Temple Compute Documentation

This site documents two products:

• Horus, an open-source workflow manager for HPC and the cloud. You describe a pipeline as a set of tasks, in YAML or Python, and Horus runs them in dependency order. It moves data between stages, sends each stage to the compute you point it at, and shows live progress in your terminal.
• Temple Compute OS, the workflow platform built on Horus. It adds a visual workflow editor, run and campaign management, and shared projects.

Horus

Get started

• Installation: get the horus command.
• Quickstart: write and run your first workflow.
• Core concepts: workflows, tasks, artifacts, and edges.
• Examples: runnable workflows you can download and adapt.

Writing workflows

• Workflows in YAML: the full schema, with a multi-task example.
• Workflows in Python: for tasks that need real logic or user prompts.
• Fan-out and fan-in and Loops: run a task once per item, or repeat a group of tasks.

Running and sharing

• Running workflows: the horus run command and the live dashboard.
• Packaging workflows: the horus package command, which zips a workflow and its inputs so it can run elsewhere.
• Sanitizing workflows: the horus sanitize command, which declares a workflow's implicit inputs so a UI can offer them.
• Handling secrets: keep credentials out of exported and packaged workflows.

Extending

• Extending Horus: add your own artifact types, tasks, runtimes, targets, and interactions as plugins.
• SDK reference: the Python API for extending the runtime.

Temple Compute OS

Get started

• Overview: the core ideas (workflows, drafts, runs, projects) and how to find your way around the app.

Workflow editor

• Designing workflows: build a workflow on the canvas in design mode, edit it as YAML, or build it with AI.
• Artifacts: workflow inputs, the artifacts panel, and searching files across runs.
• Custom input views: give an input its own editor, written as one HTML page that ships with the workflow.

Running at scale

• Runs: start a run, follow it live, read its logs and resource usage, and turn a past run back into a draft.
• Projects and campaigns: pin a finished version and launch batches of runs against it.

Sharing and extending

• Library and plugins: publish and reuse workflows, and add task and artifact kinds with plugin environments.

Reference

• Custom view protocol: the messages a custom view exchanges with the editor.