import gradio as gr
import os
import re
import time
import base64
from openai import OpenAI
from together import Together
from PIL import Image
import io
def generate_math_solution_openrouter(api_key, problem_text, history=None):
if not api_key.strip():
return "Please enter your OpenRouter API key.", history
if not problem_text.strip():
return "Please enter a math problem.", history
try:
client=OpenAI(
base_url="https://openrouter.ai/api/v1",
api_key=api_key
)
messages= [
{"role": "system", "content":
"""You are an expert math tutor who explains concepts clearly and thoroughly.
Analyze the given math problem and provide a detailed step-by-step solution.
For each step:
1. Show the mathematical operation
2. Explain why this step is necessary
3. Connect it to relevant mathematical concepts
Format your response with clear section headers using markdown.
Begin with an "Initial Analysis" section, follow with numbered steps,
and conclude with a "Final Answer" section."""},
]
if history:
for exchange in history:
messages.append({"role": "user", "content": exchange[0]}) # asking a math prob solution
if exchange[1]: # Check if there's a response
messages.append({"role": "assistant", "content": exchange[1]}) # getting the step by step solution of asked math prob
# Add the current problem
messages.append({"role": "user", "content": f"Solve this math problem step-by-step: {problem_text}"}) # what is 2+2? = problem_text
# calling the model
completion=client.chat.completions.create(
model="microsoft/phi-4-reasoning-plus:free",
messages=messages,
extra_headers={
"HTTP-Referer": "https://advancedmathtutor.edu",
"X-Title": "Advanced Math Tutor",
}
)
solution=completion.choices[0].message.content # finally you are getting solution of math prob here
# Update history
if history is None:
history = []
history.append((problem_text, solution)) # user: what is 2+2? => LLM: 4
return solution, history
except Exception as e:
error_message = f"Error: {str(e)}"
return error_message, history
# Function to convert image to base64
def image_to_base64(image_path):
if image_path is None:
return None
try:
with open(image_path, "rb") as img_file:
return base64.b64encode(img_file.read()).decode("utf-8")
except Exception as e:
print(f"Error converting image to base64: {str(e)}")
return None
# Function to generate math solution using Together AI with support for images
def generate_math_solution_together(api_key, problem_text, image_path=None, history=None):
if not api_key.strip():
return "Please enter your Together AI API key.", history
if not problem_text.strip() and image_path is None:
return "Please enter a math problem or upload an image of a math problem.", history
try:
client=Together(api_key=api_key)
messages= [
{
"role": "system",
"content": """You are an expert math tutor who explains concepts clearly and thoroughly.
Analyze the given math problem and provide a detailed step-by-step solution.
For each step:
1. Show the mathematical operation
2. Explain why this step is necessary
3. Connect it to relevant mathematical concepts
Format your response with clear section headers using markdown.
Begin with an "Initial Analysis" section, follow with numbered steps,
and conclude with a "Final Answer" section."""
}
]
# Add conversation history if it exists
if history:
for exchange in history:
messages.append({"role": "user", "content": exchange[0]})
if exchange[1]: # Check if there's a response
messages.append({"role": "assistant", "content": exchange[1]})
# Prepare the user message content # image upload + text details to help solve the problem in image
user_message_content = []
# Add text content if provided #extra info provided # image of 2+2=? + pls show detailed explanation
if problem_text.strip():
user_message_content.append({
"type": "text",
"text": f"Solve this math problem: {problem_text}"
})
else:
user_message_content.append({ #no extra info
"type": "text",
"text": "Solve this math problem from the image:" # image of 2+2=?
