Libraries

Atoms are grouped into four libraries. An engine binds one or more of them, and that decides which task names it recognises.

Library	Contains	Bound by
Common	Shell, files, network, data, secrets, containers	kis flow, AI Flow, Workflows
Infrastructure	Services, processes, ports, cron, users, gateways	kis flow
AI	Models, embeddings, vectors, retrieval, guardrails	AI Flow
Pipeline	Readers, writers, transforms over record streams	Datapipes, AI Flow (streaming forms)

A script sees the union of whatever is registered in its process. At the CLI that is effectively all four — which is why kis script can call an AI atom and a pipeline operation in the same function, and a flow generally cannot.

Which engine gets what

Engine	Common	Infrastructure	AI	Pipeline
kis flow / Automate	✓	✓
AI Flow	✓		✓	streaming forms
Workflows	✓
Datapipes				✓
kis script	✓	✓	✓	✓

Datapipes is the row to remember. It binds the pipeline library and nothing else, so there is no shell:, no ssh:, no git: in a data pipeline. See Flow engines for why that boundary exists.

Task names and namespace names

Within a library, one atom carries two names — and they are usually, but not always, the same word.

Operation	Flow task	Script namespace
Shell	shell:	shell.execute()
Object storage	s3:	s3.upload(), s3.download()
HTTP	http: or httpclient:	http.get(), http.post()
Templating	liquid: or template:	liquid.render()
Structured files	yaml:, yamlx:, json:	yaml.*
Containers	docker: or podman:	docker.*, podman.*
Kubernetes	k8s: or kubectl:	k8s.*, kubectl.*

Aliases exist where two communities call the same thing by different names. They resolve to the same atom — podman: is not a different container runtime, it is the same task under the name a podman user would look for.

Counting

121 task names are registered across the four libraries. Allowing for aliases, that is a little over a hundred distinct operations. The reference documents each one with both surfaces; these library pages are the index by group.

See also

• Atom reference — every atom, as a task and as a function
• Tasks and functions — how the two surfaces relate
• Flow engines — which engine to use