evaluate_tickets.py

import os
import time
import pandas as pd
from dotenv import load_dotenv
from openai import OpenAI, OpenAIError

load_dotenv()
client = OpenAI()

# Construye el prompt
def build_prompt(ticket, reply):
    return f"""
Evaluate the following AI-generated reply to a customer support ticket.

Ticket:
\"\"\"{ticket}\"\"\"

Reply:
\"\"\"{reply}\"\"\"

Please provide:

1. A content_score from 1 to 5 (based on relevance, correctness, and completeness).
2. A short content_explanation.
3. A format_score from 1 to 5 (based on clarity, structure, and grammar).
4. A short format_explanation.

Use the format:

Content Score: <score>
Content Explanation: <short explanation>
Format Score: <score>
Format Explanation: <short explanation>
"""

# Analiza el texto devuelto por el modelo
def parse_response(response_text):
    lines = response_text.strip().split("\n")
    try:
        content_score = int([l for l in lines if "Content Score:" in l][0].split(":")[1].strip())
        content_expl = [l for l in lines if "Content Explanation:" in l][0].split(":", 1)[1].strip()
        format_score = int([l for l in lines if "Format Score:" in l][0].split(":")[1].strip())
        format_expl = [l for l in lines if "Format Explanation:" in l][0].split(":", 1)[1].strip()
        return content_score, content_expl, format_score, format_expl
    except Exception as e:
        print("Error parsing response:\n", response_text)
        return None, "Parse error", None, "Parse error"

# Llama al modelo GPT
def evaluate_ticket(ticket, reply):
    prompt = build_prompt(ticket, reply)
    try:
        response = client.chat.completions.create(
            model="gpt-4o",
            messages=[{"role": "user", "content": prompt}],
            temperature=0.3,
        )
        return parse_response(response.choices[0].message.content)
    except OpenAIError as e:
        print(f"API error: {e}")
        return None, "API error", None, "API error"

# Ejecución principal
def main():
    df = pd.read_csv("tickets.csv")
    results = []

    for i, row in df.iterrows():
        print(f"Evaluating ticket {i + 1}/{len(df)}")
        ticket, reply = row["ticket"], row["reply"]
        content_score, content_expl, format_score, format_expl = evaluate_ticket(ticket, reply)
        results.append((content_score, content_expl, format_score, format_expl))
        time.sleep(1)

    df[["content_score", "content_explanation", "format_score", "format_explanation"]] = pd.DataFrame(results)
    df.to_csv("tickets_evaluated.csv", index=False)
    print("Done. Output saved to tickets_evaluated.csv")

if __name__ == "__main__":
    main()