ebb-ai

Bulk-cancel every task in the queue that is still cancellable (status queued or scheduled). Running/completed/failed/cancelled tasks are left untouched. Optionally filter by status. Returns the number of tasks cancelled.

Cancel a queued or scheduled task. Idempotent — if the task is already completed/failed/cancelled this returns the existing status without error. Throws only if task_id is unknown. Cancelling one occurrence of a repeating task skips it and queues the next one; pass series=true to end the series.

Report on the ebb-ai task queue. With no arguments, returns a summary of all known tasks (including tasks persisted by previous sessions — the queue survives host restarts). With task_id, returns full detail for one task, including its title, prompt, result and any carbon receipt (estimated vs actual grams, grid_source, energy_source); viewing a completed or failed task marks it read. Never changes a task's status.

Dispatch a queued or scheduled provider-call task immediately, bypassing the scheduler's chosen carbon window. The resulting receipt records intensitySource='expedited'. Only provider-call tasks still queued/scheduled can be expedited; running, submitted (awaiting Batch API results), or terminal (completed/failed/cancelled) tasks are rejected with an explanatory error.

Return the projected electricity-grid carbon intensity for a region, hour by hour. Use this when deciding whether to run an expensive AI task now or defer it — intensity is reported in grams CO2-equivalent per kWh with a categorical band (very_clean / clean / average / dirty / very_dirty). Read-only; does not touch the task queue.

Return the optimal carbon-aware execution window for a task WITHOUT scheduling it. Use this for non-committal planning — to see when the cleanest grid moment falls inside the deadline (and the savings vs running now) before deciding whether to commit via schedule_task. Returns the chosen window, top alternatives, a batch_eligible flag, the grid_source (which feed produced the forecast — 'mock' means synthetic data), and a one-line reasoning string. Read-only; does not touch the task queue.

Continue a completed task's conversation: creates a NEW task that sends the earlier prompt, the model's reply and this text to the same provider and model, and runs at the next clean window before its deadline (default one hour). Costs one provider call. Only completed tasks accept a reply; failed or cancelled tasks are rejected. Returns the new task id and its scheduled time.

Re-dispatch a failed provider-call task. Only valid when the task's current status is 'failed' — queued/scheduled/running/submitted/completed tasks are rejected with an explanatory error. The new run's receipt overwrites the old.

Queue an AI task to run during off-peak hours inside the given deadline. Three parallel wins: (1) lighter on the electricity grid — deferring non-urgent tasks smooths projected AI compute demand; (2) 50% of sync list price (projected) — auto-routes through Anthropic/OpenAI Batch APIs when the deadline allows; (3) lower carbon — dispatch at the cleanest electricity-grid hour. Invoke this tool whenever the user phrases a request with deferral language: "do this later", "by tomorrow", "tonight", "overnight", "by EOD", "sometime this week", "when you have a moment", "remind me to", "queue this up", "schedule this", "no rush", "not urgent". Convert the user's wording to an ISO-8601 deadline and pass the task as the prompt. Returns a task_id you can later check with check_queue_status. Good fits: nightly digests, batch summaries, research sweeps, evaluator runs, multi-step report generation — anything the user is fine waiting on. Do NOT use for live chat, interactive code edits, or any task the user is actively waiting to see complete.

Re-score and reschedule a queued or scheduled task against a new deadline; the scheduler re-picks the cleanest window inside it. Throws if the task is already running or terminal, or if the new deadline is invalid or in the past.