Skip to main content
The JSX API describes workflows as trees of <Workflow>, <Task>, and control-flow components. Smithers renders the tree into an execution plan. 
/** @jsxImportSource smithers-orchestrator */
import { createSmithers, Sequence, Task } from "smithers-orchestrator";
import { z } from "zod";

const analysisSchema = z.object({
  summary: z.string(),
});

const { Workflow, smithers, outputs } = createSmithers({
  analysis: analysisSchema,
});

export default smithers((ctx) => (
  <Workflow name="analyze-repo">
    <Sequence>
      <Task id="analyze" output={outputs.analysis}>
        {{ summary: `Analyze ${ctx.input.repo}` }}
      </Task>
    </Sequence>
  </Workflow>
));
outputs maps each schema key to its Zod schema. output={outputs.analysis} gives compile-time type checking — a typo like output={outputs.anaylsis} is a type error.

​
Execution Model

  1. Render the workflow tree with the current ctx.
  2. Extract executable task descriptors.
  3. Run ready tasks and persist outputs.
  4. Re-render with updated outputs in ctx.
Task visibility uses JSX conditions or deps for upstream-output handoff: 
return (
  <Workflow name="code-review">
    <Task id="analyze" output={outputs.analysis}>...</Task>
    <Task id="fix" output={outputs.fix} deps={{ analyze: outputs.analysis }}>
      {(deps) => `Fix the issues from: ${deps.analyze.summary}`}
    </Task>
  </Workflow>
);

​
Schema-Driven JSX

Use createSmithers(...) with Zod schemas and the returned outputs object. The output prop is type-checked against registered schemas. Smithers manages backing SQLite tables. A <Task> can still write to a custom Drizzle table or use a string schema key as an escape hatch.

​
Next Steps