Register searchOne with retries — Stock research: from flaky to reliable

Wrap searchOne in task(). Keep maybeFail, getExa, and the Exa call body unchanged: the workshop keeps the same transient error simulation for search. Retry logic handles most transient failures.

Import task

At the top of tasks/src/search.ts:

import { task } from '@renderinc/sdk/workflows'

Replace the export

Before

- export async function searchOne(
- _topic: string,
- spec: SearchSpec,
- index: number,
 
 
 
 
 
 
 
 
 
 
 
  ): Promise<SearchResult> {
- maybeFail(spec.query)
- // ... Exa call ...
- }

After

 
 
 
 
+ export const searchOne = task(
+ {
+   name: 'searchOne',
+   plan: 'starter',
+   timeoutSeconds: 120,
+   retry: { maxRetries: 3, waitDurationMs: 1000, backoffScaling: 1.5 },
+ },
+ async function searchOne(
+   _topic: string,
+   spec: SearchSpec,
+   index: number,
  ): Promise<SearchResult> {
 
 
 
+   maybeFail(spec.query)
+   // ... Exa call unchanged ...
+ },
+ )

Field	Why
`name: 'searchOne'`	Becomes `{service-slug}/searchOne` in `startTask`
`plan: 'starter'`	Light Exa call
`timeoutSeconds: 120`	Room for Exa latency
`retry.maxRetries: 3`	Up to four attempts when `maybeFail` throws

Close the task() call at the end of the function with }, ).

Build

Terminal

$cd tasks && npm run build
Build succeeded

What you learned

searchOne becomes export const searchOne = task({ ... }, async function ...)
maybeFail stays unchanged so retries have something to recover from
Build tasks/ before moving on