shukdevdatta123 commited on
Commit
818bad6
·
verified ·
1 Parent(s): 505d85c

Create app.py

Browse files
Files changed (1) hide show
  1. app.py +86 -0
app.py ADDED
@@ -0,0 +1,86 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import streamlit as st
2
+ import PyPDF2
3
+ import openai
4
+ import faiss
5
+ import os
6
+ from sklearn.feature_extraction.text import TfidfVectorizer
7
+ from sklearn.metrics.pairwise import cosine_similarity
8
+
9
+ # Function to extract text from a PDF file
10
+ def extract_text_from_pdf(pdf_file):
11
+ reader = PyPDF2.PdfReader(pdf_file)
12
+ text = ""
13
+ for page in reader.pages:
14
+ text += page.extract_text()
15
+ return text
16
+
17
+ # Function to generate embeddings for a piece of text
18
+ def get_embeddings(text, model="text-embedding-ada-002"):
19
+ response = openai.Embedding.create(input=[text], model=model)
20
+ return response['data'][0]['embedding']
21
+
22
+ # Function to search for similar content
23
+ def search_similar(query_embedding, index, stored_texts, top_k=3):
24
+ distances, indices = index.search([query_embedding], top_k)
25
+ results = [(stored_texts[i], distances[0][idx]) for idx, i in enumerate(indices[0])]
26
+ return results
27
+
28
+ # Streamlit app starts here
29
+ st.title("Course Query Assistant")
30
+
31
+ # Input OpenAI API key
32
+ openai_api_key = st.text_input("Enter your OpenAI API key:", type="password")
33
+
34
+ if openai_api_key:
35
+ openai.api_key = openai_api_key
36
+
37
+ # Upload course materials
38
+ uploaded_files = st.file_uploader("Upload Course Materials (PDFs)", type=["pdf"], accept_multiple_files=True)
39
+
40
+ if uploaded_files:
41
+ st.write("Processing uploaded course materials...")
42
+
43
+ # Extract text and generate embeddings for all uploaded PDFs
44
+ course_texts = []
45
+ for uploaded_file in uploaded_files:
46
+ text = extract_text_from_pdf(uploaded_file)
47
+ course_texts.append(text)
48
+
49
+ # Combine all course materials into one large text
50
+ combined_text = " ".join(course_texts)
51
+
52
+ # Split combined text into smaller chunks for embedding (max tokens ~1000)
53
+ chunks = [combined_text[i:i+1000] for i in range(0, len(combined_text), 1000)]
54
+
55
+ # Generate embeddings for all chunks
56
+ embeddings = [get_embeddings(chunk) for chunk in chunks]
57
+
58
+ # Create a FAISS index for similarity search
59
+ index = faiss.IndexFlatL2(len(embeddings[0]))
60
+ index.add(embeddings)
61
+
62
+ st.write("Course materials have been processed and indexed.")
63
+
64
+ # User query
65
+ query = st.text_input("Enter your question about the course materials:")
66
+
67
+ if query:
68
+ # Generate embedding for the query
69
+ query_embedding = get_embeddings(query)
70
+
71
+ # Search for similar chunks in the FAISS index
72
+ results = search_similar(query_embedding, index, chunks)
73
+
74
+ # Create the context for the GPT prompt
75
+ context = "\n".join([result[0] for result in results])
76
+ modified_prompt = f"Context: {context}\n\nQuestion: {query}\n\nProvide a detailed answer based on the context."
77
+
78
+ # Get the GPT-3.5-turbo response
79
+ response = openai.ChatCompletion.create(
80
+ model="gpt-3.5-turbo",
81
+ messages=[{"role": "user", "content": modified_prompt}]
82
+ )
83
+
84
+ # Display the response
85
+ st.write("### Intelligent Reply:")
86
+ st.write(response['choices'][0]['message']['content'])