Evenfire is an intelligence layer platform for running LLM agents that take real actions, on your infrastructure
TypeScript
import { describe, expect, it } from 'vitest'
import { collectWorkflowRecipeSecretRefs } from '../workflowRecipeSecretRefs'
describe('collectWorkflowRecipeSecretRefs', () => {
it('collects workload envSecret, snippet secretRef, and OAuth client refs', () => {
const refs = collectWorkflowRecipeSecretRefs({
workloads: [
{
id: 'api',
envSecret: {
name: 'workflow-api-credentials',
keys: [
{ secretKey: 'apiKey', envVar: 'API_KEY' },
{ secretKey: 'dbPassword', envVar: 'DB_PASSWORD' },
],
},
},
],
steps: [
{
id: 'snippet',
run: {
type: 'snippet',
capabilities: {
secrets: [{ secretRef: { name: 'snippet-creds', key: 'token' } }],
},
},
},
],
oauthClients: [
{
id: 'github',
clientIdRef: { name: 'github-oauth', key: 'clientId' },
clientSecretRef: { name: 'github-oauth', key: 'clientSecret' },
},
],
})
expect(
[...refs.values()].map(ref => [ref.namespace, ref.secretName, [...ref.keys].sort()])
).toEqual([
['sandbox-recipes', 'workflow-api-credentials', ['apiKey', 'dbPassword']],
['sandbox-recipes', 'snippet-creds', ['token']],
['sandbox-recipes', 'github-oauth', ['clientId', 'clientSecret']],
])
})
it('keeps transport and UI workload envSecret refs in their runtime namespaces', () => {
const refs = collectWorkflowRecipeSecretRefs({
ui: { workloadRef: 'frontend' },
workloads: [
{
id: 'transport',
transport: { type: 'streamableHttp', path: '/mcp' },
envSecret: {
name: 'transport-creds',
keys: [{ secretKey: 'apiKey', envVar: 'API_KEY' }],
},
},
{
id: 'frontend',
envSecret: {
name: 'ui-creds',
keys: [{ secretKey: 'token', envVar: 'TOKEN' }],
},
},
],
})
expect([...refs.values()].map(ref => [ref.namespace, ref.secretName, [...ref.keys]])).toEqual([
['mcp-server', 'transport-creds', ['apiKey']],
['sandbox-ui', 'ui-creds', ['token']],
])
})
it('does not invent an LLM Secret for a Codex-only agent, even if secretRef is leftover', () => {
const refs = collectWorkflowRecipeSecretRefs({
agent: {
provider: 'codex-subscription',
model: 'gpt-5.1',
secretRef: { name: 'codex-oauth', key: 'refreshToken' },
},
steps: [{ id: 'draft', instruction: 'Write a note' }],
})
expect([...refs.values()]).toEqual([])
})
it('collects an agent secretRef only when the agent uses static credentials', () => {
const refs = collectWorkflowRecipeSecretRefs({
agent: {
provider: 'openai',
model: 'gpt-5.4',
secretRef: { name: 'openai-llm', key: 'openai-api-key' },
},
})
expect([...refs.values()].map(ref => [ref.namespace, ref.secretName, [...ref.keys]])).toEqual([
['sandbox-recipes', 'openai-llm', ['openai-api-key']],
])
})
})