr.File]) -> str:
23
+ """Extracts text from multiple PDF files and stores them."""
24
+ self.documents = []
25
+ for file in files:
26
+ doc = fitz.open(stream=file.read(), filetype="pdf")
27
+ for page_num in range(len(doc)):
28
+ page = doc[page_num]
29
+ text = page.get_text()
30
+ self.documents.append({
31
+ "file_name": file.name,
32
+ "page": page_num + 1,
33
+ "content": text
34
+ })
35
+ return f"Processed {len(files)} PDFs successfully!"
36
+
37
+ def build_vector_db(self) -> str:
38
+ """Builds a vector database using the content of the PDFs."""
39
+ if not self.documents:
40
+ return "No documents to process."
41
+ contents = [doc["content"] for doc in self.documents]
42
+ self.embeddings = self.model.encode(contents, show_progress_bar=True)
43
+ self.index = faiss.IndexFlatL2(self.embeddings.shape[1])
44
+ self.index.add(np.array(self.embeddings))
45
+ return "Vector database built successfully!"
46
+
47
+ def search_documents(self, query: str, k: int = 3) -> List[Dict]:
48
+ """Searches for relevant document snippets using vector similarity."""
49
+ if not self.index:
50
+ return [{"content": "Vector database is not built."}]
51
+ query_embedding = self.model.encode([query], show_progress_bar=False)
52
+ D, I = self.index.search(np.array(query_embedding), k)
53
+ results = [self.documents[i] for i in I[0]]
54
+ return results if results else [{"content": "No relevant documents found."}]
55
+
56
+ app = MyApp()
57
+
58
+ def upload_files(files: List[gr.File]) -> str:
59
+ return app.load_pdfs(files)
60
+
61
+ def build_vector_db() -> str:
62
+ return app.build_vector_db()
63
+
64
+ def respond(message: str, history: List[Tuple[str, str]]) -> Tuple[List[Tuple[str, str]], str]:
65
+ system_message = (
66
+ "You are a helpful assistant designed to assist with studying and learning. "
67
+ "You analyze uploaded PDF documents and provide clear, concise responses "
68
+ "to any questions based on the content. You strive to be accurate, detailed, "
69
+ "and educational in your responses."
70
+ )
71
+ messages = [{"role": "system", "content": system_message}]
72
+
73
+ for user_msg, assistant_msg in history:
74
+ if user_msg:
75
+ messages.append({"role": "user", "content": user_msg})
76
+ if assistant_msg:
77
+ messages.append({"role": "assistant", "content": assistant_msg})
78
+
79
+ messages.append({"role": "user", "content": message})
80
+
81
+ # Retrieve relevant documents
82
+ retrieved_docs = app.search_documents(message)
83
+ context = "\n".join(
84
+ [f"File: {doc['file_name']}, Page: {doc['page']}\n{doc['content'][:200]}..." for doc in retrieved_docs]
85
+ )
86
+
87
+ # Generate response using the generative model
88
+ model = GenerativeModel("gemini-1.5-pro-latest")
89
+ generation_config = types.GenerationConfig(
90
+ temperature=0.7,
91
+ max_output_tokens=1024,
92
+ )
93
+
94
+ try:
95
+ response = model.generate_content([context + "\nQuestion: " + message], generation_config=generation_config)
96
+ response_content = response.text if hasattr(response, "text") else "No response generated."
97
+ except Exception as e:
98
+ response_content = f"An error occurred while generating the response: {str(e)}"
99
+
100
+ # Append the message and generated response to the chat history
101
+ history.append((message, response_content))
102
+ return history, ""
103
+
104
+ with gr.Blocks() as demo:
105
+ gr.Markdown("# Study Assistant Chatbot")
106
+ gr.Markdown("Upload your PDFs, build a vector database, and ask questions to learn efficiently.")
107
+
108
+ with gr.Row():
109
+ with gr.Column():
110
+ upload_btn = gr.File(label="Upload PDFs", file_types=[".pdf"], file_count="multiple")
111
+ upload_message = gr.Textbox(label="Upload Status", lines=2)
112
+ build_db_btn = gr.Button("Build Vector Database")
113
+ db_message = gr.Textbox(label="DB Build Status", lines=2)
114
+
115
+ upload_btn.change(upload_files, inputs=[upload_btn], outputs=[upload_message])
116
+ build_db_btn.click(build_vector_db, inputs=[], outputs=[db_message])
117
+
118
+ with gr.Column():
119
+ chatbot = gr.Chatbot(label="Chat Responses")
120
+ query_input = gr.Textbox(label="Enter your query here")
121
+ submit_btn = gr.Button("Submit")
122
+ submit_btn.click(respond, inputs=[query_input, chatbot], outputs=[chatbot, query_input])
123
+
124
+ demo.launch()