Skip to content

Latest commit

 

History

History
69 lines (48 loc) · 2.79 KB

File metadata and controls

69 lines (48 loc) · 2.79 KB

Interface Information

Following are the details of the interfaces provided by the Perception Capabilities package. A runner that intends to connect with one of these interfaces should support the input and ouput parameters described below.

ImageAnalysisRunner

Input Parameters:

Parameter Name Type Description
use_device bool Whether to use the device's vision capabilities.
prompt string The prompt with instructions on how to process the image using the LLM.

Output Parameters:

Parameter Name Type Description
text string The analysis result generated by the LLM for the given image.

SpeechRunner

Input Parameters:

Parameter Name Type Description
use_device bool Whether to use the device's audio capabilities.
text string The text to be synthesized into speech.
voice string Optional voice identifier passed through to the speech backend.
instructions string Optional synthesis instructions passed through to the speech backend.

Output Parameters:

Parameter Name Type Description
success bool Whether the speech synthesis request completed successfully.

TranscribeRunner

Input Parameters:

Parameter Name Type Description
use_device bool Whether to use the device's audio capabilities.
audio_request_window int The duration of the requested device-audio window in seconds.

Output Parameters:

Parameter Name Type Description
text string The transcription result generated for the given audio.
success bool Whether the transcription request completed successfully.

SentimentRunner

Input Parameters:

Parameter Name Type Description
text string The text to be analyzed for sentiment.
use_device bool Whether to capture device audio for analysis instead of relying only on the provided text.
audio_request_window int Requested device-audio window in seconds when use_device is enabled.

Output Parameters:

Parameter Name Type Description
text string The sentiment label returned for the analyzed input.
score double Confidence score associated with the returned sentiment label.
analyzed_text string The text that was actually analyzed by the sentiment service.