mohnish4246/ProblemSolver

1import gradio as gr2from openai import OpenAI3import pytesseract4from PIL import Image5import os6import tempfile7 8 9# --------------------------10# Setup API Key11# --------------------------12client = OpenAI(api_key=os.getenv("OPENAI_API_KEY"))13# --------------------------14# Global State15# --------------------------16math_messages = []17 18# --------------------------19# Process Uploaded/Drawn Image20# --------------------------21def process_image_with_ocr(image, shouldConvert=False):22    # Convert if needed (for Sketchpad)23    if shouldConvert:24        new_img = Image.new('RGB', size=(image.width, image.height), color=(255, 255, 255))25        new_img.paste(image, (0, 0), mask=image)26        image = new_img27 28    # Save image temporarily29    temp_file = tempfile.NamedTemporaryFile(delete=False, suffix=".png")30    image.save(temp_file.name)31 32    # Use Tesseract to extract text33    extracted_text = pytesseract.image_to_string(Image.open(temp_file.name))34    os.remove(temp_file.name)35 36    return extracted_text.strip()37 38# --------------------------39# Get Math Answer40# --------------------------41def get_math_response(image_description, user_question):42    if image_description:43        prompt = f"The following is a math problem extracted from an image:\n\n{image_description}\n\nUser question: {user_question}"44    else:45        prompt = f"User question: {user_question}"46 47    response = client.chat.completions.create(48        model="gpt-4",49        messages=[50            {"role": "system", "content": "You are a helpful math assistant. Solve problems carefully with step-by-step reasoning."},51            {"role": "user", "content": prompt}52        ],53        temperature=0.354    )55 56    return response.choices[0].message.content57# --------------------------58# Main Chat Function59# --------------------------60def math_chat_bot(image, sketchpad, question, state):61    current_tab_index = state["tab_index"]62    image_description = None63 64    if current_tab_index == 0:65        if image is not None:66            image_description = process_image_with_ocr(image)67    elif current_tab_index == 1:68        if sketchpad and sketchpad["composite"]:69            image_description = process_image_with_ocr(sketchpad["composite"], True)70 71    answer = get_math_response(image_description, question)72    yield answer73 74# --------------------------75# CSS for Inline LaTeX76# --------------------------77css = """78#qwen-md .katex-display { display: inline; }79#qwen-md .katex-display>.katex { display: inline; }80#qwen-md .katex-display>.katex>.katex-html { display: inline; }81"""82 83# --------------------------84# Gradio Interface85# --------------------------86with gr.Blocks(css=css) as demo:87    gr.HTML("""88    <p align="center"><img src="https://modelscope.oss-cn-beijing.aliyuncs.com/resource/qwen.png" style="height: 60px"/></p>89    <center><font size=8>📖 Qwen2.5-Math Demo</font></center>90    <center><font size=3>This WebUI uses Qwen2-VL for OCR and Qwen2.5-Math for reasoning. Input images or math problems below.</font></center>91    """)92 93    state = gr.State({"tab_index": 0})94 95    with gr.Row():96        with gr.Column():97            with gr.Tabs() as input_tabs:98                with gr.Tab("Upload"):99                    input_image = gr.Image(type="pil", label="Upload")100                with gr.Tab("Sketch"):101                    input_sketchpad = gr.Sketchpad(type="pil", label="Sketch", layers=False)102 103            def on_tab_select(e: gr.SelectData, s): s["tab_index"] = e.index104            input_tabs.select(fn=lambda e: {"tab_index": e.index}, inputs=None, outputs=state)105 106            input_text = gr.Textbox(label="Enter your question")107            with gr.Row():108                clear_btn = gr.ClearButton([input_image, input_sketchpad, input_text])109                submit_btn = gr.Button("Submit", variant="primary")110 111        with gr.Column():112            output_md = gr.Markdown(label="Answer", elem_id="qwen-md",113                                    latex_delimiters=[114                                        {"left": "\\(", "right": "\\)", "display": True},115                                        {"left": "\\[", "right": "\\]", "display": True},116                                        {"left": "\\begin{equation}", "right": "\\end{equation}", "display": True},117                                        {"left": "\\begin{align}", "right": "\\end{align}", "display": True},118                                        {"left": "\\begin{gather}", "right": "\\end{gather}", "display": True},119                                    ])120 121        submit_btn.click(122            fn=math_chat_bot,123            inputs=[input_image, input_sketchpad, input_text, state],124            outputs=output_md125        )126 127# --------------------------128# Launch129# --------------------------130if __name__ == "__main__":131    demo.launch()