Bot Version 1.1.0

app/bot.py

@@ -9,8 +9,7 @@ from discord.ext import tasks
 from .config import settings
 from .db import init_db, insert_check, fetch_checks_since,fetch_month_checks
 from .utils import (
-    check_site, summarize_counts, get_site_names,
-    render_bar, compute_uptime, format_detection_reason,
+    check_site, summarize_counts, get_site_names, render_bar, compute_uptime, format_detection_reason
 )
  
 TOKEN = settings.discord_secret_key
@@ -24,6 +23,130 @@ SITE_CHOICES = [
     for site in MONITORED_SITES
 ]
 
+# ---------------------------------------------------------------------------
+# State tracking — populated on startup, updated each poll
+# ---------------------------------------------------------------------------
+# Holds the last known result string per site name: {"Site1": "up", ...}
+last_known_state: dict[str, str] = {}
+ 
+# Tracks when we last sent an alert per site to enforce the cooldown
+last_alerted_at: dict[str, datetime] = {}
+
+
+# ---------------------------------------------------------------------------
+# Alert helpers
+# ---------------------------------------------------------------------------
+ 
+# Transitions that warrant an alert
+ALERT_TRANSITIONS = {
+    ("up",       "down"),
+    ("up",       "degraded"),
+    ("degraded", "down"),
+    ("down",     "up"),       # recovery
+    ("degraded", "up"),       # recovery
+}
+ 
+INCIDENT_EMOJI = {
+    "down":     "🟥",
+    "degraded": "🟨",
+    "up":       "🟩",
+}
+ 
+ 
+async def maybe_send_alert(
+    site_name: str,
+    prev: str,
+    curr: str,
+    result: dict,
+    now: datetime,
+) -> None:
+    """Send an alert to the alerts channel if conditions are met."""
+ 
+    # Alerts disabled
+    if not settings.alert_channel_id:
+        return
+ 
+    # Not a transition we care about
+    if (prev, curr) not in ALERT_TRANSITIONS:
+        return
+ 
+    is_recovery = curr == "up"
+ 
+    # Cooldown — recoveries always bypass so you always know when a site is back
+    if not is_recovery:
+        last = last_alerted_at.get(site_name)
+        cooldown = timedelta(minutes=settings.alert_cooldown_minutes)
+        if last and (now - last) < cooldown:
+            return
+ 
+    last_alerted_at[site_name] = now
+ 
+    channel = client.get_channel(settings.alert_channel_id)
+    if channel is None:
+        print(f"Alert channel {settings.alert_channel_id} not found.")
+        return
+ 
+    await channel.send(_build_alert_message(site_name, prev, curr, result, now))
+ 
+ 
+def _build_alert_message(
+    site_name: str,
+    prev: str,
+    curr: str,
+    result: dict,
+    now: datetime,
+) -> str:
+    is_recovery = curr == "up"
+    emoji = INCIDENT_EMOJI.get(curr, "⬛")
+    title = "RECOVERED" if is_recovery else curr.upper()
+    timestamp = now.strftime("%H:%M UTC")
+ 
+    lines = [
+        f"{emoji} **{title}** | {site_name}",
+        f"Status: `{prev}` → `{curr}`",
+    ]
+ 
+    if is_recovery:
+        if result.get("latency_ms") is not None:
+            lines.append(f"Latency: {result['latency_ms']}ms")
+        # Approximate downtime from DB
+        downtime = _approximate_downtime(site_name, now)
+        if downtime:
+            lines.append(f"Downtime: ~{downtime}")
+    else:
+        reason = format_detection_reason(result.get("detection_reason"))
+        if reason:
+            lines.append(f"Reason: {reason}")
+        if result.get("http_status"):
+            lines.append(f"HTTP status: {result['http_status']}")
+ 
+    lines.append(f"Checked: {timestamp}")
+    return "\n".join(lines)
+ 
+ 
+def _approximate_downtime(site_name: str, now: datetime) -> str | None:
+    """
+    Walk back through recent checks to find the last 'up' row and return
+    a human-readable duration string, e.g. '~45 min' or '~2 hr 10 min'.
+    """
+    rows = fetch_checks_since(site_name, now - timedelta(days=1))
+    last_up_at = None
+    for checked_at, result, *_ in reversed(rows):
+        if result == "up":
+            last_up_at = datetime.fromisoformat(checked_at)
+            break
+ 
+    if last_up_at is None:
+        return None
+ 
+    delta = now - last_up_at
+    total_minutes = int(delta.total_seconds() // 60)
+    if total_minutes < 60:
+        return f"{total_minutes} min"
+    hours, mins = divmod(total_minutes, 60)
+    return f"{hours} hr {mins} min" if mins else f"{hours} hr"
+
+
 intents = discord.Intents.default()
 client = discord.Client(intents=intents)
 tree = app_commands.CommandTree(client)
@@ -33,38 +156,55 @@ session: aiohttp.ClientSession | None = None
 async def on_ready():
     global session
     init_db()
-
+ 
     if session is None:
         session = aiohttp.ClientSession(headers={"User-Agent": "VoteUptimeBot/1.0"})
-
+ 
+    # Seed state from the most recent DB row per site so we don't false-alert
+    # on restart.
+    for site in MONITORED_SITES:
+        rows = fetch_checks_since(
+            site["name"], datetime.now(timezone.utc) - timedelta(hours=1)
+        )
+        if rows:
+            last_known_state[site["name"]] = rows[-1][1]  # result column
+ 
     if not poll_sites.is_running():
         poll_sites.start()
-
+ 
     await tree.sync()
     print(f"Logged in as {client.user}")
 
-@tasks.loop(minutes=15)
+@tasks.loop(minutes=settings.poll_interval_minutes)
 async def poll_sites():
-    now = datetime.now(timezone.utc).isoformat()
+    now = datetime.now(timezone.utc)
+    now_iso = now.isoformat()
  
     for site in MONITORED_SITES:
         result = await check_site(session, site)
-        # Combine notes + detection_reason into the notes field so the DB
-        # schema doesn't need a migration.  error_type already carries reason.
+        site_name = site["name"]
+        new_state = result["result"]
+ 
         notes = result["notes"]
         if result.get("detection_reason") and not notes:
             notes = format_detection_reason(result["detection_reason"])
  
         insert_check(
-            site_name=site["name"],
-            checked_at=now,
+            site_name=site_name,
+            checked_at=now_iso,
             http_status=result["http_status"],
             latency_ms=result["latency_ms"],
-            result=result["result"],
-            error_type=result["error_type"],   # now carries detection_reason too
+            result=new_state,
+            error_type=result["error_type"],
             notes=notes,
         )
-
+ 
+        # --- Alert logic ---
+        prev_state = last_known_state.get(site_name)
+        last_known_state[site_name] = new_state
+ 
+        if prev_state is not None and prev_state != new_state:
+            await maybe_send_alert(site_name, prev_state, new_state, result, now)
 
 @poll_sites.before_loop
 async def before_poll_sites():