Update app.py

app.py

@@ -1,75 +1,97 @@
 import gradio as gr
-import openai
 import pandas as pd
 import os
+import google.generativeai as genai
+import fitz  # PyMuPDF to read PDF
+
+# Load Gemini API key from Hugging Face secret
+genai.configure(api_key=os.getenv("AIzaSyDX-JAUDT7NB65Ry2P5ymEFlzKNohWKTlg"))
+model = genai.GenerativeModel("gemini-2.0-flash")
+
+# PDF reading utility
+def extract_text_from_pdf(pdf_file):
+    text = ""
+    doc = fitz.open(pdf_file.name)
+    for page in doc:
+        text += page.get_text()
+    return text.strip()
+
+# Gemini prompting logic
+def evaluate_prompt(question, solution, instructions):
+    prompt = f"""
+You are an expert math evaluator. Follow these instructions carefully:
 
-openai.api_key = os.getenv("OPENAI_API_KEY")  # Add this in the HF Secrets tab
-
-# 🔍 Prompt Generator
-def generate_prompt(question, solution, instructions):
-    return f"""
 {instructions}
 
 ---
 
-Question:
+**Question:**
 {question}
 
-Student's Solution:
+**Student's Solution:**
 {solution}
 
-Evaluate the student's solution and provide feedback.
+Evaluate the solution. Say whether it's correct or not. If incorrect, give a clear and detailed explanation of why it's wrong and how to fix it.
 
-End your response with:
-Is the solution correct? [Yes/No]
+Respond in this format:
+Correct?: [Yes/No]
+Reasoning: <explanation>
 """
-
-# 🧠 Evaluation Function
-def evaluate_csv(file, instructions):
-    df = pd.read_csv(file.name)
-    feedback_list = []
-    status_list = []
-
-    for idx, row in df.iterrows():
-        q = row['question']
-        s = row['solution']
-        prompt = generate_prompt(q, s, instructions)
-
+    response = model.generate_content(prompt)
+    return response.text.strip()
+
+# Handler for single text input
+def evaluate_text_inputs(question, solution, instructions_file, instructions_text):
+    instructions = instructions_text or extract_text_from_pdf(instructions_file) if instructions_file else ""
+    result = evaluate_prompt(question, solution, instructions)
+    return result
+
+# Handler for CSV batch input
+def evaluate_csv(csv_file, instructions_file, instructions_text):
+    instructions = instructions_text or extract_text_from_pdf(instructions_file) if instructions_file else ""
+    df = pd.read_csv(csv_file.name)
+    
+    evaluations = []
+    for _, row in df.iterrows():
+        q = row.get('question', '')
+        s = row.get('solution', '')
         try:
-            response = openai.ChatCompletion.create(
-                model="gpt-4",  # or gpt-3.5-turbo
-                messages=[{"role": "user", "content": prompt}],
-                temperature=0.2
-            )
-            output = response.choices[0].message.content
-            feedback_list.append(output)
-
-            # Extract Yes/No from output
-            status = "Yes" if "Yes" in output.splitlines()[-1] else "No"
-            status_list.append(status)
-
+            eval_result = evaluate_prompt(q, s, instructions)
+            evaluations.append(eval_result)
         except Exception as e:
-            feedback_list.append(f"Error: {e}")
-            status_list.append("Error")
-
-    df['AI_Feedback'] = feedback_list
-    df['Is_Correct'] = status_list
-
-    output_path = "evaluated_output.csv"
+            evaluations.append(f"Error: {e}")
+    
+    df['evaluation'] = evaluations
+    output_path = "evaluated_results.csv"
     df.to_csv(output_path, index=False)
     return output_path
 
-# 🎛️ Gradio UI
+# Gradio UI
 with gr.Blocks() as demo:
-    gr.Markdown("# 📊 Solution Evaluator (CSV)")
-    gr.Markdown("Upload your CSV with `question` and `solution` columns. Provide evaluation instructions.")
-    with gr.Row():
-        csv_input = gr.File(label="Upload CSV", file_types=[".csv"])
-        instructions_input = gr.Textbox(lines=6, label="Evaluation Instructions")
-    output_file = gr.File(label="Download Evaluated CSV")
-
-    run_btn = gr.Button("Evaluate Solutions")
-
-    run_btn.click(fn=evaluate_csv, inputs=[csv_input, instructions_input], outputs=output_file)
+    gr.Markdown("## 🤖 Solution Evaluator (Gemini API)")
+    gr.Markdown("Evaluate student solutions using Google's Gemini API.")
+
+    with gr.Tab("📝 Single Evaluation"):
+        with gr.Row():
+            question_input = gr.Textbox(label="Question", lines=3)
+            solution_input = gr.Textbox(label="Solution", lines=6)
+        with gr.Row():
+            instructions_text = gr.Textbox(label="Instructions (text)", lines=6)
+            instructions_file = gr.File(label="Upload Instructions PDF", file_types=[".pdf"])
+        output_single = gr.Textbox(label="Evaluation Result", lines=8)
+        btn_single = gr.Button("Evaluate Solution")
+        btn_single.click(fn=evaluate_text_inputs,
+                         inputs=[question_input, solution_input, instructions_file, instructions_text],
+                         outputs=output_single)
+
+    with gr.Tab("📄 Batch CSV Evaluation"):
+        csv_file = gr.File(label="Upload CSV (columns: question, solution)", file_types=[".csv"])
+        inst_text = gr.Textbox(label="Instructions (text)", lines=6)
+        inst_pdf = gr.File(label="Upload Instructions PDF", file_types=[".pdf"])
+        output_csv = gr.File(label="Download Evaluated CSV")
+        btn_batch = gr.Button("Evaluate All")
+        btn_batch.click(fn=evaluate_csv,
+                        inputs=[csv_file, inst_pdf, inst_text],
+                        outputs=output_csv)
 
 demo.launch()