{"id":10021533,"date":"2026-11-02T09:58:18","date_gmt":"2026-11-02T01:58:18","guid":{"rendered":"https:\/\/itmax.cn\/?p=10021533"},"modified":"2026-11-02T09:58:27","modified_gmt":"2026-11-02T01:58:27","slug":"promptfoo-20261102","status":"publish","type":"post","link":"https:\/\/itmax.cn\/?p=10021533","title":{"rendered":"\u6539\u63d0\u793a\u8bcd\u522b\u518d\u9760\u624b\u611f\u4e86\uff5cpromptfoo \u88ab OpenAI \u6536\u7f16\u540e\uff0c\u6211\u628a\u5b83\u88c5\u4e86\u4e00\u904d"},"content":{"rendered":"<h1 id=\"promptfoo-openai\">\u6539\u63d0\u793a\u8bcd\u522b\u518d\u9760\u624b\u611f\u4e86\uff5cpromptfoo \u88ab OpenAI \u6536\u7f16\u540e\uff0c\u6211\u628a\u5b83\u88c5\u4e86\u4e00\u904d<\/h1>\n<figure style=\"margin:2em 0;text-align:center;\">\n<img decoding=\"async\" src=\"https:\/\/itmax.cn\/wp-content\/uploads\/2026\/11\/promptfoo-cover.png\" alt=\"promptfoo \u5c01\u9762\u56fe\uff1aGitHub 26,708 \u661f\uff0c\u6700\u65b0\u7248\u672c 0.125.1\uff0c\u8fd1\u56db\u5468 441 \u6b21\u63d0\u4ea4\" style=\"max-width:100%;height:auto;border-radius:8px;\"><figcaption style=\"font-size:0.9em;color:#666;margin-top:0.5em;\">promptfoo\uff1a\u8dd1\u5728\u7ec8\u7aef\u91cc\u7684\u5927\u6a21\u578b<code>\u56de\u5f52\u6d4b\u8bd5<\/code>\u4e0e\u7ea2\u961f\u626b\u63cf CLI\uff0cOpenAI \u6536\u7f16\u540e\u4ecd\u7ef4\u6301 MIT \u5f00\u6e90<\/figcaption><\/figure>\n<h2 id=\"_1\">\u4e0a\u7ebf\u524d\u90a3\u4e00\u4e0b\uff0c\u6211\u5fc3\u91cc\u6ca1\u5e95<\/h2>\n<p>\u5148\u8bf4\u4e00\u4e2a\u6211\u81ea\u5df1\u683d\u8fc7\u7684\u8ddf\u5934\u3002\u53bb\u5e74\u7ed9\u4e00\u4e2a\u5ba2\u670d\u573a\u666f\u8c03\u7cfb\u7edf\u63d0\u793a\u8bcd\uff0c\u6539\u4e86\u4e09\u7248\uff0c\u6bcf\u7248\u6211\u624b\u52a8\u5582\u5341\u51e0\u6761\u7528\u6237\u8f93\u5165\uff0c\u770b\u7740\u8f93\u51fa\u89c9\u5f97\u300c\u597d\u50cf\u987a\u773c\u4e86\u300d\uff0c\u5c31\u53d1\u4e0a\u53bb\u4e86\u3002\u7ed3\u679c\u7b2c\u4e8c\u5929\u8fd0\u8425\u627e\u8fc7\u6765\uff1a\u6709\u4e00\u7c7b\u95ee\u9000\u6b3e\u6d41\u7a0b\u7684\u8f93\u5165\uff0c\u65b0\u7248\u672c\u628a\u5173\u952e\u6b65\u9aa4\u8bf4\u4e22\u4e86\u3002\u5341\u51e0\u6761\u6837\u672c\u91cc\u6ca1\u6709\u8fd9\u4e00\u7c7b\uff0c\u6211\u7684\u300c\u624b\u611f\u300d\u5728\u5b83\u9762\u524d\u4e00\u6587\u4e0d\u503c\u3002<\/p>\n<p>\u66f4\u96be\u53d7\u7684\u662f\uff0c\u6211\u8bf4\u4e0d\u6e05\u8fd9\u4e00\u7248\u5230\u5e95\u6bd4\u4e0a\u4e00\u7248\u597