xr_ai_tools.vision

VLM tools over single images, image collections, and timed video frames.

Classes

ImageQueryRequest	An image reference and the question to answer from it.
MultiImageQueryRequest	An ordered image collection and one question spanning the images.
VideoQueryRequest	Chronological image frames and a temporal question about them.
ImageQueryResult	A complete VLM answer and visual-input availability state.
ImageQueryChunk	One incremental text fragment from a streaming VLM answer.
ImageQueryTool	Answer a question about one caller-selected image.
MultiImageQueryTool	Answer one question across an ordered image collection.
VideoQueryTool	Answer one temporal question across timestamped image frames.
StreamingImageQueryTool	Stream an answer about one caller-selected image.

Module Contents

class xr_ai_tools.vision.ImageQueryRequest

An image reference and the question to answer from it.

image: xr_ai_tools.image.ImageReference

Image selected by another tool or caller.

query: str

Question to answer from the image.

class xr_ai_tools.vision.MultiImageQueryRequest

An ordered image collection and one question spanning the images.

images: list[xr_ai_tools.image.ImageReference]

Ordered images selected by other tools or the caller.

query: str

Question to answer from the images.

class xr_ai_tools.vision.VideoQueryRequest

Chronological image frames and a temporal question about them.

frames: list[xr_ai_tools.image.TimedImage]

Chronologically ordered image frames with Unix timestamps.

query: str

Question to answer from the timed frames.

validate_timeline() → VideoQueryRequest

Require frames to be supplied in chronological order.

class xr_ai_tools.vision.ImageQueryResult

A complete VLM answer and visual-input availability state.

text: str

Complete answer text.

available: bool

Whether the supplied visual input produced a usable answer.

class xr_ai_tools.vision.ImageQueryChunk

One incremental text fragment from a streaming VLM answer.

text: str

Partial fragment of the streamed answer text.

class xr_ai_tools.vision.ImageQueryTool(

*,

images: xr_ai_tools.image.ImageRegistry,

vlm: xr_ai_models.VLMService,

system_prompt: str = '',

)

Answer a question about one caller-selected image.

class xr_ai_tools.vision.MultiImageQueryTool(

*,

images: xr_ai_tools.image.ImageRegistry,

vlm: xr_ai_models.VLMService,

system_prompt: str = '',

)

Answer one question across an ordered image collection.

class xr_ai_tools.vision.VideoQueryTool(

*,

images: xr_ai_tools.image.ImageRegistry,

vlm: xr_ai_models.VLMService,

system_prompt: str = '',

)

Answer one temporal question across timestamped image frames.

class xr_ai_tools.vision.StreamingImageQueryTool(

*,

images: xr_ai_tools.image.ImageRegistry,

vlm: xr_ai_models.VLMService,

system_prompt: str = '',

)

Stream an answer about one caller-selected image.