Komponenten

Classify Image (Zero Shot SigLIP)

@google

Scores each Image against free-form text candidate_labels with a SigLIP/CLIP model, builds one ranked result bounded by threshold and top_k, and projects it onto the connected output shape (top class, ranked list, label, pairs, or class id).

Input Audio HTTP

Exposes a public audio ingress that accepts client-pushed uploads over HTTP POST or WebSocket, decodes them, and emits AudioFrame chunks paired with a per-upload session-id String. Use it when a deployment needs to take audio from outside the graph.

Output Audio HTTP

Encodes each AudioFrame and ships it live over WebSocket and/or async HTTP POST, with raw PCM modes emitting headerless samples and container modes emitting one self-contained file per frame, so a backend can stream audio to browsers or an external recorder.

Estimate Camera Poses (COLMAP)

@colmap

Structure-from-motion over an overlapping [Image] batch of one static scene, emitting a deploy-time-resolved oneof: per-registered-image [CameraPose] extrinsics, per-image [CameraIntrinsics], or the sparse coloured PointCloud of the reconstruction.

Input XLSX File

Reads an Excel worksheet from each workbook in the bound directory and emits one Analytics.TabularDataset per file, choosing the sheet by name or the active one. Use it to pull spreadsheet data into a tabular pipeline.

Unwarp Document (DvD)

@hanquansanren

Rectifies a curved, folded, or crumpled document Image with the DvD coordinate-diffusion model, which runs an iterative denoising loop to predict a backward warp and emits the flattened Image for OCR.

Extract Text (RapidOCR)

@rapidai

Runs fast ONNX-based RapidOCR (PP-OCRv5) detection and recognition over each Image, emitting either a joined String transcript or per-line regions with Polygon or Rectangle bounds and optional confidence.

Classify Speech Emotion (emotion2vec)

@funaudiollm

Classifies speech affect in each AudioFrame with an emotion2vec model, emitting ranked [Classification] pairs or a single top-label String chosen by the connected output slot. Audio is resampled to 16 kHz internally for language-robust affect labelling.

Classify Audio Zero Shot (HuggingFace)

@huggingface

Zero-shot audio classifier: self-downloads a HuggingFace CLAP checkpoint and ranks a runtime label set against each AudioFrame by audio-text similarity. Emits [Classification], a top Classification, or raw [Double] scores. Model hot-swaps.

Equalize Histogram (OpenCV)

@opencv

Applies CLAHE or global histogram equalization to each Image on a CPU, enhancing the luma channel in a chosen working space (or the whole image in grayscale) and emitting a contrast-enhanced Image.

Input PDF File

Reads each PDF in the bound directory and emits a oneof t[String, Image, Document, Document.Page] chosen by what it is wired to: a text consumer receives one String per document (the extracted text), an image consumer receives one Image per page (the rasterized page), a Document consumer receives one fully-structured document per file, a Document.Page consumer receives one structured page streamed as it renders. Use it to feed native PDFs into language backends, to rasterize pages for OCR when the source has no text layer, or to take a page apart into the blocks and typography it was set with.

Input Image File

Decodes images from the bound directory and emits each as a BGR Image at a paced rate, optionally resizing and looping. Use it for offline batches or deterministic image test sets.

Restore Face (CodeFormer)

@sczhou

CodeFormer blind face restoration on an Image. Faces are detected, restored with a transformer codebook prior and pasted into the upscaled frame; the sink picks the arm: restored Image, face count, or per-face crops.

Estimate Metric Depth (UniDepth V2)

@lpiccinelli

Intrinsics-free metric monocular depth (UniDepth V2): one forward over an Image jointly recovers metric depth and camera intrinsics, emitting a deploy-time-resolved MetricDepthMap, CameraIntrinsics, or a camera-frame PointCloud.

Output Image HTTP

Encodes each Image as PNG or JPEG and ships it over any subset of three live egress modes — HTTP POST/PUT, binary WebSocket broadcast, or an MJPEG stream — with optional per-frame metadata carried as headers or a paired WebSocket frame.

Filter

Gates a value stream t by a paired Bool: forwards the value that tick when the Bool is true, drops it when false. The output type is exactly t; the Bool decides pass-through only, never the type, and never appears on the output.

Extract Text Embedding (HuggingFace)

@huggingface

Encodes each String into a dense Embedding by loading any HuggingFace encoder directly in-process with AutoModel and AutoTokenizer, pooled by CLS or mean and optionally L2-normalised. No cloud API. Suited to retrieval, clustering, similarity, and RAG encoding.