MiniMax H3 Delete Task (Ace Data Cloud)

Every task you've ever submitted to the Ace Data Cloud MiniMax service sits in your account history, with its prompt, its inputs and its output URLs. This node clears them out - by one ID, or in bulk. It generates nothing, costs nothing, and is the least glamorous node in the pack, which is exactly why most people never wire it until their history is a scroll of dead UUIDs.

What it sends

The node posts to https://api.acedata.cloud/minimax/tasks with action: delete baked in - there's no action selector here, so its whole identity is deletion. The same service call backs the delete option on MiniMax H3 List Tasks; this is the version you keep at the end of a graph so you can't trip over it.

confirm is the guard, and it's enforced locally before any request leaves the machine: leave it off and you get "Set confirm to true to perform this deletion or archive operation." Here's the wrinkle - the node writes every widget value into the request payload, and a boolean false counts as set, so the toggle overrides parameters_json. Pasting {"confirm": true} into the JSON while the switch is off will never delete anything. You have to flip the switch itself.

id deletes one task. ids deletes several - and it forgives formatting: a JSON array works, and so does one ID per line in the widget, because the client splits plain text into a list for this field. limit, offset, created_at_min and created_at_max are the same windowing fields the lookup path uses, so a filtered bulk delete is possible. That combination is worth pausing on: a filter typo plus a flipped toggle is how you lose a batch of task records in one click. Use List Tasks to see exactly what your window matches first, then delete.

Outputs

The six outputs every node in this pack returns - status, task_id, media_urls_json, first_media_url, result_json, trace_id - are mostly ceremonial on a delete. Expect status: succeeded if the service accepted the request, empty URL fields, and result_json holding the scrubbed response. There's also an on-canvas text line reporting the status, task ID and URL count, which is genuinely useful here since a delete has no visible artifact. If the ID was already gone, you get an HTTP error instead of a friendly no-op.

A word on safety: deletions here are not read-only, so the client gives the request exactly one attempt. No automatic retry after a connection failure - which is correct behaviour for a destructive call, but it means a network hiccup leaves you unsure whether it happened. Re-run it if you're unsure; deleting an already-deleted task just errors, it doesn't delete something else.

Install

cd ComfyUI/custom_nodes && git clone https://github.com/AceDataCloud/MiniMaxH3ComfyUI
python -m pip install -r requirements.txt

Then restart ComfyUI with ACEDATACLOUD_API_KEY (or ACEDATACLOUD_API_TOKEN) set in the ComfyUI process environment. Manager route: search MiniMax H3, confirm publisher acedatacloud, registry ace-minimax-h3. Dependencies are requests and jsonschema; no models, no GPU, nothing large to download. Billing isn't a concern for this node - deleting a task doesn't refund credits and doesn't consume any - but it does destroy the record you'd need in order to prove a payment or debug a bad render.

Troubleshooting

• It refuses to run. Nine times out of ten that's the confirm toggle, not a bug. The error message is explicit; read it rather than adding more to parameters_json.
• Deleted a task, now the output URL is gone too. Keep the clip, not just the link. Anything you actually want long-term should be through MiniMax H3 Load Video and a save node, or downloaded to disk, before you tidy up.
• Bulk delete accidentally matched more than expected. Nothing in the node tells you what it's about to hit - no preview, no count. Check with List Tasks using the same filters first. If you're not sure, delete one ID at a time.
• HTTP error with a terse message. That's deliberate. This pack strips upstream payloads and internal fields from errors so credentials and request bodies can't leak into your workflow, your screenshots or an issue thread. Use the task ID or trace ID if you need the provider to look it up.