ReasoningEvaluateNode class

A media node that evaluates video frames against a ReasoningSpec. Subscribes to a video source, sends frames to the evaluation LLM, and emits tool calls and responses as events.

Signature:

export declare class ReasoningEvaluateNode extends AutoProcessorMediaNode<"video" | "audio" | "subtitle">

Extends: AutoProcessorMediaNode<"video" | "audio" | "subtitle">

Methods

Method	Modifiers	Description
enrolFace(name, image, instance)		Enrol a named face in the running faces tool’s gallery from an encoded image (JPEG/PNG). Subsequent detections of this person carry their identity. Enrolling an existing name replaces that entry. Resolves to the gallery names after the operation; rejects if the image can’t be used (undecodable, no confident face) — the node keeps running.
enrolVoice(name, wav, instance)		Enrol a named voice in the running speaker_id tool’s gallery from a 16kHz mono WAV (PCM16 or float32). Subsequent scored utterances matching this voice carry the name. Enrolling an existing name replaces that entry. Resolves to the gallery names after the operation; rejects if the WAV can’t be used — the node keeps running.
listIdentities(tool, instance)		List the names enrolled in the running faces / speaker_id tool’s gallery.
removeIdentity(tool, name, instance)		Remove a named identity from the running faces / speaker_id tool’s gallery. Subsequent detections of that person are anonymous again. Resolves to the remaining gallery names.
updateProvider(provider, contextDocuments)		Hot-swap the LLM provider while the node is running. The current analysis state is carried forward so the session continues seamlessly.
updateSpec(spec)		Hot-swap the ReasoningSpec while the node is running.

Method

Modifiers

Description

enrolFace(name, image, instance)

Enrol a named face in the running faces tool’s gallery from an encoded image (JPEG/PNG). Subsequent detections of this person carry their identity. Enrolling an existing name replaces that entry. Resolves to the gallery names after the operation; rejects if the image can’t be used (undecodable, no confident face) — the node keeps running.

enrolVoice(name, wav, instance)

Enrol a named voice in the running speaker_id tool’s gallery from a 16kHz mono WAV (PCM16 or float32). Subsequent scored utterances matching this voice carry the name. Enrolling an existing name replaces that entry. Resolves to the gallery names after the operation; rejects if the WAV can’t be used — the node keeps running.

listIdentities(tool, instance)

List the names enrolled in the running faces / speaker_id tool’s gallery.

removeIdentity(tool, name, instance)

Remove a named identity from the running faces / speaker_id tool’s gallery. Subsequent detections of that person are anonymous again. Resolves to the remaining gallery names.

updateProvider(provider, contextDocuments)

Hot-swap the LLM provider while the node is running. The current analysis state is carried forward so the session continues seamlessly.

updateSpec(spec)

Hot-swap the ReasoningSpec while the node is running.