AI capabilities¶
The AI capabilities group provides platform-side model endpoints: text generation, speech recognition, object detection, audio sentiment classification, and image-to-video generation. Apart from video generation, which uses the asynchronous "create a task, then query its status" pattern, every endpoint returns its result in a single request.
These endpoint paths sit directly under /api/v5/ and do not require repository parameters such as owner or repo.
Objects¶
| Object | Description | Related endpoint |
|---|---|---|
| Text generation | Chat completions compatible with the OpenAI Chat Completions protocol | Text generation |
| Speech recognition | Audio to text; supports mp3, wav, m4a, flac, and webm | Automatic speech recognition |
| Object detection | Detect objects in an image with a YOLO model and return bounding boxes with class confidences | Object detection |
| Image-to-video generation | Asynchronous task: create the task first, then query its status using the task information | Image-to-video generation (create task) |
| Audio classification | Return an audio vector and scores for the three sentiment dimensions (arousal, dominance, valence) | Audio classification |
Identification¶
- No repository or organization identifier is needed; paths look like
POST /api/v5/chat/completions. - Video generation identifies progress with a "task": create it with
POST /api/v5/video/generateand query it withPOST /api/v5/video/status. - Authentication still uses the platform access token; see Authentication.
Common tasks¶
| Task | Endpoint |
|---|---|
| Text generation | POST /api/v5/chat/completions |
| Automatic speech recognition | POST /api/v5/audio/transcriptions |
| Object detection | POST /api/v5/detect/yolo |
| Image-to-video generation (create task) | POST /api/v5/video/generate |
| Image-to-video generation (query status) | POST /api/v5/video/status |
| Audio classification | POST /api/v5/audio/classification |
Notes¶
- Querying video generation status is also a
POST:POST /api/v5/video/status, notGET. This is the easiest mistake to make when editing a request out of habit. - Video generation is asynchronous:
generateonly creates the task and you must pollstatusfor the result. The other endpoints return synchronously. - Text generation is compatible with the OpenAI Chat Completions protocol, so you can use an OpenAI client directly and only need to change the base URL and the token.
- Speech recognition only accepts specific audio formats (mp3, wav, m4a, flac, webm); transcode any other format first.
Endpoint list¶
This module contains 6 endpoints in total.
| # | Endpoint | Method and path | Description |
|---|---|---|---|
| 1 | Text generation | POST /api/v5/chat/completions |
Generates text. The endpoint is based on the OpenAI-compatible Chat Completions protocol. With… |
| 2 | Automatic speech recognition | POST /api/v5/audio/transcriptions |
Provides automatic speech recognition: accepts an audio file (mp3, wav, m4a, flac, and webm are supported) and returns… |
| 3 | Object detection | POST /api/v5/detect/yolo |
Performs object detection with a YOLO model on an image and returns the detections, including bounding box coordinates, classes, and confidences… |
| 4 | Image-to-video generation (create task) | POST /api/v5/video/generate |
Call this endpoint to generate a video from an image (create the task). |
| 5 | Image-to-video generation (query status) | POST /api/v5/video/status |
Call this endpoint to generate a video from an image (query the status). |
| 6 | Audio classification | POST /api/v5/audio/classification |
Performs sentiment classification on an uploaded audio file and returns the audio vector representation, scores for the three sentiment dimensions (arousal, dominance, valence), and the formatted… |
Markdown source: View on AtomGit·Content from the open-source project gitcode-dev/atomgit-api-doc