openpencil/scripts/ab-corpus/clients/deepseek.ts
2026-05-03 21:00:00 +08:00

49 lines
1.6 KiB
TypeScript

/**
* DeepSeek wrapper. OpenAI-compatible endpoint at api.deepseek.com.
* Key in `DEEPSEEK_API_KEY`. Latest flagship at time of wiring is
* `deepseek-v4-pro`; `deepseek-v4-flash` is the lighter variant; the
* older `deepseek-chat` / `deepseek-reasoner` IDs are deprecated
* (sunset 2026-07-24) but the harness routes them anyway so corpus
* comparisons against pre-deprecation runs still work.
*
* Docs: https://api-docs.deepseek.com/zh-cn/
*/
import type { ChatCallResult } from './openai-compat';
import { callOpenAICompat } from './openai-compat';
const DEEPSEEK_BASE_URL = process.env.DEEPSEEK_BASE_URL ?? 'https://api.deepseek.com';
const DEEPSEEK_API_KEY_ENV = 'DEEPSEEK_API_KEY';
export interface CallDeepSeekArgs {
model: string;
system: string;
user: string;
temperature?: number;
maxTokens?: number;
}
export async function callDeepSeek(args: CallDeepSeekArgs): Promise<ChatCallResult> {
const apiKey = process.env[DEEPSEEK_API_KEY_ENV];
if (!apiKey) {
throw new Error(
`${DEEPSEEK_API_KEY_ENV} not set — export it before running with --live, or pass --dry-run to use fixtures`,
);
}
return callOpenAICompat({
baseURL: DEEPSEEK_BASE_URL,
apiKey,
model: args.model,
system: args.system,
user: args.user,
temperature: args.temperature,
maxTokens: args.maxTokens,
label: 'deepseek',
// ab-v2 garbage attribution: 6 of 40 deepseek-v4-pro runs returned
// empty content (server-side flakiness, not model output). retries=1
// recovered most but ab-v3 still leaked 6 empty-content errors.
// Bumped to 2 alongside ark to share the exponential backoff window.
retries: 2,
});
}