Add Issue Tracking Framework and Initial Monitoring Issues

docs/issues/1-bug-seen-submission-race-condition.md

@@ -0,0 +1,123 @@
+# Seen Submission Race Condition
+
+**Priority**: 1 **Type**: bug **Status**: open **Created**: 2025-07-04
+**Updated**: 2025-07-04
+
+## Description
+
+Race condition exists between filtering submissions and marking them as seen,
+potentially allowing duplicate notifications if the monitoring loop runs
+multiple times rapidly or if API responses change between calls.
+
+## Reproduction Steps
+
+1. Set up monitoring targets with very short poll intervals (difficult to
+   reproduce consistently)
+2. Force multiple rapid checks using `.trigger` command
+3. Monitor for duplicate notifications of the same submission
+4. Check database for timing issues in `seen_submissions` table
+
+## Expected vs Actual Behavior
+
+- **Expected**: Each submission should only be notified once, regardless of
+  timing
+- **Actual**: Rapid polling or API inconsistencies could cause duplicate
+  notifications
+
+## Technical Details
+
+### Code Location
+
+- **File**: `src/cogs/runner.py`
+- **Functions**: `filter_new_submissions()` (line 386-398),
+  `post_submissions()`, `mark_submissions_seen()`
+
+### Race Condition Flow
+
+```python
+# Time T1: First check starts
+submissions = await fetch_submissions_for_location(...)  # API call
+new_submissions = self.db.filter_new_submissions(channel_id, submissions)  # Query DB
+
+# Time T2: Second check starts (before first completes)
+submissions_2 = await fetch_submissions_for_location(...)  # Same API response
+new_submissions_2 = self.db.filter_new_submissions(channel_id, submissions_2)  # Same query result
+
+# Time T3: Both post notifications
+await self.post_submissions(new_submissions)  # Posts duplicates
+await self.post_submissions(new_submissions_2)  # Posts duplicates
+
+# Time T4: Both mark as seen
+self.db.mark_submissions_seen(channel_id, submission_ids)  # Second call gets IntegrityError
+```
+
+### Database Protection
+
+- `seen_submissions` table has unique constraint on
+  `(channel_id, submission_id)`
+- `mark_submissions_seen()` handles `IntegrityError` gracefully
+- **Problem**: Notifications already sent before constraint violation
+
+### Potential Triggers
+
+- Multiple rapid manual checks (`.trigger` command)
+- Cloud Run scaling events causing multiple instances
+- API response timing variations
+- Database connection delays
+
+## Proposed Solutions
+
+### Option 1: Atomic Transaction
+
+```python
+async def process_submissions_atomically(self, channel_id, submissions):
+    async with self.db.transaction():
+        new_submissions = self.db.filter_new_submissions(channel_id, submissions)
+        if new_submissions:
+            self.db.mark_submissions_seen(channel_id, [s["id"] for s in new_submissions])
+            await self.post_submissions(new_submissions)
+```
+
+### Option 2: Channel-Level Locking
+
+```python
+class Runner:
+    def __init__(self):
+        self.channel_locks = {}
+
+    async def run_checks_for_channel(self, channel_id, config, is_manual_check=False):
+        if channel_id not in self.channel_locks:
+            self.channel_locks[channel_id] = asyncio.Lock()
+
+        async with self.channel_locks[channel_id]:
+            # Existing check logic
+```
+
+### Option 3: Submission ID Tracking
+
+```python
+# Track recently processed submission IDs in memory
+self.recently_processed = {}  # {channel_id: set(submission_ids)}
+
+async def filter_new_submissions(self, channel_id, submissions):
+    # Filter by database AND recent memory
+    recent = self.recently_processed.get(channel_id, set())
+    truly_new = [s for s in db_filtered if s["id"] not in recent]
+    return truly_new
+```
+
+## Acceptance Criteria
+
+- [ ] No duplicate notifications under rapid polling scenarios
+- [ ] Multiple manual checks don't cause duplicates
+- [ ] Database integrity maintained
+- [ ] Performance impact minimal
+- [ ] Cloud Run scaling doesn't trigger race conditions
+- [ ] Test with concurrent `.trigger` commands
+
+## Notes
+
+- Difficult to reproduce consistently in testing
+- May require load testing to verify fix
+- Related to Cloud Run scaling and multiple instances
+- Consider distributed locking for multi-instance deployments

