Logs & Async Tasks

The platform stores request details, drawing records and async tasks such as video and music separately. After signing in, open Usage Logs or Task Logs from the left navigation. This page covers all three data sections, task status and how to query them through the API.

The three sections live at:

• /usage-logs/common: general API usage logs and billing details;
• /usage-logs/drawing: Midjourney drawing tasks;
• /usage-logs/task: general async tasks such as Suno and video.

Note

Logs only contain records from your own account. They are retained for a limited period, so export anything you need for long-term accounting.

Usage logs

Click Usage Logs in the left sidebar, or go to /usage-logs/common, to see API requests and billing details.

Filters

Filter	Description
Time range	Filters by record creation time. Both ends are included in the query range.
Model name	Filters by model name.
Group	Filters by the group the request used.
Log type	All types, top-up, consumption, management, system, error, refund, sign-in.
Token name	Filters by API key name.
Request ID	Locates one request by the platform request ID.
Upstream request ID	Locates a request by the ID the upstream service returned. Including this ID when you report a problem makes it much faster to investigate.

Page statistics

The values in the filter bar follow the current filters:

• Usage: the total quota of the consumption records in the current filter range, viewable by time, model, group and key.
• RPM: consumption requests in the last 60 seconds.
• TPM: total input and output tokens in the last 60 seconds.

List and details

The list has columns for time and type, token, model, stream, tokens, cost, latency and details.

• Tokens: shows input tokens / output tokens. When caching is used, cache read and write counts may also appear.
• Cost: shows the quota consumed or refunded for that record.
• Latency: shows the response time for consumption or error records that have timing data. Streaming requests may additionally show first response time (FRT) and throughput.
• Details: click a record to open the details dialog. Depending on the log type and the fields actually stored, it may include the request ID, model mapping, error or refund reason, token breakdown, billing breakdown, cache/audio/image usage and stream status — not every record shows every block.

Why a record may be missing

Log writes lag slightly. A request you just sent can take a few seconds to appear, so refresh before concluding it is missing.

Not every failed response produces an Error record: requests rejected before they reach a model (an invalid key, insufficient quota) usually produce no consumption log at all.

The Record IP address preference in your profile only controls whether the request IP is stored in consumption and error logs. It does not affect whether the record itself is created.

Task logs

Task Logs contains two subsections, Drawing logs and Task logs:

• Drawing logs: Midjourney drawing tasks.
• Task logs: async tasks such as text-to-video, image-to-video and audio generation.

Drawing logs

Click Task Logs → Drawing logs in the left sidebar, or go to /usage-logs/drawing, to see Midjourney drawing task records.

Drawing logs can be filtered by:

• Date range;
• Midjourney task ID.

The list includes submission time and status, action type, task ID, latency, submission result, progress, image, prompt and failure reason. Click an image to preview it, click a prompt to see it in full, and click a failure reason to see the complete error message.

Drawing statuses are: Not started, Queued, In progress, Success, Failure, Window waiting.

Note

Drawing status is updated by background polling, so it lags by tens of seconds. There is no need to refresh repeatedly. Tasks still unfinished after roughly an hour are marked as failed, and the pre-charged quota is refunded under the rules below.

Async task logs

Click Task Logs → Task logs in the left sidebar, or go to /usage-logs/task, to see async task logs.

Async endpoints for video, music and similar do not return a final result when the request is submitted. On successful submission the platform returns a public task ID and stores a task record, then a background job polls the upstream service and updates the status, progress and result.

Filtering tasks

Task logs can be filtered by:

• Date range;
• Task ID.

List and results

The task list shows submission and completion time, task ID, platform and action, latency, status, progress and details:

• When a Suno task succeeds and returns an audio address, the details column offers an audio preview;
• Some video tasks show a preview entry after they succeed. You can also request /v1/videos/{task_id}/content with the same account’s sign-in session or API key to fetch the video content of a completed task. That endpoint validates task ownership and returns nothing when the task is unfinished or belongs to another user;
• Other failed records show a failure reason in the details column, which opens the full message when clicked.

Task status and background processing

Async task logs use the following status values. Different platforms may spell a status slightly differently, so handle them leniently when reading through the API.

Status	Meaning
Not started	The task record exists but processing has not begun.
Queued	Submitted upstream, waiting for further processing.
In queue	Entered the processing queue.
In progress	Being processed.
Success	Finished; the result can be viewed or fetched.
Failure	Processing failed. The failure reason is written to the task record when the upstream provides one.
Unknown	The internally defined fallback for an unknown status.

Task status is updated by background polling, roughly every 15 seconds, so a freshly submitted task lags briefly. When polling for a result through the API, match that cadence instead of looping at high frequency. Tasks that never finish move to a failed terminal state after a timeout.

