Operator catalog — by family

Glyph ships 78 operators across 24 families. Each family page lists its operators with config schemas and input/output ports. For a single visual grid of everything, see the operator catalog.

Families

Family	Operators	What it does
Agent	1	LLM-driven agents that call tools in a loop until a goal is reached.
Cache	1	Redis-backed key/value cache scoped to the run’s document.
Code	1	Sandboxed user code execution.
Connector	4	Outbound calls to third-party services (auth-gated by secret refs).
Data	3	Pure-function data shaping: parse, transform, list ops.
Database	2	Read/write the tenant Postgres schema.
Document	4	Read/write the current document’s metadata and CRDT subdocs.
Eval	4	Grade outputs against assertions, regexes, and judge prompts.
Guardrails	4	Input/output filtering and cost capping.
HTTP	1	Outbound HTTP, SSRF-guarded.
Human	2	Block the run until a human approves or supplies input.
Input	5	Source values into the graph (text, file, secret, sample data).
Knowledge	4	Vector-backed semantic search over per-document collections.
Logic	7	Control flow: if/else, switch, loops, joins, waits.
Media	5	Image / audio / video generation, transcription, OCR.
Model	5	Direct LLM calls — responses, classification, extraction, embeddings, reranking.
Output	5	Terminal nodes: return, persist, notify, export, preview.
Prompt	2	Compose prompt messages from templates and runtime variables.
Storage	1	S3-backed object storage scoped to the run’s document.
Tools	3	Surface external tools (MCP, function-call style).
Trigger	7	Event sources that start a run.
Vars	1	Per-run scratchpad — set/get keys across the graph.
Web	3	HTTP fetch + article scraping.
Workflow	3	Invoke other workflows and return values to a caller.