})
# Add image if provided
if image_path:
# Convert image to base64
base64_image = image_to_base64(image_path)
if base64_image:
user_message_content.append({
"type": "image_url", #together url format for img support in case of image models
"image_url": {
"url": f"data:image/jpeg;base64,{base64_image}"
}
})
# Add the user message with content
messages.append({
"role": "user",
"content": user_message_content
})
# Create the completion /calling the model
response = client.chat.completions.create(
model="meta-llama/Llama-Vision-Free",
messages=messages,
stream=False
)
solution = response.choices[0].message.content
# Update history - for simplicity, just store the text problem
if history is None:
history = []
history.append((problem_text if problem_text.strip() else "Image problem", solution))
return solution, history
except Exception as e:
error_message = f"Error: {str(e)}"
return error_message, history
### what the gradio website interface will look like? answer korbo
def create_demo():
with gr.Blocks(theme=gr.themes.Soft(primary_hue="blue")) as demo:
gr.Markdown("# 📚 Advanced Math Tutor")
gr.Markdown("""
This application provides step-by-step solutions to math problems using advanced AI models.
Choose between OpenRouter's Phi-4-reasoning-plus for text-based problems or Together AI's
Llama-Vision for problems with images.
""")
### pages/tabs = 2 tabs
with gr.Tabs():
# Text-based problem solver (OpenRouter)
with gr.TabItem("Text Problem Solver (OpenRouter)/PHI"):
with gr.Row():
with gr.Column(scale=1):
openrouter_api_key = gr.Textbox(
label="OpenRouter API Key",
placeholder="Enter your OpenRouter API key (starts with sk-or-)",
type="password"
)
text_problem_input = gr.Textbox(
label="Math Problem",
placeholder="Enter your math problem here...",
lines=5
)
example_problems = gr.Examples(
examples=[
["Solve the quadratic equation: 3x² + 5x - 2 = 0"],
["Find the derivative of f(x) = x³ln(x)"],
["Calculate the area of a circle with radius 5 cm"],
["Find all values of x that satisfy the equation: logâ‚‚(x-1) + logâ‚‚(x+3) = 5"]
],
inputs=[text_problem_input],
label="Example Problems"
)
with gr.Row():
openrouter_submit_btn = gr.Button("Solve Problem", variant="primary")
openrouter_clear_btn = gr.Button("Clear")
with gr.Column(scale=2):
openrouter_solution_output = gr.Markdown(label="Solution")
# Store conversation history (invisible to user)
openrouter_conversation_history = gr.State(value=None)
# Button actions
openrouter_submit_btn.click(
fn=generate_math_solution_openrouter,
inputs=[openrouter_api_key, text_problem_input, openrouter_conversation_history],
outputs=[openrouter_solution_output, openrouter_conversation_history]
)
openrouter_clear_btn.click(
fn=lambda: ("", None),
inputs=[],
outputs=[openrouter_solution_output, openrouter_conversation_history]
)
# Image-based problem solver (Together AI)
with gr.TabItem("Image Problem Solver (Together AI)"):
with gr.Row():
with gr.Column(scale=1):
together_api_key = gr.Textbox(
label="Together AI API Key",
placeholder="Enter your Together AI API key",
type="password"
)
together_problem_input = gr.Textbox(
label="Problem Description (Optional)",
placeholder="Enter additional context for the image problem...",
lines=3
)
together_image_input = gr.Image(
label="Upload Math Problem Image",
type="filepath"
)
with gr.Row():
together_submit_btn = gr.Button("Solve Problem", variant="primary")
together_clear_btn = gr.Button("Clear")
with gr.Column(scale=2):
together_solution_output = gr.Markdown(label="Solution")
# Store conversation history (invisible to user)
together_conversation_history = gr.State(value=None)
# Button actions
together_submit_btn.click(
fn=generate_math_solution_together,
inputs=[together_api_key, together_problem_input, together_image_input, together_conversation_history],
outputs=[together_solution_output, together_conversation_history]
)
together_clear_btn.click(
fn=lambda: ("", None),
inputs=[],
outputs=[together_solution_output, together_conversation_history]
)
return demo
### launch the app
if __name__ == "__main__":
demo=create_demo()
demo.launch()