6 |
7 | ## Overview
8 | AI-Tutor, an educational web app, integrates the latest OpenAI's Assistant API and Retrieval-Augmented Generation (RAG) to offer personalized tutoring across various academic courses. It adapts to specific course materials, ensuring a dynamic, responsive, and tailored learning experience.
9 |
10 | ## Core Features
11 | - **Versatile Application**: Adaptable to any course with the requirement of uploading course-specific materials.
12 | - **Intelligent Tutoring**: Tailored responses that align with course content, overcoming information hallucination.
13 | - **User Interface**: Streamlit-based, intuitive UI hosted on the Free Streamlit Community Cloud.
14 | - **Data Privacy and Security**: Adherence to OpenAI's stringent privacy practices.
15 |
16 | ## Technical Composition
17 | - **Core Technologies**: Utilizes LLM and RAG from OpenAI's Assistants API for core AI functionalities.
18 | - **Programming Language**: Primarily Python.
19 | - **Hosting Server**: Free Streamlit Community Cloud.
20 |
21 | ## Methodology
22 | AI-Tutor employs a combination of advanced AI technologies to enhance the learning experience:
23 |
24 | - **LLM API**: This forms the backbone of AI-Tutor's intelligence. It enables the platform to understand complex queries, retrieve information efficiently, and provide accurate, context-aware responses.
25 |
26 | - **Retrieval-Augmented Generation (RAG)**: RAG is a key component that supplements the AI's knowledge base. By accessing external data sources, RAG ensures that the AI's responses are not only relevant but also grounded in verified information. This feature is crucial for maintaining the accuracy and reliability of the educational content provided by AI-Tutor.
27 |
28 | - *Mitigating Information Hallucination*: RAG effectively addresses the challenge of AI-generated misinformation by validating responses against trusted external sources.
29 | - *Personalized Learning Experience*: The use of diverse data sources enables AI-Tutor to tailor its responses to specific educational contexts and individual learner needs.
30 |
31 | 
32 |
33 |
34 |
35 |
36 | ## Web App Components
37 |
38 |
39 |
40 |
41 | - **Sidebar**:
42 | - Option to delete all uploaded materials.
43 | - Generate and download Q&A records in HTML format.
44 | - **Main Page**:
45 | - API Key input.
46 | - Upload feature for course materials.
47 | - Interactive Q&A section.
48 | - Display and archive of Q&A records.
49 | - Feature to pose new questions.
50 |
51 | ## Getting Started
52 | 1. Access [AI-Tutor Streamlit App](https://aitutor-gawywv3h6qfwzzvikfzkpl.streamlit.app/).
53 | 2. Enter your OpenAI API Key.
54 | 3. Upload course materials for a custom tutoring session.
55 | 4. Interact with the AI assistant for course-specific queries.
56 | 5. Download the Q&A session transcript in HTML format.
57 |
58 | ## Benefits
59 | - **Enhanced Learning Experience**: Integrates RAG with AI technologies for a personalized educational journey.
60 | - **Data Privacy Assurance**: Strong commitment to user data protection.
61 | - **Current and Relevant Responses**: Ensures up-to-date and accurate information.
62 | - **Broad Course Coverage**: Applicable across various educational fields and disciplines.
63 | - **Interactive Tutoring**: Real-time feedback and dynamic learning sessions.
