Documentation menu

Step library

A step is one entry in a workflow's step_configs. Each step names an activity — the unit of work Jetty knows how to run — plus the config that activity needs. For how steps chain together and pass data, see workflows & steps. This page is the catalog of activities you can reach for.

Activities fall into four families: AI models, control flow, data processing, and evaluation. The tables below list a representative slice of each, not the whole registry.

AI models

These call language, image, and video models. Each family needs the matching provider key configured; see runtimes and the connected-accounts idea for how Jetty resolves credentials at run time.

CategoryActivityWhat it does
AI modelslitellm_chatOne chat completion against any of 100+ providers via LiteLLM.
AI modelslitellm_batchMany completions in one step: fan out a list of requests.
AI modelsGemini stepsPrompt and reader activities plus gemini_image_generator.
AI modelsAWS BedrockInvoke models hosted on Bedrock.
AI modelsreplicate_text2imageReplicate models for text-to-image and the rest of the Replicate catalog.

Control flow

These shape how a workflow runs: branching, looping, fanning out across many inputs, and pulling data back out of prior runs.

CategoryActivityWhat it does
Control flowlist_emit_awaitFan out over a list and run children in parallel.
Control flowconditional_branchTake one path or another based on a value.
Control flownumber_sequence_generatorProduce a numeric range to iterate over.
Control flowloop_counterTrack iterations in a loop.
Control flowextract_from_trajectoriesPull values out of earlier trajectories into the current run.

Data processing

The glue steps: moving text and files around, attaching metadata, and reaching out over HTTP.

CategoryActivityWhat it does
Data processingtext_concatenateJoin text fragments into one string.
Data processingread / download / save file stepsRead, fetch, and persist files through Jetty storage.
Data processingmetadataAttach or read metadata on an asset.
Data processingwebhookCall out to an external HTTP endpoint mid-workflow.

Evaluation

These grade and analyze runs, the steps that drive trajectory-based evaluation.

CategoryActivityWhat it does
Evaluationsimple_judgeLLM-as-judge: score on a 1–5 scale or a categorical rubric.
Evaluationselect_trajectoriesFilter prior trajectories into a set to analyze.
Evaluationvisualize_correlationPlot how scores or signals relate across runs.

Discovering steps programmatically

This catalog is a snapshot, while the live registry is the source of truth. List the full set of step templates from your editor with list-step-templates (MCP) or listStepTemplates (SDK), then inspect any one to see its exact config shape.


Ready to wire these together? See examples & recipes for copy-paste workflow JSON.