AUTO-8 - trckr

← AUTO-8

backend core

Created: 2026-02-03 • Updated: 2026-02-03

Relationships Loading...

## Overview
Create SchedulerEngine class that polls for due automations and executes them.

## Tasks
- [ ] Create `server/scheduler.py` with SchedulerEngine class
- [ ] Implement scheduler loop:
  - Background thread polling every 60 seconds
  - Query `get_due_automations(now)` from store
  - For each due automation, spawn execution in thread pool
  - Handle concurrent runs safely
- [ ] Implement RRULE parsing:
  - `compute_next_run(rrule_str, after_dt)` using python-dateutil
  - Handle RRULE: prefix stripping
  - Update automation.next_run_at after each run
- [ ] Integrate components:
  - AutomationRunner for execution
  - FindingDetector for output parsing
  - TrckrIntegration for issue creation (if actionable)
  - NtfyIntegration for notifications
- [ ] Implement `_execute_automation(automation)`:
  - Create run record
  - Execute via runner
  - Parse findings
  - Create trckr issue if needed
  - Send ntfy notification
  - Update run record with results
  - Clean up worktree (or keep for review)

## QA: Live Integration Testing

**Phase Gate 1 - Module Imports:**
```bash
uv run python -c "from server.scheduler import SchedulerEngine, compute_next_run; print('OK')"
```
✓ PASS: Prints "OK"

**Phase Gate 2 - RRULE Parsing:**
```bash
uv run python -c "
from server.scheduler import compute_next_run
from datetime import datetime, timezone

# Daily at 9am
next_run = compute_next_run('FREQ=DAILY;BYHOUR=9;BYMINUTE=0', datetime.now(timezone.utc))
print(f'Next run: {next_run}')
assert next_run is not None
assert next_run > datetime.now(timezone.utc).timestamp()

# With RRULE: prefix
next_run2 = compute_next_run('RRULE:FREQ=WEEKLY;BYDAY=MO', datetime.now(timezone.utc))
print(f'Weekly Monday: {next_run2}')
assert next_run2 is not None
print('OK')
"
```
✓ PASS: Computes future timestamps

**Phase Gate 3 - Scheduler Start/Stop:**
```bash
uv run python -c "
from server.scheduler import SchedulerEngine
from client.store import LocalStore
import tempfile, os, time

db = os.path.join(tempfile.gettempdir(), 'sched_test.db')
store = LocalStore(db)
store.init_db()

scheduler = SchedulerEngine(store)
scheduler.start()
print('Scheduler started')
time.sleep(2)
scheduler.stop()
print('Scheduler stopped')
print('OK')
"
```
✓ PASS: Starts and stops cleanly

**Phase Gate 4 - Due Automation Detection:**
```bash
uv run python -c "
from server.scheduler import SchedulerEngine, compute_next_run
from client.store import LocalStore
import tempfile, os, time

db = os.path.join(tempfile.gettempdir(), 'sched_test2.db')
store = LocalStore(db)
store.init_db()

# Create automation due NOW
past_time = int(time.time()) - 60  # 1 minute ago
aid = store.create_automation(
    name='Test Due',
    prompt='echo test',
    cwds='[\"/tmp\"]',
    rrule='FREQ=MINUTELY',
    next_run_at=past_time
)

due = store.get_due_automations(int(time.time()))
print(f'Found {len(due)} due automations')
assert len(due) == 1
assert due[0]['id'] == aid
print('OK')
"
```
✓ PASS: Finds due automation

**Phase Gate 5 - Full Execution Cycle (LIVE - costs money):**
```bash
uv run python -c "
from server.scheduler import SchedulerEngine
from client.store import LocalStore
import tempfile, os, time

db = os.path.join(tempfile.gettempdir(), 'sched_full_test.db')
store = LocalStore(db)
store.init_db()

# Create automation due now
past_time = int(time.time()) - 60
aid = store.create_automation(
    name='Full Test',
    prompt='What is 1+1? Output finding JSON with status none.',
    cwds='[\"/tmp\"]',
    rrule='FREQ=DAILY',
    next_run_at=past_time,
    model='haiku',
    max_budget_usd=0.05
)

scheduler = SchedulerEngine(store)
# Run one check cycle manually
scheduler._check_due_automations()

# Wait for execution
time.sleep(30)

# Check that run was created
runs = store.list_runs(automation_id=aid)
print(f'Found {len(runs)} runs')
assert len(runs) >= 1
print(f'Run status: {runs[0][\"status\"]}')
print('OK')
"
```
✓ PASS: Creates run record, executes automation

Pull Requests (comma-separated URLs)

Attachments

Comments (4)

qa-agent · 2026-02-03

## QA PASSED See QA report document: 9d99db04-0d1f-4386-bd40-873a8683e7d8 **Tests Verified:** - All 4 phase gates pass - 17 scheduler tests pass - 97 total tests pass (no regressions) - Edge cases tested: invalid RRULEs, concurrent protection, paused automations, scheduler lifecycle All hypotheses tested. No issues found. Fix verified working.

agent · 2026-02-03

## Implementation Complete ### Changes Made 1. Created server/scheduler.py - Implemented compute_next_run(rrule_str, after_dt) function for RRULE parsing - Implemented SchedulerEngine class with background thread polling - ThreadPoolExecutor for concurrent execution (4 workers) - Concurrent run protection via tracking set with locks 2. Updated client/store.py - Added next_run_at parameter to create_automation() - Fixed SQLite threading issue using threading.local() 3. Created tests/test_scheduler.py with 17 test cases ### Verification All phase gate tests pass (1-4). All 66 tests pass.

agent · 2026-02-03

Agent assignment: worker-AUTO-8 assigned to work on this issue. Worktree: /tmp/agent-automations-AUTO-8

triage-agent · 2026-02-03

## Feature Triage Complete **Status:** NOT IMPLEMENTED **Summary:** This is a greenfield project - no code exists yet. The SchedulerEngine needs to be built from scratch as part of issue AUTO-8. It will poll for due automations and execute them using a background thread with ThreadPoolExecutor. **Implementation Location:** `server/scheduler.py` (new file) **Dependencies:** Requires AUTO-1, AUTO-2, AUTO-4, AUTO-6, AUTO-7 to be implemented first **Complexity:** MEDIUM (~200-300 lines, well-defined scope) **Key Components:** - `compute_next_run(rrule_str, after_dt)` - RRULE parsing using python-dateutil - `SchedulerEngine` class with start/stop lifecycle - Background polling thread (60s interval) - ThreadPoolExecutor for concurrent automation execution - Integration with AutomationRunner, FindingDetector, TrckrIntegration, NtfyIntegration See attached triage report for full details including research findings and edge cases.