Catalog Functions

Ignite ships a public catalog of prebuilt functions you can invoke directly — no code to write, no deploy. They’re all compiled (tiny artifacts, fast cold starts), CPU-only, and MCP-enabled so agents can call them too.

Browse them with ListPublicCatalog, inspect one with GetPublicCatalogDetail (its input/output JSON Schema tells you the exact payload), and call them like any app — see Invocation. They live in the public org, so any authenticated caller can invoke them.

These are great composable building blocks for document/RAG pipelines: extract → chunk → transform a source, then analyze it.

Analysis

Text, data, and vision analysis — no ML model required (pure-Rust algorithms).

Function	What it does
`analyze-graph`	Graph/network analysis — PageRank, shortest path (Dijkstra), connected components, degree centrality. Input is an edge list.
`detect-anomalies`	Anomalies in time-series via statistical methods (z-score, IQR, MAD).
`detect-language`	Language detection with `whatlang` — 69 languages, trigram-based.
`extract-keywords`	Keywords / keyphrases via RAKE, TF-IDF, or TextRank. Sub-millisecond per document.
`track-objects-bytetrack`	Multi-object tracking across video frames (ByteTrack — Kalman filter + Hungarian algorithm).

Ingest — Chunk

Split a source into segments ready for embedding or downstream processing.

Function	What it does
`chunk-text`	Split text into smaller pieces while preserving semantic context.
`chunk-file`	Split documents by structure — PDF pages, DOCX paragraphs, ZIP/TAR entries, CSV/JSONL records, or byte windows.
`chunk-image`	Split images into tiles, or return the whole image with optional resizing.
`chunk-audio`	Segment audio by time and generate a spectrogram image per segment.
`chunk-video`	Extract keyframes via ffmpeg, returned as base64-encoded images.

Ingest — Extract

Pull text and media out of common document formats.

Function	What it does
`extract-pdf-text`	Extract text from PDFs (parallel page processing).
`extract-pdf-images`	Extract images from PDFs (parallel page processing).
`extract-docx`	Extract text from DOCX (parses `word/document.xml`).
`extract-xlsx`	Extract text from Excel `.xlsx` spreadsheets.
`extract-html`	Strip HTML tags and return plain text.
`extract-email`	Parse `.eml` files — headers, body text, attachments.

Ingest — Transform

Reshape, clean, redact, or secure data in flight.

Function	What it does
`normalize-text`	Unicode normalization, case conversion, whitespace trimming.
`redact-pii`	Redact PII patterns (SSN, email, phone) from text.
`redact-pii-enhanced`	36+ PII pattern types with redact / mask / hash modes (GDPR/HIPAA-oriented).
`encrypt-data`	AES-256-GCM encrypt/decrypt, SHA-256/512 hashing, HMAC sign/verify (via `ring`).
`geo-process`	Geospatial ops — Haversine/Vincenty distance, point-in-polygon, bounding box, centroid, area, simplify.
`resize-image`	Resize images with aspect-ratio preservation.
`crop-regions`	Crop rectangular regions from an image — consumes bounding boxes from detection/segmentation.

Example

Function	What it does
`hello-world`	Minimal example demonstrating the Ignite function platform — a good first thing to invoke.

Each catalog function declares its metadata (name, category, tags, input/output JSON Schema, mcp_enabled, resource tier) in a manifest, and is published to the public catalog. To build your own instead of using a prebuilt one, see Code & Runtimes and the recipes.