Evenfire is an intelligence layer platform for running LLM agents that take real actions, on your infrastructure
TypeScript
import { describe, expect, it, vi } from 'vitest'
import type { PromptBridgeTarget } from '../domain/types'
import { PluginWorkloadSdkCredentialBrokerClient } from './credentialBrokerClient'
const target: PromptBridgeTarget = {
targetRef: 'primary-openai',
provider: 'openai',
model: 'gpt-5.4-mini',
credentialSlot: 'openai-api-key',
}
function response(status: number, body: unknown): Response {
return new Response(JSON.stringify(body), {
status,
headers: { 'Content-Type': 'application/json' },
})
}
describe('PluginWorkloadSdkCredentialBrokerClient', () => {
it('redeems one exact target ticket using the live runtime token', async () => {
const fetchImpl = vi.fn().mockResolvedValue(
response(200, {
provider: target.provider,
model: target.model,
credentialSlot: target.credentialSlot,
credentials: { 'openai-api-key': 'provider-secret' },
llmSecretName: 'chatllm-api-keys',
})
)
const client = new PluginWorkloadSdkCredentialBrokerClient({
baseUrl: 'http://wrc:8082/',
recipeNamespace: 'sandbox-recipes',
recipeName: 'recipe-a',
getAccessToken: () => 'runtime-jwt',
fetchImpl: fetchImpl as unknown as typeof fetch,
})
const result = await client.resolve({
invocationId: 'inv-1',
attemptGeneration: 1,
target,
credentialTicket: 'signed-ticket',
})
expect(result.keys.openai?.['openai-api-key']).toBe('provider-secret')
const [url, init] = fetchImpl.mock.calls[0] as [string, RequestInit]
expect(url).toBe('http://wrc:8082/api/v1/workflow/recipe-a/plugin-workload-sdk/credentials')
expect((init.headers as Record<string, string>).Authorization).toBe('Bearer runtime-jwt')
expect(JSON.parse(String(init.body))).toEqual({
recipeNamespace: 'sandbox-recipes',
invocationId: 'inv-1',
attemptGeneration: 1,
target,
credentialTicket: 'signed-ticket',
})
})
it.each([403, 503])('collapses broker %s into a terminal non-enumerable error', async status => {
const fetchImpl = vi
.fn()
.mockResolvedValue(response(status, { error: 'secret-name-that-must-not-leak' }))
const client = new PluginWorkloadSdkCredentialBrokerClient({
baseUrl: 'http://wrc:8082',
recipeNamespace: 'sandbox-recipes',
recipeName: 'recipe-a',
getAccessToken: () => 'runtime-jwt',
fetchImpl: fetchImpl as unknown as typeof fetch,
})
const error = await client
.resolve({
invocationId: 'inv-1',
attemptGeneration: 1,
target,
credentialTicket: 'signed-ticket',
})
.catch(caught => caught)
expect(error).toMatchObject({ code: 'provider_unavailable', retryable: false })
expect(String(error.message)).not.toContain('secret-name-that-must-not-leak')
})
it('rejects a response for a different target and never returns its credentials', async () => {
const fetchImpl = vi.fn().mockResolvedValue(
response(200, {
provider: 'claude',
model: target.model,
credentialSlot: target.credentialSlot,
credentials: { 'claude-api-key': 'wrong-secret' },
llmSecretName: 'wrong-secret-name',
})
)
const client = new PluginWorkloadSdkCredentialBrokerClient({
baseUrl: 'http://wrc:8082',
recipeNamespace: 'sandbox-recipes',
recipeName: 'recipe-a',
getAccessToken: () => 'runtime-jwt',
fetchImpl: fetchImpl as unknown as typeof fetch,
})
await expect(
client.resolve({
invocationId: 'inv-1',
attemptGeneration: 1,
target,
credentialTicket: 'signed-ticket',
})
).rejects.toMatchObject({ code: 'provider_unavailable', retryable: false })
})
})