d\u8fd8\u662f\u574f\u3002\u6ca1\u6709\u57fa\u7ebf\uff0c\u6ca1\u6709\u8bb0\u5f55\uff0c\u6539\u574f\u4e86\u90fd\u5bdf\u89c9\u4e0d\u5230\u3002\u505a\u8fc7\u51e0\u5e74\u540e\u7aef\u7684\u4eba\u5bf9\u8fd9\u79cd\u611f\u89c9\u5e94\u8be5\u4e0d\u964c\u751f\u2014\u2014\u5c31\u50cf\u5199\u4ee3\u7801\u4e0d\u5199\u6d4b\u8bd5\uff0c\u5168\u9760\u4e0a\u751f\u4ea7\u73af\u5883\u770b\u65e5\u5fd7\u3002\u6d4b\u8bd5\u7f3a\u5931\u7684\u4ee3\u4ef7\u4ece\u6765\u4e0d\u662f\u300c\u6d4b\u8bd5\u300d\u672c\u8eab\uff0c\u800c\u662f\u4f60\u6c38\u8fdc\u5728\u8d4c\uff0c\u8d4c\u6ce8\u662f\u7528\u6237\u7684\u4fe1\u4efb\u3002<\/p>\n<p>\u4eca\u5e74\u4e09\u6708\uff0cOpenAI \u5b98\u5ba3\u628a promptfoo \u6536\u7f16\u4e86\uff08\u5b98\u65b9\u516c\u544a\u6807\u9898\u5c31\u662f\u300cOpenAI \u5c06\u6536\u8d2d Promptfoo\u300d\uff09\uff0c\u56e2\u961f\u6574\u4f53\u52a0\u5165\uff0c\u6280\u672f\u4f1a\u5e76\u8fdb\u5b83\u7684 Frontier \u5e73\u53f0\u3002\u6d88\u606f\u51fa\u6765\u540e\u6211\u8eab\u8fb9\u4e0d\u5c11\u4eba\u7684\u7b2c\u4e00\u53cd\u5e94\u662f\uff1a\u8fd9\u5f00\u6e90\u5de5\u5177\u8fd8\u80fd\u7528\u591a\u4e45\uff1f\u4f1a\u4e0d\u4f1a\u6162\u6162\u53d8\u6210\u6536\u8d39\u5e73\u53f0\u7684\u94a9\u5b50\uff1f\u5e26\u7740\u8fd9\u4e2a\u7591\u95ee\uff0c\u6211\u51b3\u5b9a\u628a\u5b83\u88c5\u4e00\u904d\uff0c\u628a\u8fd9\u7b14\u8d26\u7b97\u6e05\u695a\u3002<\/p>\n<p>\u987a\u5e26\u4ea4\u4ee3\u4e00\u4e0b\u80cc\u666f\uff1a\u8fd9\u4e2a\u9879\u76ee 2023 \u5e74 4 \u6708\u5c31\u5efa\u4ed3\u4e86\uff0c\u521b\u59cb\u4eba Ian Webster \u548c Michael D&#8217;Angelo \u5e26\u961f\uff0cTypeScript \u5199\u7684\uff0cMIT \u534f\u8bae\uff0c\u4ece\u300c\u5bf9\u6bd4\u51e0\u4e2a\u6a21\u578b\u7684\u63d0\u793a\u8bcd\u6548\u679c\u300d\u8fd9\u79cd\u5c0f\u5de5\u5177\u4e00\u8def\u957f\u6210\u73b0\u5728\u8fd9\u4e2a\u4f53\u91cf\u3002\u80fd\u88ab\u6536\u7f16\uff0c\u672c\u8eab\u8bf4\u660e\u5b83\u8e29\u4e2d\u4e86\u4e00\u4e2a\u771f\u5b9e\u9700\u6c42\u2014\u2014\u5927\u6a21\u578b\u5e94\u7528\u6d4b\u8d77\u6765\u592a\u6ca1\u7ae0\u6cd5\u4e86\u3002<\/p>\n<h2 id=\"_2\">\u5b83\u5230\u5e95\u7ba1\u54ea\u51e0\u644a\u4e8b<\/h2>\n<p>\u88c5\u5b8c\u7ffb\u4e86\u6587\u6863\u548c\u6e90\u7801\u76ee\u5f55\uff0c\u6211\u8ba4\u4e3a\u5b83\u5b9e\u9645\u662f\u56db\u644a\u4e8b\uff0c\u6210\u8272\u4e0d\u4e00\u6837\uff1a<\/p>\n<p><strong>\u7b2c\u4e00\u644a\uff0c\u8bc4\u4f30\u77e9\u9635\uff0c\u8fd9\u662f\u5b83\u7684\u57fa\u672c\u76d8\u3002<\/strong> \u4e00\u4e2a