Scheduler Tools¶

The Scheduler Tools component provides LangChain tools for creating, pausing, resuming, stopping, and listing scheduled recurring jobs. This allows agents to set up automated workflow executions on intervals without human intervention.

Component type: scheduler_tools

How It Works¶

This component registers five tools with the parent agent. Scheduled jobs use Pipelit's self-rescheduling architecture -- each job runs as an RQ task that, on completion, enqueues itself for the next run via Queue.enqueue_in(). This approach avoids external cron dependencies and supports exponential backoff on failure.

Each scheduled job tracks:

• Run count -- how many times it has executed.
• Error count -- how many failures have occurred.
• Status -- active, paused, done, dead.
• Interval -- seconds between successful runs.

Ports¶

Outputs¶

Configuration¶

This tool has no configurable settings.

Usage¶

Connect this tool to an agent via the green diamond tool handle:

flowchart LR
    Sched[Scheduler Tools] -.->|tool| Agent
    SH[System Health] -.->|tool| Agent
    Model[AI Model] -.->|llm| Agent
    Trigger[Chat Trigger] --> Agent

Tools Provided¶

This component registers five tools with the agent:

create_schedule¶

Create a scheduled job that runs a workflow on a recurring interval.

create_schedule(
    name: str,
    workflow_id: int,
    interval_seconds: int,
    description: str = "",
    trigger_node_id: str | None = None,
    total_repeats: int = 0,
    max_retries: int = 3,
    timeout_seconds: int = 600,
) -> str

Returns: {"success": true, "schedule_id": "...", "name": "...", "status": "active"}

pause_schedule¶

Pause a running scheduled job. It stops rescheduling until resumed.

pause_schedule(schedule_id: str) -> str

Returns: {"success": true, "schedule_id": "...", "status": "paused"}

resume_schedule¶

Resume a paused scheduled job.

resume_schedule(schedule_id: str) -> str

Returns: {"success": true, "schedule_id": "...", "status": "active"}

stop_schedule¶

Permanently delete a scheduled job.

stop_schedule(schedule_id: str) -> str

Returns: {"success": true, "schedule_id": "...", "deleted": true}

list_schedules¶

List scheduled jobs with optional filters.

list_schedules(
    status: str | None = None,
    workflow_id: int | None = None,
    limit: int = 10,
) -> str

Returns: JSON with results list, each containing schedule_id, name, status, workflow_id, interval_seconds, run_count, and error_count.

Example¶

An agent setting up hourly monitoring:

User: Run the site-monitor workflow every hour.

Agent thinking: I need to find the workflow ID and create a schedule.

Tool call: list_schedules(workflow_id=12)
Result: {"success": true, "results": []}

Tool call: create_schedule(
    name="Hourly Site Monitor",
    workflow_id=12,
    interval_seconds=3600,
    description="Check site health every hour",
    max_retries=3
)
Result: {"success": true, "schedule_id": "sched-abc123", "name": "Hourly Site Monitor", "status": "active"}

Agent: Done! The site-monitor workflow will run every hour.
       Schedule ID: sched-abc123

Later, pausing the schedule:

User: Pause the monitoring, we're doing maintenance.

Tool call: pause_schedule(schedule_id="sched-abc123")
Result: {"success": true, "schedule_id": "sched-abc123", "status": "paused"}