Failure, timeout and refunds

Once a task reaches a failed terminal state, the system usually tries to refund any quota that was pre-charged at submission, and writes a refund log entry on success. A successful refund clears the pending refund quota on the task record and reduces the accumulated usage. The request count is not rolled back.

The following can make the page status and the balance temporarily inconsistent:

• The task never had pre-charged quota (the cost shows as 0);
• The refund has not gone through yet, and the system is holding the quota for a retry or manual reconciliation. Keep the task ID and contact support in this case.

Caution

When investigating a refund, check the wallet balance or remaining subscription quota (and the quota of the API key in use) together with the refund records in the general usage log. Do not judge whether a refund completed from the Failure status alone.

API reference

Log query endpoints

To pull usage into your own accounting or monitoring system, use the read-only endpoints below. They authenticate with your sign-in session and always return results scoped to your own account.

Data	Endpoint	Time parameter unit
General logs	GET /api/log/self	seconds (created_at)
Log statistics	GET /api/log/self/stat	seconds
Midjourney drawing	GET /api/mj/self	milliseconds (submit_time)
Async tasks	GET /api/task/self	seconds (submit_time)

These endpoints use the pagination parameters p and page_size. Page numbers start at 1, and a page_size above 100 is truncated to 100.

In addition, GET /api/log/token reads a key’s recent logs using the API key itself, with no sign-in session — handy for a server-side usage check. It returns at most the 1000 most recent records, and works even if the key has expired or run out of quota.

Common query parameters for GET /api/task/self: p, page_size, task_id, start_timestamp, end_timestamp; it also accepts platform, action and status.

The task endpoints use Unix second timestamps for start_timestamp and end_timestamp. The Midjourney drawing endpoints use milliseconds.

Async generation and task endpoints

The endpoints below submit and query async video, music and drawing tasks. They all authenticate with an API key. The video content endpoint also accepts a sign-in session, and enforces task ownership so it only ever returns your own tasks.

Which platforms are available depends on the models currently offered — filter by model type in the Model Marketplace to check.

Endpoint	Path	Description
OpenAI video generation	POST /v1/videos	Creates an OpenAI-compatible async video generation task and returns a task identifier on success.
OpenAI video task query	GET /v1/videos/{task_id}	Returns the status, progress and result of a video generation task.
Video content retrieval	GET /v1/videos/{task_id}/content	Streams back the finished video content. The system fetches it upstream from the task result or decodes a stored data URL.
Video remix	POST /v1/videos/{video_id}/remix	Submits a new remix async task based on an existing video.
Legacy video generation	POST /v1/video/generations	The compatible legacy video generation path, which creates an async video task.
Legacy video task query	GET /v1/video/generations/{task_id}	Returns the status and result of a legacy video generation task.
Kling text-to-video	POST /kling/v1/videos/text2video	Submits a text-to-video task in the Kling format. The system converts the request format first.
Kling image-to-video	POST /kling/v1/videos/image2video	Submits an image-to-video task in the Kling format.
Kling task query	GET /kling/v1/videos/text2video/{task_id} GET /kling/v1/videos/image2video/{task_id}	Returns the result of a Kling text-to-video or image-to-video task.
Jimeng official endpoint	POST /jimeng/	Compatible with the official Jimeng API format, using parameters such as Action and Version to distinguish submission from query.
Suno task submission	POST /suno/submit/{action}	Submits a Suno music generation task. The action parameter determines the task type.
Suno task query	POST /suno/fetch GET /suno/fetch/{id}	Returns the Suno task list, or the status and result of one task.
Midjourney task submission	POST /mj/submit/{action} POST /{mode}/mj/submit/{action}	Supports actions such as imagine, describe, blend, change, simple-change, edits, video, action, modal and shorten.
Midjourney task query	GET /mj/task/{id}/fetch GET /mj/task/{id}/image-seed POST /mj/task/list-by-condition	Returns the task status, image seed, or a task list by condition. Paths with a {mode} prefix are also supported.
Midjourney image proxy	GET /mj/image/{id} GET /{mode}/mj/image/{id}	Fetches or proxies the generated image by Midjourney task ID.
Midjourney extended actions	POST /mj/insight-face/swap POST /mj/submit/upload-discord-images	Extended features such as face swap and uploading Discord images.

Relationship to the dashboard

Logs are a record-by-record breakdown, while the dashboard aggregates by time, model and group. They refresh at different times: a new request appears in the logs first, and chart data only updates after aggregation, so a short-lived mismatch between the two is normal.

When investigating a call, start in the usage log and locate the request by time, model, request ID or upstream request ID. For an async request, note the task_id the submission endpoint returned and query the final task status and failure details. When you need support, quoting the request ID or task ID is the fastest route.