Files
Agent-Workbench/tests/fixtures/cache-usage-model.mjs
T
2026-11-07 11:20:02 +08:00

48 lines
4.6 KiB
JavaScript

import http from 'node:http';
export const usageSamples = {
chat: { protocol: 'chat-completions', raw: { prompt_tokens: 1000, completion_tokens: 100, total_tokens: 1100,
prompt_tokens_details: { cached_tokens: 600, cache_write_tokens: 300, uncached_tokens: 100 }, completion_tokens_details: { reasoning_tokens: 40 } } },
responses: { protocol: 'responses', raw: { input_tokens: 12000, output_tokens: 100, total_tokens: 12100,
input_tokens_details: { cached_tokens: 4000, cache_write_tokens: 7000, cache_write_tokens_details: { ttl_30m_tokens: 3000, ttl_12h_tokens: 4000 } }, output_tokens_details: { reasoning_tokens: 40 } } },
messages: { protocol: 'anthropic-messages', raw: { input_tokens: 50, output_tokens: 12, cache_read_input_tokens: 100, cache_creation_input_tokens: 248,
cache_creation: { ephemeral_5m_input_tokens: 148, ephemeral_1h_input_tokens: 100 } } },
};
export const usageModelId = 'owned-cache-usage';
const event = value => 'data: ' + (typeof value === 'string' ? value : JSON.stringify(value)) + '\r\n\r\n';
const messageEvent = (type, value) => 'event: ' + type + '\r\n' + event({ type, ...value });
export function usageResponse(protocol, usage, text = '这是本机用量夹具的完整回复,没有执行工具。') {
const output = { type: 'message', id: 'msg_owned_usage', role: 'assistant', status: 'completed', content: [{ type: 'output_text', text }] };
if (protocol === 'responses') return { id: 'resp_owned_usage', model: usageModelId, status: 'completed', output: [output], ...(usage ? { usage } : {}) };
if (protocol === 'anthropic-messages') return { id: 'msg_owned_usage', type: 'message', model: usageModelId, role: 'assistant', content: [{ type: 'text', text }], stop_reason: 'end_turn', stop_sequence: null, ...(usage ? { usage } : {}) };
return { id: 'chat_owned_usage', model: usageModelId, choices: [{ index: 0, message: { role: 'assistant', content: text }, finish_reason: 'stop' }], ...(usage ? { usage } : {}) };
}
export function usageStream(protocol, usage, terminal = true) {
const final = usageResponse(protocol, usage);
if (protocol === 'responses') return event({ type: 'response.created', sequence_number: 0, response: { id: final.id, model: usageModelId } })
+ (terminal ? event({ type: 'response.completed', sequence_number: 1, response: final }) : '');
if (protocol === 'anthropic-messages') return messageEvent('message_start', { message: { ...final, content: [], stop_reason: null, usage: { ...usage, output_tokens: 0 } } })
+ messageEvent('content_block_start', { index: 0, content_block: { type: 'text', text: '' } })
+ messageEvent('content_block_delta', { index: 0, delta: { type: 'text_delta', text: final.content[0].text } })
+ messageEvent('content_block_stop', { index: 0 })
+ messageEvent('message_delta', { delta: { stop_reason: 'end_turn', stop_sequence: null }, usage: { output_tokens: usage.output_tokens } })
+ (terminal ? messageEvent('message_stop', {}) : '');
return event({ id: final.id, model: usageModelId, choices: [{ index: 0, delta: { role: 'assistant', content: final.choices[0].message.content }, finish_reason: null }] })
+ event({ id: final.id, model: usageModelId, choices: [{ index: 0, delta: {}, finish_reason: 'stop' }] })
+ (usage ? event({ id: final.id, model: usageModelId, choices: [], usage }) : '') + (terminal ? event('[DONE]') : '');
}
export async function cacheUsageFixture(t, { protocol, raw, streaming = false, terminal = true, status = 200 }) {
const requests = [], responses = [];
const server = http.createServer(async (request, response) => {
const chunks = []; for await (const chunk of request) chunks.push(chunk);
requests.push({ url: request.url, input: JSON.parse(Buffer.concat(chunks)) });
const value = status === 200 ? usageResponse(protocol, raw) : { error: { type: 'owned_usage_error', code: 'owned_usage_400', message: '自有上游原错误,不重试' }, usage: raw };
const bytes = Buffer.from(streaming && status === 200 ? usageStream(protocol, raw, terminal) : JSON.stringify(value)); responses.push(bytes);
response.writeHead(status, { 'Content-Type': streaming && status === 200 ? 'text/event-stream' : 'application/json', 'x-request-id': 'owned-usage-request' });
for (let offset = 0; offset < bytes.length; offset += 7) response.write(bytes.subarray(offset, offset + 7)); response.end();
});
await new Promise(resolve => server.listen(0, '127.0.0.1', resolve));
t.after(() => new Promise(resolve => { server.closeAllConnections(); server.close(resolve); }));
return { baseUrl: 'http://127.0.0.1:' + server.address().port + '/v1', requests, responses };
}