64 |
65 | ## References
66 | - [OpenAI Assistant API Documentation](https://platform.openai.com/docs/guides/assistants)
67 | - [Introduction to Retrieval-Augmented Generation (RAG)](https://www.datastax.com/blog/2020/10/introducing-retrieval-augmented-generation-rag)
68 | - [Vector Database Similarity Search](https://www.infoworld.com/article/3634357/what-is-vector-search-better-search-through-ai.html)
69 | - [OpenAI Privacy and Security Practices](https://openai.com/security)
70 |
--------------------------------------------------------------------------------
/app.py:
--------------------------------------------------------------------------------
1 | # Built-in modules
2 | import io
3 | import os
4 | import subprocess
5 |
6 | # Third-party modules
7 | import streamlit as st
8 | from PIL import Image
9 | from openai import OpenAI
10 |
11 | # Local modules
12 | import api_handler
13 | from api_handler import send_query_get_response
14 | from chat_gen import generate_html
15 | from file_upload import upload_files_to_assistant, attach_files_to_assistant, check_and_upload_files
16 |
17 | logo=Image.open('logo.png')
18 | sb_logo=Image.open('sb_logo.png')
19 |
20 |
21 | c1, c2 = st.columns([0.9, 3.2])
22 |
23 | with c1:
24 | st.caption('')
25 | st.caption('')
26 | st.image(logo,width=120)
27 |
28 | with c2:
29 |
30 | st.title('EduMentor : An AI-Enhanced Tutoring System')
31 |
32 |
33 | # RAG Function Description
34 | st.markdown("## AI Tutor Description")
35 | rag_description = """
36 | EduMentor leverages the cutting-edge RAG (Retrieval-Augmented Generation) function to provide in-depth, contextually rich answers to complex educational queries. This AI-driven approach combines extensive knowledge retrieval with dynamic response generation, offering students a deeper, more nuanced understanding of subjects and fostering a more interactive, exploratory learning environment.
37 | """
38 | st.markdown(rag_description)
39 |
40 | # OpenAI API Key Input
41 | api_key = st.text_input(label='Enter your OpenAI API Key', type='password')
42 |
43 | if api_key:
44 | # If API key is entered, initialize the OpenAI client and proceed with app functionality
45 | client = OpenAI(api_key=api_key)
46 | assistant_id = 'asst_konaahsahZ0UhK82guGBvn6m'
47 |
48 | # File Handling Section
49 | files_info = check_and_upload_files(client, assistant_id)
50 |
51 | st.markdown(f'Number of files uploaded in the assistant: :blue[{len(files_info)}]')
52 | st.divider()
53 |
54 | # Sidebar for Additional Features
55 | st.sidebar.header('EduMentor: AI-Tutor')
56 | st.sidebar.image(logo,width=120)
57 | st.sidebar.caption('Made by D')
58 | # Adding a button in the sidebar to delete all files from the assistant
59 | if st.sidebar.button('Delete All Files from Assistant'):
60 | # Retrieve all file IDs associated with the assistant
61 | assistant_files_response = client.beta.assistants.files.list(assistant_id=assistant_id)
62 | assistant_files = assistant_files_response.data
63 |
64 | # Delete each file
65 | for file in assistant_files:
66 | file_id = file.id
67 | client.beta.assistants.files.delete(assistant_id=assistant_id, file_id=file_id)
68 | st.sidebar.success(f'Deleted file: {file_id}')
69 |
70 | if st.sidebar.button('Generate Chat History'):
71 | html_data = generate_html(st.session_state.messages)
72 | st.sidebar.download_button(label="Download Chat History as HTML",
73 | data=html_data,
74 | file_name="chat_history.html",
75 | mime="text/html")
76 |
77 |
78 | # Main Chat Interface
79 | st.subheader('Q&A record with AI-Tutor 📜')
80 | st.caption('You can choose to download the chat history in either PDF or HTML format using the options in the sidebar on the left.')
81 | if "messages" not in st.session_state:
82 | st.session_state.messages = []
83 |
84 | for message in st.session_state.messages:
85 | with st.chat_message(message["role"]):
86 | st.markdown(message["content"], unsafe_allow_html=True)
87 |
88 | if prompt := st.chat_input("Welcome and ask a question to the AI tutor"):
89 | st.session_state.messages.append({"role": "user", "content": prompt})
90 | with st.chat_message("user"):
91 | st.markdown(prompt)
92 |
93 | with st.chat_message("assistant", avatar='👨🏻🏫'):
94 | message_placeholder = st.empty()
95 | with st.spinner('Thinking...'):
96 | response = send_query_get_response(client,prompt,assistant_id)
97 | message_placeholder.markdown(response)
98 | st.session_state.messages.append({"role": "assistant", "content": response})
99 |
100 | else:
101 | # Prompt for API key if not entered
102 | st.warning("Please enter your OpenAI API Key to use EduMentor.")
103 |
--------------------------------------------------------------------------------
5 | 
39 |