app.py

# slack_reporter/app.py – nyckeltal från nystartad JSONL-logg
import os
import time
import schedule
import pandas as pd
import requests
import logging
import re
from datetime import datetime, timedelta
from huggingface_hub import hf_hub_download
import gradio as gr

# --- Loggning ---
logging.basicConfig(
    filename="slack_reporter.log",
    level=logging.INFO,
    format="%(asctime)s %(levelname)s %(message)s"
)

# --- Konfiguration ---
REPO_ID = "ChargeNodeEurope/logfiles"
LOG_FILENAME = "logs_v2/conversation_log_v2.txt"
WEBHOOK_URL = os.environ.get("SLACK_WEBHOOK_URL")
HF_TOKEN = os.environ.get("HF_TOKEN")

# --- Funktion: Skicka Slack-meddelande med nyckeltal ---
def send_summary_to_slack(summary_text):
    payload = {"text": f"📊 Daglig sammanställning:\n{summary_text}"}
    try:
        response = requests.post(WEBHOOK_URL, json=payload)
        logging.info(f"Slack response status: {response.status_code}")
        logging.info(f"Slack response text: {response.text}")
        response.raise_for_status()
        logging.info("Slack-meddelande postat.")
    except Exception as e:
        logging.error(f"Fel vid Slack-post: {e}", exc_info=True)
        raise

# --- Funktion: Bygg nyckeltalstext ---
def generate_summary(df):
    now = datetime.now()
    yesterday = now - timedelta(days=1)
    last_14 = now - timedelta(days=14)

    # Konvertera timestamp till datetime
    df["timestamp"] = pd.to_datetime(df["timestamp"], errors="coerce")
    # Filtrera loggar för senaste 24h och 14 dagar
    df_day = df[df["timestamp"] >= yesterday.replace(hour=0, minute=0, second=0)]
    df_14 = df[df["timestamp"] >= last_14]

    summary = []
    summary.append(f"📅 Datum: {now.strftime('%Y-%m-%d')}")
    summary.append(f"🗨️ Antal meddelanden senaste 24h: {len(df_day)}")
    summary.append(f"👥 Unika sessioner senaste 24h: {df_day['session_id'].nunique()}")

    # Beräkna snitt antal meddelanden per session (om minst en session finns)
    unique_sessions = df_day['session_id'].nunique()
    avg_messages_per_session = round(len(df_day) / unique_sessions, 2) if unique_sessions > 0 else 0
    summary.append(f"📊 Meddelanden per session (snitt): {avg_messages_per_session}")

    # Snittsvarstid (om kolumnen finns och har giltiga värden)
    if "response_time" in df_day.columns and not df_day["response_time"].empty:
        avg_response_time = round(df_day["response_time"].mean(), 2)
        summary.append(f"🕒 Snittsvarstid senaste 24h: {avg_response_time} sekunder")
    else:
        summary.append("🕒 Snittsvarstid senaste 24h: N/A")

    summary.append(f"📊 Antal meddelanden (14 dagar): {len(df_14)}")
    summary.append(f"📈 Unika användare (14 dagar): {df_14['user_id'].nunique()}")

    # Likes och dislikes under senaste 24h
    if "feedback" in df_day.columns:
        df_day_feedback = df_day.dropna(subset=["feedback"])
        likes = df_day_feedback[df_day_feedback["feedback"] == "up"].shape[0]
        dislikes = df_day_feedback[df_day_feedback["feedback"] == "down"].shape[0]
    else:
        likes, dislikes = 0, 0
    summary.append(f"👍 Likes senaste 24h: {likes}")
    summary.append(f"👎 Dislikes senaste 24h: {dislikes}")

    # Nyckeltal kring webbläsare: visa de tre mest använda
    if "browser" in df_day.columns:
        browser_counts = df_day["browser"].value_counts()
        top_browsers = browser_counts.head(3)
        browser_summary = "🌐 Browser-fördelning senaste 24h: " + ", ".join([f"{browser}: {count}" for browser, count in top_browsers.items()])
    else:
        browser_summary = "🌐 Browser-fördelning senaste 24h: Ingen data"
    summary.append(browser_summary)

    # Tre exempel på frågor och svar med tumme ner
    if "feedback" in df_day.columns:
        df_down = df_day[df_day["feedback"] == "down"].dropna(subset=["user_message", "bot_reply"])
    else:
        df_down = pd.DataFrame()
    if not df_down.empty:
        examples = df_down.head(3)
        examples_text = "❌ Exempel på frågor och svar med tumme ner:\n"
        for idx, row in examples.iterrows():
            examples_text += f"• Fråga: {row['user_message']}\n  Svar: {row['bot_reply']}\n"
    else:
        examples_text = "❌ Inga exempel med tumme ner hittades senaste 24h."
    summary.append(examples_text)

    return "\n".join(summary)

# --- Huvudfunktion ---
def run_report():
    try:
        logging.info("Startar rapportgenerering...")
        log_path = hf_hub_download(
            repo_id=REPO_ID,
            filename=LOG_FILENAME,
            repo_type="dataset",
            token=HF_TOKEN
        )

        try:
            df = pd.read_json(log_path, lines=True)
            logging.info("Loggfil läst som JSONL.")
        except Exception as e:
            logging.error(f"Kunde inte läsa logg som JSONL: {e}", exc_info=True)
            return "❌ Loggfilen kunde inte läsas som JSONL."

        if df.empty:
            logging.warning("Inga loggrader hittades – avbryter.")
            return "⚠️ Inga loggar att rapportera."

        summary = generate_summary(df)
        send_summary_to_slack(summary)
        return "✅ Slack-meddelande skickat."

    except Exception as e:
        logging.error(f"Fel vid körning: {e}", exc_info=True)
        return f"❌ Fel vid körning: {e}"

# --- Schemaläggning ---
schedule.every().day.at("08:20").do(run_report)

def run_schedule():
    while True:
        schedule.run_pending()
        time.sleep(60)

import threading
threading.Thread(target=run_schedule, daemon=True).start()

# --- Gradio-knapp ---
with gr.Blocks() as app:
    gr.Markdown("# Slack Reporter – Nyckeltal")
    btn = gr.Button("Skicka rapport nu")
    output = gr.Textbox()
    btn.click(fn=run_report, outputs=output)

if __name__ == "__main__":
    app.launch()