0d3cb498a3
CI / Shell Format Check (push) Has been cancelled
CI / Check Ruby (3.4) (push) Has been cancelled
CI / CI Config (push) Has been cancelled
CI / Test on Node ${{ matrix.node }} and ${{ matrix.os }}${{ matrix.shard && format(' (shard {0}/3)', matrix.shard) || '' }} (push) Has been cancelled
CI / Build on Node ${{ matrix.node }} (push) Has been cancelled
CI / Style Check (push) Has been cancelled
CI / Generate Assets (push) Has been cancelled
CI / Check Python (3.14) (push) Has been cancelled
CI / Check Python (3.9) (push) Has been cancelled
CI / Build Docs (push) Has been cancelled
CI / Code Scan Action (push) Has been cancelled
CI / Site tests (push) Has been cancelled
CI / webui tests (push) Has been cancelled
CI / Run Integration Tests (push) Has been cancelled
CI / Run Smoke Tests (push) Has been cancelled
CI / Go Tests (push) Has been cancelled
CI / Share Test (push) Has been cancelled
CI / Redteam (Production API) (push) Has been cancelled
CI / Redteam (Staging API) (push) Has been cancelled
CI / GitHub Actions Lint (push) Has been cancelled
CI / Check Ruby (3.0) (push) Has been cancelled
release-please / release-please (push) Has been cancelled
release-please / build (push) Has been cancelled
release-please / publish-npm (push) Has been cancelled
release-please / publish-npm-backfill (push) Has been cancelled
release-please / docker (push) Has been cancelled
release-please / publish-code-scan-action (push) Has been cancelled
release-please / attest-code-scan-action (push) Has been cancelled
Deploy local.promptfoo.app / Deploy to Cloudflare Pages (push) Has been cancelled
Test and Publish Multi-arch Docker Image / test (push) Has been cancelled
Test and Publish Multi-arch Docker Image / build-docker-and-push-digests (map[digest-suffix:linux-amd64 platform:linux/amd64 runner:ubuntu-latest]) (push) Has been cancelled
Test and Publish Multi-arch Docker Image / build-docker-and-push-digests (map[digest-suffix:linux-arm64 platform:linux/arm64 runner:ubuntu-24.04-arm]) (push) Has been cancelled
Test and Publish Multi-arch Docker Image / merge-docker-digests (push) Has been cancelled
Test and Publish Multi-arch Docker Image / Attest Multi-arch Image (push) Has been cancelled
Validate Renovate Config / Validate Renovate Configuration (push) Has been cancelled
327 lines
9.9 KiB
TypeScript
327 lines
9.9 KiB
TypeScript
import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest';
|
|
import { sanitizeProvider } from '../../../src/models/evalResult';
|
|
import { neverGenerateRemote } from '../../../src/redteam/remoteGeneration';
|
|
import {
|
|
createMockProvider,
|
|
createProviderResponse,
|
|
type MockApiProvider,
|
|
} from '../../factories/provider';
|
|
|
|
import type { ApiProvider, CallApiContextParams } from '../../../src/types/index';
|
|
|
|
const mockFetchWithProxy = vi.fn();
|
|
const mockRenderPrompt = vi.fn();
|
|
|
|
vi.mock('../../../src/util/fetch/index', () => ({
|
|
fetchWithProxy: (...args: unknown[]) => mockFetchWithProxy(...args),
|
|
}));
|
|
|
|
vi.mock('../../../src/evaluatorHelpers', () => ({
|
|
renderPrompt: (...args: unknown[]) => mockRenderPrompt(...args),
|
|
}));
|
|
|
|
vi.mock('../../../src/globalConfig/accounts', () => ({
|
|
getUserEmail: vi.fn().mockReturnValue('test@example.com'),
|
|
}));
|
|
|
|
vi.mock('../../../src/redteam/remoteGeneration', () => ({
|
|
getRemoteGenerationExplicitlyDisabledError: vi.fn(
|
|
(strategyName) =>
|
|
`${strategyName} requires remote generation, which has been explicitly disabled.`,
|
|
),
|
|
getRemoteGenerationHeaders: vi.fn((extra) => ({ 'Content-Type': 'application/json', ...extra })),
|
|
getRemoteGenerationUrl: vi.fn().mockReturnValue('http://test.api/generate'),
|
|
neverGenerateRemote: vi.fn().mockReturnValue(false),
|
|
}));
|
|
|
|
describe('BestOfNProvider - Runtime Behavior', () => {
|
|
let BestOfNProvider: typeof import('../../../src/redteam/providers/bestOfN').default;
|
|
let mockTargetProvider: MockApiProvider;
|
|
|
|
const createMockContext = (targetProvider: ApiProvider): CallApiContextParams => ({
|
|
originalProvider: targetProvider,
|
|
vars: { input: 'test input' },
|
|
prompt: { raw: 'test prompt', label: 'test' },
|
|
});
|
|
|
|
beforeEach(async () => {
|
|
vi.clearAllMocks();
|
|
vi.mocked(neverGenerateRemote).mockReset();
|
|
vi.mocked(neverGenerateRemote).mockReturnValue(false);
|
|
mockRenderPrompt.mockReset();
|
|
mockRenderPrompt.mockImplementation((_prompt: unknown, vars: unknown) => {
|
|
const input =
|
|
typeof vars === 'object' &&
|
|
vars !== null &&
|
|
'input' in vars &&
|
|
typeof (vars as { input?: unknown }).input === 'string'
|
|
? (vars as { input: string }).input
|
|
: undefined;
|
|
return input || 'rendered prompt';
|
|
});
|
|
|
|
// Dynamic import after mocks are set up
|
|
const module = await import('../../../src/redteam/providers/bestOfN');
|
|
BestOfNProvider = module.default;
|
|
|
|
mockTargetProvider = createMockProvider({
|
|
response: createProviderResponse({ output: 'target response' }),
|
|
});
|
|
|
|
// Mock successful response from remote API
|
|
mockFetchWithProxy.mockResolvedValue({
|
|
json: async () => ({
|
|
modifiedPrompts: ['candidate 1', 'candidate 2'],
|
|
}),
|
|
});
|
|
});
|
|
|
|
afterEach(() => {
|
|
vi.clearAllMocks();
|
|
});
|
|
|
|
it('should pass abortSignal to fetchWithProxy', async () => {
|
|
const provider = new BestOfNProvider({
|
|
injectVar: 'input',
|
|
});
|
|
|
|
const abortController = new AbortController();
|
|
const context = createMockContext(mockTargetProvider);
|
|
|
|
await provider.callApi('test prompt', context, { abortSignal: abortController.signal });
|
|
|
|
expect(mockFetchWithProxy).toHaveBeenCalledWith(
|
|
expect.any(String),
|
|
expect.any(Object),
|
|
abortController.signal,
|
|
);
|
|
});
|
|
|
|
it('should include target context in remote generation requests', async () => {
|
|
const provider = new BestOfNProvider({
|
|
injectVar: 'input',
|
|
targetId: 'cloud-target-123',
|
|
});
|
|
|
|
await provider.callApi('test prompt', createMockContext(mockTargetProvider));
|
|
|
|
const request = mockFetchWithProxy.mock.calls[0]?.[1] as { body?: string } | undefined;
|
|
expect(request?.body).toBeDefined();
|
|
expect(JSON.parse(request?.body ?? '{}')).toMatchObject({
|
|
targetId: 'cloud-target-123',
|
|
task: 'jailbreak:best-of-n',
|
|
});
|
|
});
|
|
|
|
it('should pass options to target provider callApi', async () => {
|
|
const provider = new BestOfNProvider({
|
|
injectVar: 'input',
|
|
});
|
|
|
|
const abortController = new AbortController();
|
|
const context = createMockContext(mockTargetProvider);
|
|
const options = { abortSignal: abortController.signal };
|
|
|
|
await provider.callApi('test prompt', context, options);
|
|
|
|
// The target provider should be called with the options
|
|
expect(mockTargetProvider.callApi).toHaveBeenCalledWith(
|
|
expect.any(String),
|
|
expect.any(Object),
|
|
options,
|
|
);
|
|
});
|
|
|
|
it('should re-throw AbortError and not swallow it', async () => {
|
|
const provider = new BestOfNProvider({
|
|
injectVar: 'input',
|
|
});
|
|
|
|
const abortError = new Error('The operation was aborted');
|
|
abortError.name = 'AbortError';
|
|
mockFetchWithProxy.mockRejectedValue(abortError);
|
|
|
|
const context = createMockContext(mockTargetProvider);
|
|
|
|
await expect(provider.callApi('test prompt', context)).rejects.toThrow(
|
|
'The operation was aborted',
|
|
);
|
|
});
|
|
|
|
it('should swallow non-AbortError exceptions and return error response', async () => {
|
|
const provider = new BestOfNProvider({
|
|
injectVar: 'input',
|
|
});
|
|
|
|
const regularError = new Error('Network error');
|
|
mockFetchWithProxy.mockRejectedValue(regularError);
|
|
|
|
const context = createMockContext(mockTargetProvider);
|
|
|
|
const result = await provider.callApi('test prompt', context);
|
|
|
|
// Non-AbortError should be caught and returned as an error response
|
|
expect(result.error).toContain('Network error');
|
|
});
|
|
|
|
it.each([
|
|
42,
|
|
true,
|
|
null,
|
|
{ prompt: 'candidate 0' },
|
|
])('should skip non-string candidate prompt from remote generation: %j', async (invalidPrompt) => {
|
|
const provider = new BestOfNProvider({
|
|
injectVar: 'input',
|
|
});
|
|
const context = createMockContext(mockTargetProvider);
|
|
|
|
mockFetchWithProxy.mockResolvedValue({
|
|
json: async () => ({
|
|
modifiedPrompts: [invalidPrompt, 'candidate 2'],
|
|
}),
|
|
});
|
|
|
|
await provider.callApi('test prompt', context);
|
|
|
|
expect(mockRenderPrompt).toHaveBeenCalledTimes(1);
|
|
expect(mockTargetProvider.callApi).toHaveBeenCalledTimes(1);
|
|
expect(mockTargetProvider.callApi).toHaveBeenCalledWith(
|
|
'candidate 2',
|
|
expect.any(Object),
|
|
undefined,
|
|
);
|
|
});
|
|
|
|
it.each([
|
|
'file://etc/passwd',
|
|
' FILE://etc/passwd',
|
|
'\tFiLe://etc/passwd',
|
|
'package:@promptfoo/fake:getSecret',
|
|
' PACKAGE:@promptfoo/fake:getSecret',
|
|
'\tPaCkAgE:@promptfoo/fake:getSecret',
|
|
])('should skip unsafe candidate prompt from remote generation: %s', async (unsafePrompt) => {
|
|
const provider = new BestOfNProvider({
|
|
injectVar: 'input',
|
|
});
|
|
const context = createMockContext(mockTargetProvider);
|
|
|
|
mockFetchWithProxy.mockResolvedValue({
|
|
json: async () => ({
|
|
modifiedPrompts: [unsafePrompt, 'candidate 2'],
|
|
}),
|
|
});
|
|
|
|
await provider.callApi('test prompt', context);
|
|
|
|
expect(mockTargetProvider.callApi).toHaveBeenCalledTimes(1);
|
|
expect(mockTargetProvider.callApi).toHaveBeenCalledWith(
|
|
'candidate 2',
|
|
expect.any(Object),
|
|
undefined,
|
|
);
|
|
});
|
|
|
|
it('should pass the injected variable through renderPrompt without special loading or template rendering', async () => {
|
|
const provider = new BestOfNProvider({
|
|
injectVar: 'input',
|
|
});
|
|
const context = createMockContext(mockTargetProvider);
|
|
|
|
await provider.callApi('test prompt', context);
|
|
|
|
expect(mockRenderPrompt).toHaveBeenCalledWith(
|
|
context.prompt,
|
|
{
|
|
...context.vars,
|
|
input: 'candidate 1',
|
|
},
|
|
context.filters,
|
|
mockTargetProvider,
|
|
['input'],
|
|
);
|
|
});
|
|
});
|
|
|
|
describe('BestOfNProvider - Config Serialization', () => {
|
|
let BestOfNProvider: typeof import('../../../src/redteam/providers/bestOfN').default;
|
|
|
|
beforeEach(async () => {
|
|
vi.clearAllMocks();
|
|
vi.mocked(neverGenerateRemote).mockReset();
|
|
vi.mocked(neverGenerateRemote).mockReturnValue(false);
|
|
const module = await import('../../../src/redteam/providers/bestOfN');
|
|
BestOfNProvider = module.default;
|
|
});
|
|
|
|
afterEach(() => {
|
|
vi.clearAllMocks();
|
|
});
|
|
|
|
it('should expose config property with all constructor options', () => {
|
|
const options = {
|
|
injectVar: 'query',
|
|
maxConcurrency: 5,
|
|
nSteps: 10,
|
|
maxCandidatesPerStep: 20,
|
|
};
|
|
|
|
const provider = new BestOfNProvider(options);
|
|
|
|
expect(provider.config).toEqual({
|
|
injectVar: 'query',
|
|
maxConcurrency: 5,
|
|
nSteps: 10,
|
|
maxCandidatesPerStep: 20,
|
|
});
|
|
});
|
|
|
|
it('should apply default maxConcurrency when not provided', () => {
|
|
const provider = new BestOfNProvider({
|
|
injectVar: 'query',
|
|
});
|
|
|
|
expect(provider.config.maxConcurrency).toBe(3);
|
|
});
|
|
|
|
it('should throw an actionable error when remote generation is explicitly disabled', () => {
|
|
vi.mocked(neverGenerateRemote).mockReturnValue(true);
|
|
|
|
expect(() => new BestOfNProvider({ injectVar: 'query' })).toThrow(
|
|
'Best-of-N strategy requires remote generation, which has been explicitly disabled.',
|
|
);
|
|
});
|
|
|
|
it('should preserve config through sanitizeProvider for database storage', () => {
|
|
const provider = new BestOfNProvider({
|
|
injectVar: 'query',
|
|
maxConcurrency: 5,
|
|
nSteps: 10,
|
|
});
|
|
|
|
const sanitized = sanitizeProvider(provider);
|
|
|
|
expect(sanitized.id).toBe('promptfoo:redteam:best-of-n');
|
|
expect(sanitized.config).toEqual({
|
|
injectVar: 'query',
|
|
maxConcurrency: 5,
|
|
nSteps: 10,
|
|
maxCandidatesPerStep: undefined,
|
|
});
|
|
});
|
|
|
|
it('should allow recreating provider from sanitized config', () => {
|
|
const originalProvider = new BestOfNProvider({
|
|
injectVar: 'query',
|
|
maxConcurrency: 5,
|
|
nSteps: 10,
|
|
});
|
|
|
|
// Simulate what happens when saved to database and loaded via retry strategy
|
|
const sanitized = sanitizeProvider(originalProvider);
|
|
const recreatedProvider = new BestOfNProvider(sanitized.config);
|
|
|
|
expect(recreatedProvider.config).toEqual(originalProvider.config);
|
|
expect(recreatedProvider.id()).toBe(originalProvider.id());
|
|
});
|
|
});
|