Components

Delay by One

Delays a stream t by one tick: emits the previous-tick datum each tick, with a configured initial_value on the first tick. Type preserving, holding exactly one datum of state so live and delayed streams can be compared tick-to-tick.

Reidentify Objects (Torchreid)

@kaiyangzhou

Appearance re-identification over a video Image and a tracker's [Track] stream. Each track's box is cropped, OSNet-embedded, matched against a per-identity bank, and its volatile id is rewritten to a stable global id. The sink resolves the output arm.

Visualize

Generic overlay renderer. Accepts a base Image plus bounding boxes, segmentation masks, landmarks, points, polygons, vectors, or caption strings and stacks all overlays in input order, emitting the composited result as an Image.

Apply Image Morphology (OpenCV)

@opencv

Applies an OpenCV morphological operation to an Image with a chosen structuring element, emitting a downstream-resolved oneof t[Image, Mask] — the filtered frame in its input color space, or a re-binarised foreground Mask.

Classify Audio Event (HuggingFace)

@huggingface

Runs any HuggingFace audio-classification checkpoint on each AudioFrame; emits ranked [Classification], a top Classification, or the top label as String. Sigmoid (multi-label) vs softmax (single-label) is auto-picked per checkpoint.

Convert Point Cloud to Mesh (Open3D)

@intel

Reconstructs a triangle Mesh from an aligned PointCloud using Open3D Screened-Poisson or Ball-Pivoting reconstruction, estimating point normals when the input lacks them and cleaning degenerate / duplicate geometry before emitting.

Evaluate Expression

Generic sandboxed expression node: binds variadic inputs positionally as x0..xN plus per-replica state and message_count, runs the configured expression, and emits whatever it returns, pairing with branch or filter for boolean gating.

Detect Human Posture (RuleBased PPL)

Rule-based posture gate: consumes Landmarks2d.Human22 keypoints plus the frame width and height and emits Bool true when the geometric score for the chosen posture clears a threshold. No ML model — a CPU-only deterministic gate over a pose estimator.

Upscale Image (OSEDiff)

@cswry

Restores and super-resolves an Image with OSEDiff, a LoRA-finetuned Stable Diffusion 2.1 that maps a degraded input to a clean result in a SINGLE denoising step, emitting the enlarged Image at the requested factor.

Input JSON HTTP

Exposes a typed JSON ingress over HTTP POST or WebSocket, decoding each payload against the type the downstream port demands and emitting it as a t value. Use it for structured JSON ingest rather than raw-byte ingress.

Transcribe Audio (Parakeet)

@nvidia

NVIDIA Parakeet / Nemotron (NeMo) transcriber. Resamples each AudioFrame to 16 kHz and emits oneof t[String, [Subtitle], {text: String, language: String}] — plain text, timed segments, or text with the language it was recognised as. mode picks whole-frame transcription or cache-aware streaming, chunk_ms sets the streaming latency, and language conditions a multilingual checkpoint or asks it to detect the language itself.

Detect Image Anomaly (PaDiM)

@cealist

PaDiM anomaly detector: takes an Image and emits {score: Double, heatmap: Image} — an image-level anomaly score plus a per-pixel heatmap Image. Use when a class-specific feature distribution is fitted ahead of time and per-pixel localisation is needed.

Upscale Image (Real ESRGAN)

@xinntao

Upscales an Image with a Real-ESRGAN GAN generator (a photo or anime checkpoint chosen by model) and emits a super-resolved Image at outscale. Tuned for noisy or blurry input; tile_size bounds VRAM.

Detect Zone Transition

Consumes per-id [(UInt64, Bool)] in-zone flags, keeps a rolling history per tracked id, and emits a downstream-resolved oneof t[[UInt64], [Observation.Event]] — per-id transition codes, or enter/exit events.

Reconstruct Scene (Depth Anything 3)

@bytedance

Feed-forward 3D reconstruction from a single Image or a set of [Image] views with Depth Anything 3. Recovers per-view [MetricDepthMap], [CameraIntrinsics], [CameraPose], or a fused PointCloud, selected downstream.

Collect Speech Turn

@bleenco

Buffers an AudioFrame stream guided by VAD spans ([{start: Double, end: Double}]) and a turn-end Double probability, emitting one padded utterance AudioFrame per completed speech turn plus a Bool collecting flag.

Measure Image Sharpness (OpenCV)

@opencv

Scores each Image with a gradient-based sharpness measure against a cutoff, emitting a score plus pass flag (or a bare Double, Bool, or [Analytics.Metric] list per the downstream) so blurry frames can be dropped ahead of OCR. CPU-only.

Extract Text (Ollama)

@ollama

Sends each Image to a vision model served by a local Ollama runtime over its chat endpoint, then emits a joined String transcript or per-block regions with Polygon/Rectangle bounds and confidence.