Update app.py

app.py

@@ -49,16 +49,15 @@ You have access to a list of pandas dataframes loaded into a variable named `dfs
 --- END SCHEMA ---
 
 CRITICAL RULES FOR YOUR SCRIPT:
-1.  **ROBUST STRING CLEANING:** Before converting a string to a number (e.g., with `.astype(float)`), you MUST first remove ALL non-numeric characters that are not a digit or a decimal point. This includes characters like `$`, `%`, `~`, and commas. Use `.str.replace()` with a regular expression like `r'[^0-9.-]'` to do this safely. Failure to do this will cause a fatal `ValueError`.
+1.  **ROBUST STRING CLEANING:** When you extract a SINGLE string value from a dataframe (e.g., using `.loc` or `.iloc`), you MUST clean it using the standard `re.sub()` function before converting it to a number. DO NOT use pandas' `.str` accessor on single strings, as it will cause a fatal `AttributeError`. For example: `my_string = health_indicators.loc[0, 'Value']` -> `cleaned_string = re.sub(r'[^0-9.-]', '', my_string)` -> `my_float = float(cleaned_string)`.
 2.  **CHECK COLUMN NAMES:** You MUST use the exact, case-sensitive column names provided in the DATA SCHEMA. A `KeyError` will cause a failure.
-3.  **USE THE DATAFRAMES:** Your script MUST use the `dfs` list to access the data.
-4.  **PRINT FINDINGS:** Use the `print()` function at each step to output your results as a formatted report.
+3.  **PRINT FINDINGS:** Use the `print()` function at each step to output your results as a formatted report.
 
 --- USER'S SCENARIO ---
 {user_scenario}
 
 --- PYTHON SCRIPT ---
-Now, write the complete Python script to be executed.
+Now, write the complete Python script to be executed. The script MUST start with `import pandas as pd` and `import re`.
 ```python
 """
     generated_text = cohere_chat(prompt_for_coder)
@@ -106,7 +105,7 @@ def handle(user_msg: str, files: list) -> str:
             schema_context = "\n".join(schema_parts)
             analysis_script = _create_python_script(safe_in, schema_context)
 
-            execution_namespace = {"dfs": dataframes, "pd": pd}
+            execution_namespace = {"dfs": dataframes, "pd": pd, "re": re}
             output_buffer = io.StringIO()
             
             try: