ai-functions
Version:
Core AI primitives for building intelligent applications
165 lines (164 loc) • 7.36 kB
JavaScript
/**
* E2E Tests for AWS Bedrock Flex Tier Processing
*
* These tests hit real AWS Bedrock APIs to verify the flex tier actually works.
* Requires AWS credentials:
* - AWS_ACCESS_KEY_ID
* - AWS_SECRET_ACCESS_KEY
* - AWS_REGION (optional, defaults to us-east-1)
*
* Run with: npx vitest run test/e2e-bedrock.test.ts
*/
import { describe, it, expect, beforeAll, afterAll } from 'vitest';
import { configure, resetContext, getExecutionTier, isFlexAvailable } from '../src/context.js';
// Import the Bedrock adapter to register it
import { configureAWSBedrock, bedrockFlexAdapter } from '../src/batch/bedrock.js';
import '../src/batch/bedrock.js';
// Skip tests if no AWS credentials
// NOTE: Unlike OpenAI/Google, Bedrock via AI Gateway still requires AWS SigV4 signing
// Gateway alone is NOT sufficient - you need AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY
const hasCredentials = !!(process.env.AWS_ACCESS_KEY_ID && process.env.AWS_SECRET_ACCESS_KEY);
// Claude Opus 4.5 model ID on Bedrock
const CLAUDE_OPUS_MODEL = 'anthropic.claude-opus-4-20250514-v1:0';
// Fallback to Sonnet if Opus isn't available
const CLAUDE_SONNET_MODEL = 'anthropic.claude-sonnet-4-20250514-v1:0';
describe.skipIf(!hasCredentials)('E2E Bedrock Flex Tier Processing', () => {
beforeAll(() => {
// Configure Bedrock - can use either direct AWS credentials or AI Gateway
configureAWSBedrock({
region: process.env.AWS_REGION || 'us-east-1',
s3Bucket: process.env.BEDROCK_BATCH_S3_BUCKET || 'dummy-bucket',
// AI Gateway configuration (optional)
gatewayUrl: process.env.AI_GATEWAY_URL,
gatewayToken: process.env.AI_GATEWAY_TOKEN,
});
configure({
provider: 'bedrock',
model: CLAUDE_SONNET_MODEL, // Use Sonnet by default (more available)
batchMode: 'auto',
flexThreshold: 3,
batchThreshold: 500,
});
});
afterAll(() => {
resetContext();
});
describe('Flex Adapter Direct', () => {
it('processes a small batch via flex adapter', async () => {
const items = [
{
id: 'item_1',
prompt: 'What is 2 + 2? Reply with just the number.',
status: 'pending',
},
{
id: 'item_2',
prompt: 'What is 3 + 3? Reply with just the number.',
status: 'pending',
},
{
id: 'item_3',
prompt: 'What is 4 + 4? Reply with just the number.',
status: 'pending',
},
];
const results = await bedrockFlexAdapter.submitFlex(items, { model: CLAUDE_SONNET_MODEL });
expect(results).toHaveLength(3);
results.forEach((result) => {
expect(result.status).toBe('completed');
expect(result.result).toBeDefined();
});
// Check the actual responses contain numbers
const result1 = results.find((r) => r.id === 'item_1');
const result2 = results.find((r) => r.id === 'item_2');
const result3 = results.find((r) => r.id === 'item_3');
expect(String(result1?.result)).toContain('4');
expect(String(result2?.result)).toContain('6');
expect(String(result3?.result)).toContain('8');
}, 60000); // 60 second timeout for Bedrock
it('reports token usage', async () => {
const items = [
{
id: 'usage_test',
prompt: 'Count from 1 to 5',
status: 'pending',
},
];
const results = await bedrockFlexAdapter.submitFlex(items, { model: CLAUDE_SONNET_MODEL });
expect(results[0].usage).toBeDefined();
expect(results[0].usage?.promptTokens).toBeGreaterThan(0);
expect(results[0].usage?.completionTokens).toBeGreaterThan(0);
expect(results[0].usage?.totalTokens).toBeGreaterThan(0);
}, 60000);
it('handles system prompts', async () => {
const items = [
{
id: 'system_test',
prompt: 'What should I say?',
options: {
system: 'You are a helpful assistant that always responds with exactly "Hello World"',
},
status: 'pending',
},
];
const results = await bedrockFlexAdapter.submitFlex(items, { model: CLAUDE_SONNET_MODEL });
expect(results[0].status).toBe('completed');
expect(String(results[0].result).toLowerCase()).toContain('hello');
}, 60000);
});
describe('Execution Tier Selection', () => {
it('selects immediate tier for < flexThreshold items', () => {
configure({ batchMode: 'auto', flexThreshold: 5, batchThreshold: 500 });
expect(getExecutionTier(1)).toBe('immediate');
expect(getExecutionTier(4)).toBe('immediate');
});
it('selects flex tier for flexThreshold to < batchThreshold items', () => {
configure({ batchMode: 'auto', flexThreshold: 5, batchThreshold: 500 });
expect(getExecutionTier(5)).toBe('flex');
expect(getExecutionTier(100)).toBe('flex');
});
it('reports flex as available for bedrock', () => {
configure({ provider: 'bedrock' });
expect(isFlexAvailable()).toBe(true);
});
});
describe('Concurrent Processing', () => {
it('processes multiple items concurrently', async () => {
// Create 5 items to test concurrent processing
const items = Array.from({ length: 5 }, (_, i) => ({
id: `concurrent_${i}`,
prompt: `What is ${i} + 1? Reply with just the number.`,
status: 'pending',
}));
const startTime = Date.now();
const results = await bedrockFlexAdapter.submitFlex(items, { model: CLAUDE_SONNET_MODEL });
const duration = Date.now() - startTime;
expect(results).toHaveLength(5);
results.forEach((result) => {
expect(result.status).toBe('completed');
});
console.log(`Processed 5 items in ${duration}ms`);
}, 120000);
});
});
// Optional test with Claude Opus 4.5 (may not be available in all regions)
describe.skipIf(!hasCredentials)('E2E Bedrock with Claude Opus 4.5', () => {
it.skip('processes with Claude Opus 4.5', async () => {
const items = [
{
id: 'opus_test',
prompt: 'Write a haiku about AI.',
status: 'pending',
},
];
try {
const results = await bedrockFlexAdapter.submitFlex(items, { model: CLAUDE_OPUS_MODEL });
expect(results[0].status).toBe('completed');
expect(results[0].result).toBeDefined();
console.log('Claude Opus 4.5 response:', results[0].result);
}
catch (error) {
console.log('Claude Opus 4.5 not available in this region:', error);
}
}, 120000);
});