Update app.py

app.py

@@ -11,50 +11,53 @@ import re
 import ast
 import time
 from collections import deque
-from dotenv import load_dotenv
 import gradio as gr
 from google import genai
 
-load_dotenv()
-
 
 # ---------------------------------------------------------------------------
 # MODULE 3 — Pre-processing Module
 # ---------------------------------------------------------------------------
-def pre_processing(description, code):
-    build_prompt = f"""
-#ROLE
-You are a Python code reviewer. Your goal is to determine if the provided Python code strictly complies with the requirements mentioned by the user.
- 
-#CONTEXT
-Developers may create code that does not entirely meet the specified requirements. You will examine the relationship between a natural-language description and a Python code sample and create a structured evaluation report.
- 
-#INPUT DATA
-Description: {description}
-Code: ```python
-{code}
+def build_prompt(description: str, code: str) -> str:
+    prompt = f"""<start_of_turn>user
+### ROLE
+You are an expert Python code reviewer. Your sole task is to determine whether the
+provided Python code correctly implements the behaviour described in the user's
+requirements description.
+### CONTEXT
+Developers sometimes write code that does not fully satisfy the requirements they
+were given. You will analyse the semantic relationship between a natural-language
+description and a Python code snippet, then produce a structured evaluation report.
+### INPUT DATA
+**Requirements Description:**
+{description.strip()}
+**Python Code:**
+```python
+{code.strip()}
 ```
- 
-#TASK
-1. Carefully read the requirements description.
-2. Examine each line of the Python code.
+### TASK
+1. Read the requirements description carefully.
+2. Analyse the Python code line by line.
 3. Determine whether the code fulfils ALL requirements stated in the description.
-4. Estimate the accuracy in percentages to assess how accurately the code matches the description.
-5. List specific requirements that are either absent or incorrectly applied.
-6. Base your evaluation solely on static code analysis and logical reasoning.
- 
-#CONSTRAINTS
-1. Do not execute the code.
-2. The output must be in the specified format.
-3. The output should be clear and concise.
- 
-#OUTPUT FORMAT (to be strictly followed without any changes)
-Result: either "Pass" or "Fail" (Pass if the code is functional and matches the description, Fail if it has bugs or doesn't match).
-Accuracy: an integer from 0 to 100 representing how well the code matches the description and is bug-free.
-Review: a short prose explanation of the result (e.g. "Code correctly handles all described requirements including edge cases.").    
-Issues: The value should be a single string containing a numbered list. For each item, include the severity (Error, Warning, or Info) followed by a dash and description of the issue (e.g. "Missing type hints on function signature"). List each issue on a new line starting with a number. Use newline characters (\n) to separate each line.
+4. Estimate an accuracy percentage (0–100) reflecting how completely the code
+   matches the description.
+5. List any specific requirements that are missing or incorrectly implemented.
+### CONSTRAINTS
+- Do NOT execute the code.
+- Base your evaluation solely on static code analysis and logical reasoning.
+- Keep feedback concise, clear, and actionable.
+- Your response MUST follow the output format exactly.
+### OUTPUT FORMAT
+Respond only with the following structure — no extra text before or after:
+RESULT: <PASS or FAIL>
+ACCURACY: <integer 0-100>%
+SUMMARY: <one sentence overall assessment>
+ISSUES:
+- <issue 1, or "None" if code fully matches the description>
+- <issue 2>
 ...
- 
+<end_of_turn>
+<start_of_turn>model
 """
     return prompt