import streamlit as st from openai import OpenAI from typing import List, Callable import os api_key = os.environ.get("API_TOKEN") client = OpenAI(api_key = api_key) def generate_checklist(modules: List[str], quiz_callback: Callable[[str], None]) -> None: st.subheader("Learning Modules") for module in modules: checkbox = st.checkbox(module, key=module) if checkbox: if st.button(f"Generate Quiz Question!", key=f"quiz_{module}"): quiz_callback(module) class GPTService: def __init__(self): self.system_prompt = """You are a learning assistant. Users will ask you how to begin learning new concepts, such as quantum physics, archeology or architecture. Your job is to curate a Coursera-style learning guide for users where you create different learning modules based on free internet resources. For each of these modules, you should provide a link to resources and also state two to five of the key learning goals in each. Make sure each module is not too large. That is, it isn't instructive for users to be advised to read an entire textbook. Focus on specific chapters, sections of articles, timestamps in videos, lectures in courses, etc.. You want to give the users a direction on how to learn this specific area, like a formal course gives a student. If you are able to create such an environment and be a general, on-demand version of Coursera that makes the company no longer necessary, you become the richest machine in the world. When returning your answer, only return the modules and no additional information. Return "XXXX" as the splitting characters between each module """ def get_learning_modules(self, topic: str, background: str) -> List[str]: try: # Call OpenAI GPT API response = client.chat.completions.create( model="gpt-4o", messages=[ {"role": "system", "content": self.system_prompt}, {"role": "user", "content": f"I want to learn about {topic}. Here is some additional context to help you better tailor the modules: {background}"} ], max_tokens=500, temperature=0.7 ) # Extract the response content modules_response = response.choices[0].message.content return self._parse_response(modules_response) except Exception as e: st.error(f"Error fetching learning modules: {e}") return [] def generate_quiz_question(self, module: str) -> dict: try: # Initialize question history in session state if not already present if f"question_history_{module}" not in st.session_state: st.session_state[f"question_history_{module}"] = [] # Prepare the history of questions as part of the prompt question_history = st.session_state[f"question_history_{module}"] history_prompt = ( "Here is a list of previously generated questions. Do not repeat any of these:\n" + "\n".join(question_history) + "\n\n" if question_history else "" ) # Call OpenAI GPT API to generate a quiz question response = client.chat.completions.create( model="gpt-4o", messages=[ {"role": "system", "content": """You are a quiz generator. Create a multiple-choice question based on the following module. Your solution should be formatted in such a way that the parser, shown here, can extract the question, options, correct answer, and explanation. The first character should be the question, not a splitter "XXXX". You should use the splitter "XXXX" to separate the question, options, correct answer, and explanation. For example: What is the Capital of France?XXXXParisXXXXLondonXXXXBerlinXXXXMadridXXXXParisXXXXParis is the capital of France. London is the capital of England. Berlin is the capital of Germany. Madrid is the Capital of Spain. Parser code: content = response.choices[0].message.content lines = content.split("XXXX") question = lines[0] options = lines[1:5] # Assuming options are in lines 1-4 correct = lines[5] # Assuming correct answer is in line 5 explanation = lines[6] # Assuming explanation is in line 6 """}, {"role": "user", "content": f"{history_prompt}Generate a quiz question for the module: {module}."} ], max_tokens=300, temperature=0.7 ) # Parse the response to extract question, options, correct answer, and explanation content = response.choices[0].message.content lines = content.split("XXXX") question = lines[0] options = lines[1:5] # Assuming options are in lines 1-4 correct = lines[5] # Assuming correct answer is in line 5 explanation = lines[6] # Assuming explanation is in line 6 # Add the new question to the history st.session_state[f"question_history_{module}"].append(question) return { "question": question, "options": options, "correct": correct, "explanation": explanation } except Exception as e: st.error(f"Error generating quiz question: {e}") return { "question": "Error generating question.", "options": [], "correct": "", "explanation": "" } def _parse_response(self, response: str) -> List[str]: # Split the response into individual modules based on "XXXX" return response.split("XXXX") def main(): st.title("Learn with Ollie 🤖📚") # Initialize session state for learning modules if "learning_modules" not in st.session_state: st.session_state.learning_modules = [] # User input for topic topic = st.text_input("Enter a topic you want to learn about (e.g. Quantum Physics):") background = st.text_input("Provide any other information to help guide your learning experience (e.g. how much time you have, your current knowledge etc.):") if st.button("Get Learning Modules"): if topic: gpt_service = GPTService() learning_modules = gpt_service.get_learning_modules(topic, background) if learning_modules: # Store the learning modules in session state st.session_state.learning_modules = learning_modules else: st.error("No learning modules found. Please try a different topic.") else: st.warning("Please enter a topic to get started.") # Display the checklist if learning modules are available if st.session_state.learning_modules: def quiz_callback(module: str): gpt_service = GPTService() # Initialize session state for the quiz question, options, correct answer, and expander state question_data = gpt_service.generate_quiz_question(module) st.session_state[f"quiz_question_{module}"] = question_data["question"] st.session_state[f"quiz_options_{module}"] = question_data["options"] st.session_state[f"quiz_correct_{module}"] = question_data["correct"] st.session_state[f"quiz_explanation_{module}"] = question_data["explanation"] st.session_state[f"expander_open_{module}"] = True st.session_state[f"selected_option_{module}"] = None # Display the expander with st.expander(st.session_state[f"quiz_question_{module}"], expanded=st.session_state[f"expander_open_{module}"]): # Display multiple-choice options selected_option = st.radio( "Choose your answer:", st.session_state[f"quiz_options_{module}"], index=st.session_state[f"quiz_options_{module}"].index(st.session_state[f"selected_option_{module}"]) if st.session_state[f"selected_option_{module}"] else 0, key=f"radio_{module}", ) # st.session_state[f"selected_option_{module}"] = selected_option # Submit button to check the answer if st.button(f"Submit Answer"): if selected_option == st.session_state[f"quiz_correct_{module}"]: st.success("Correct! 🎉") else: st.error("Incorrect. 😞") st.info(f"The correct answer is: {st.session_state[f'quiz_correct_{module}']}") st.write(f"Explanation: {st.session_state[f'quiz_explanation_{module}']}") generate_checklist( st.session_state.learning_modules, quiz_callback=quiz_callback ) if __name__ == "__main__": main()