docs/issues/1-bug-startup-duplicate-notifications.md

@@ -0,0 +1,90 @@
+# Startup Duplicate Notifications
+
+**Priority**: 1 **Type**: bug **Status**: open **Created**: 2025-07-04
+**Updated**: 2025-07-04
+
+## Description
+
+When the bot starts up, it sends notifications for old submissions that have
+already been shown previously. The startup check bypasses normal polling logic
+and treats all submissions as "new" even though they were previously processed.
+
+## Reproduction Steps
+
+1. Start the bot with existing monitoring targets
+2. Observe that one or more channels receive notifications
+3. Check log timestamps - notifications are for submissions that occurred before
+   the current session
+4. Compare with previous bot runs - same submissions are re-notified
+
+## Expected vs Actual Behavior
+
+- **Expected**: Startup should only initialize monitoring, not send duplicate
+  notifications
+- **Actual**: Startup sends notifications for previously seen submissions
+
+## Technical Details
+
+### Code Location
+
+- **File**: `src/cogs/runner.py`
+- **Function**: `_run_startup_checks()` (lines 457-508)
+- **Issue**: Calls `run_checks_for_channel(is_manual_check=False)` during
+  startup
+
+### Root Cause
+
+1. `_run_startup_checks()` runs immediately when bot becomes ready
+2. Calls `run_checks_for_channel()` with `is_manual_check=False`
+3. Since `last_poll_at` is None, API fetches all recent submissions without date
+   filtering
+4. `filter_new_submissions()` has no seen submissions in database yet (fresh
+   start)
+5. All submissions get posted as notifications
+6. Submissions are then marked as seen, preventing future duplicates
+
+### Database State
+
+- `seen_submissions` table is empty on startup
+- `last_poll_at` is None for channels that haven't been polled
+- API returns submissions from past 24 hours without filtering
+
+## Proposed Solution
+
+### Option 1: Change Startup to Manual Check
+
+```python
+# In _run_startup_checks()
+result = await self.run_checks_for_channel(channel_id, config, is_manual_check=True)
+```
+
+This would show "Last 5 submissions" format instead of posting as notifications.
+
+### Option 2: Mark as Seen Without Posting
+
+```python
+# In _run_startup_checks()
+submissions = await self._fetch_submissions_for_channel(channel_id, config)
+if submissions:
+    self.db.mark_submissions_seen(channel_id, [s["id"] for s in submissions])
+# Don't call post_submissions()
+```
+
+### Option 3: Skip Startup Checks Entirely
+
+Remove `_run_startup_checks()` and let normal 1-minute polling handle first
+check.
+
+## Acceptance Criteria
+
+- [ ] Bot startup does not trigger duplicate notifications
+- [ ] Normal polling behavior unchanged
+- [ ] Manual check commands still show recent submissions
+- [ ] No regression in monitoring functionality
+- [ ] Test with multiple channels and monitoring targets
+
+## Notes
+
+- This issue is particularly noticeable with active monitoring targets
+- Users report confusion when old submissions appear as "new"
+- May be related to timezone handling in date filtering logic

docs/issues/2-bug-timezone-handling-inconsistency.md

@@ -0,0 +1,113 @@
+# Timezone Handling Inconsistency
+
+**Priority**: 2 **Type**: bug **Status**: open **Created**: 2025-07-04
+**Updated**: 2025-07-04
+
+## Description
+
+Inconsistent timezone handling across the application may cause polling schedule
+issues, incorrect date filtering in API calls, and database timestamp problems.
+
+## Reproduction Steps
+
+1. Review datetime handling in different parts of the codebase
+2. Check for naive vs timezone-aware datetime objects
+3. Monitor for polling schedule irregularities
+4. Verify API date filtering works correctly across time zones
+
+## Expected vs Actual Behavior
+
+- **Expected**: All datetime operations should be timezone-aware and consistent
+- **Actual**: Mix of naive and timezone-aware datetime objects causing potential
+  issues
+
+## Technical Details
+
+### Code Locations
+
+- **File**: `src/database.py` - `get_channel_config()` (lines 113-131)
+- **File**: `src/cogs/runner.py` - Various datetime operations
+- **File**: `src/api.py` - Date filtering in API calls
+
+### Specific Issues
+
+#### Database Operations
+
+```python
+# In get_channel_config() - timezone conversion happens here
+if config.last_poll_at and config.last_poll_at.tzinfo is None:
+    config.last_poll_at = config.last_poll_at.replace(tzinfo=timezone.utc)
+```
+
+#### Polling Schedule Logic
+
+```python
+# In _should_poll_channel() - may use naive datetime
+current_time = datetime.now()  # Potentially naive
+time_since_last_poll = current_time - config.last_poll_at  # Mixed types
+```
+
+#### API Date Filtering
+
+```python
+# Date calculations for min_date_of_submission parameter
+yesterday = datetime.now() - timedelta(days=1)  # Potentially naive
+date_str = yesterday.strftime("%Y-%m-%d")  # May not account for timezone
+```
+
+### Potential Impact
+
+- **Polling Schedule**: Channels may poll too frequently or not frequently
+  enough
+- **API Filtering**: Submissions might be missed or duplicated due to date range
+  issues
+- **Database Consistency**: Timestamps may not align properly
+- **Multi-timezone Users**: Different behavior based on server vs user timezones
+
+## Proposed Solution
+
+### Centralized Timezone Handling
+
+```python
+# Create utility functions for consistent datetime handling
+from datetime import datetime, timezone
+import pytz
+
+def now_utc() -> datetime:
+    """Get current UTC time, always timezone-aware"""
+    return datetime.now(timezone.utc)
+
+def ensure_utc(dt: datetime) -> datetime:
+    """Ensure datetime is timezone-aware and in UTC"""
+    if dt.tzinfo is None:
+        # Assume naive datetime is UTC
+        return dt.replace(tzinfo=timezone.utc)
+    return dt.astimezone(timezone.utc)
+
+def format_api_date(dt: datetime) -> str:
+    """Format datetime for API calls consistently"""
+    return ensure_utc(dt).strftime("%Y-%m-%d")
+```
+
+### Update All Datetime Operations
+
+1. Replace `datetime.now()` with `now_utc()`
+2. Ensure all database datetime fields are timezone-aware
+3. Standardize API date formatting
+4. Add timezone validation in tests
+
+## Acceptance Criteria
+
+- [ ] All datetime objects are timezone-aware
+- [ ] Polling schedules work correctly regardless of server timezone
+- [ ] API date filtering is consistent and accurate
+- [ ] Database timestamps are properly handled
+- [ ] No regression in existing functionality
+- [ ] Add timezone-specific tests
+
+## Notes
+
+- May require database migration if existing timestamps are naive
+- Consider impact on existing data
+- Test with different server timezone configurations
+- Document timezone assumptions for deployment