Module

pixano.inference.provider

Abstract inference provider interface. This module defines the abstract interface that all inference providers must implement. Pixano uses this interface to communicate with different backends (pixano-inference, OpenAI, Gemini, local models, etc.) in a uniform way. Only discovery methods are abstract; each task method defaults to raising `TaskNotSupportedError`, so a provider implements only what its backend supports.

Classes

class

InferenceProvider

(ABC)

Abstract interface for inference backends. Example: ```python from pixano.inference import get_provider provider = get_provider("pixano-inference", url="http://localhost:8000") models = await provider.list_models() result = await provider.image_mask_generation(input_data) ```

Attributes

Name	Type	Default	Description
`name`	`str`	—	Provider name (e.g., 'pixano-inference', 'openai', 'gemini').

Methods

 cancel_video_mask_generation_job 
(job_id, timeout)
  → VideoMaskGenerationJobStatus

Cancel an asynchronous video mask generation job.

Parameter	Type	Default
`job_id`	`str`	—
`timeout`	`float`	`30.0`

 close 
()
  → None

Release any resources held by the provider (e.g. HTTP connections).

 detection 
(input_data, timeout)
  → DetectionResult

Detect objects in an image using zero-shot detection.

Parameter	Type	Default
`input_data`	`DetectionInput`	—
`timeout`	`float`	`60.0`

 embedding 
(input_data, timeout)
  → EmbeddingResult

Embed an image or text into a shared vector space (CLIP-style).

Parameter	Type	Default
`input_data`	`EmbeddingInput`	—
`timeout`	`float`	`60.0`

 get_server_info 
()
  → ServerInfo

Get server information (version, models, models_to_task).

 get_video_mask_generation_job 
(job_id, timeout)
  → VideoMaskGenerationJobStatus

Fetch the current status of an asynchronous video mask generation job.

Parameter	Type	Default
`job_id`	`str`	—
`timeout`	`float`	`30.0`

 image_mask_generation 
(input_data, timeout)
  → ImageMaskGenerationResult

Generate masks for an image.

Parameter	Type	Default
`input_data`	`ImageMaskGenerationInput`	—
`timeout`	`float`	`60.0`

 list_models 
(task)
  → list[ModelInfo]

List available models, optionally filtered by task.

Parameter	Type	Default
`task`	`InferenceTask \| None`	`None`

 submit_video_mask_generation_job 
(input_data, timeout)
  → VideoMaskGenerationJobStatus

Submit an asynchronous video mask generation job.

Parameter	Type	Default
`input_data`	`VideoMaskGenerationInput`	—
`timeout`	`float`	`30.0`

 video_mask_generation 
(input_data, timeout)
  → VideoMaskGenerationResult

Generate masks for video frames (synchronous).

Parameter	Type	Default
`input_data`	`VideoMaskGenerationInput`	—
`timeout`	`float`	`120.0`

 vlm 
(input_data, timeout)
  → VLMResult

Generate text conditioned on images.

Parameter	Type	Default
`input_data`	`VLMInput`	—
`timeout`	`float`	`60.0`