YAML \u6587\u4ef6\u5199\u6e05\u695a\u4e09\u6837\u4e1c\u897f\uff1a\u63d0\u793a\u8bcd\uff08\u53ef\u4ee5\u591a\u4e2a\u7248\u672c\u5e76\u5b58\uff09\u3001\u6a21\u578b provider\u3001\u6d4b\u8bd5\u7528\u4f8b\u52a0\u65ad\u8a00\u3002\u8dd1\u4e00\u6761\u547d\u4ee4\uff0c\u8f93\u51fa\u4e00\u5f20\u300c\u7528\u4f8b \u00d7 \u63d0\u793a\u8bcd\u7248\u672c\u300d\u7684\u901a\u8fc7\u7387\u77e9\u9635\u3002\u6211\u90a3\u4e2a\u5ba2\u670d\u573a\u666f\u7684\u75db\u70b9\uff0c\u5b83\u7ed9\u7684\u5c31\u662f\u8fd9\u4e48\u4e2a\u4e1c\u897f\uff1a\u6bcf\u6539\u4e00\u7248\u63d0\u793a\u8bcd\uff0c\u8dd1\u4e00\u904d\u77e9\u9635\uff0c\u6570\u5b57\u6446\u5728\u90a3\u91cc\uff0c\u597d\u8fd8\u662f\u574f\u4e0d\u7528\u731c\u3002\u77e9\u9635\u8fd8\u662f\u53cc\u5411\u7684\u2014\u2014\u4f60\u53ef\u4ee5\u540c\u65f6\u6a2a\u5411\u6bd4\u63d0\u793a\u8bcd\u7248\u672c\uff0c\u4e5f\u53ef\u4ee5\u7eb5\u5411\u6bd4 GPT\u3001Claude\u3001Gemini\u3001DeepSeek \u4e43\u81f3\u672c\u5730 Ollama \u7aef\u70b9\uff0c\u5b98\u65b9\u652f\u6301\u7684 provider \u6709\u516d\u5341\u591a\u4e2a\uff0c\u4e3b\u6d41\u7684\u4e91\u5382\u5546\u57fa\u672c\u90fd\u9f50\u4e86\u3002<\/p>\n<p><strong>\u7b2c\u4e8c\u644a\uff0c\u65ad\u8a00\u4f53\u7cfb\uff0c\u8fd9\u662f\u77e9\u9635\u7684\u7075\u9b42\u3002<\/strong> \u786e\u5b9a\u6027\u65ad\u8a00\u6709 <code>contains<\/code>\u3001<code>equals<\/code>\u3001<code>similar<\/code>\uff08\u57fa\u4e8e embedding \u76f8\u4f3c\u5ea6\u8bbe\u9608\u503c\uff09\u3001<code>javascript<\/code>\uff08\u81ea\u5df1\u5199\u51fd\u6570\u5224\u65ad\u8f93\u51fa\uff09\uff1b\u6a21\u578b\u6253\u5206\u65ad\u8a00\u6709 <code>llm-rubric<\/code>\u3001<code>factuality<\/code> \u8fd9\u7c7b\uff0c\u8ba9\u4e00\u4e2a\u66f4\u5f3a\u7684\u6a21\u578b\u5f53\u88c1\u5224\uff0c\u88c1\u5224\u53ef\u4ee5\u8ddf\u88ab\u6d4b\u6a21\u578b\u4e0d\u662f\u540c\u4e00\u5bb6\u3002\u6211\u7684\u5224\u65ad\u662f\uff1a\u786e\u5b9a\u6027\u65ad\u8a00\u662f\u771f\u80fd\u7528\uff0c\u95ed\u7740\u773c\u775b\u4fe1\uff1b\u6a21\u578b\u6253\u5206\u90a3\u51e0\u4e2a\u8981\u770b\u88c1\u5224\u6a21\u578b\u6210\u8272\uff0c\u88c1\u5224\u81ea\u5df1\u90fd\u4f1a\u6296\uff0c\u540c\u4e00\u4efd\u8f93\u51fa\u4e24\u6b21\u6253\u5206\u53ef\u80fd\u7ed9\u51fa\u4e0d\u540c\u7ed3\u8bba\uff0c\u5f97\u5206\u53ea\u80fd\u5f53\u53c2\u8003\uff0c\u522b\u5f53\u95e8\u7981\u7684\u786c\u6307\u6807\u3002\u8981\u5f53\u95e8\u7981\uff0c\u5c31\u914d <code>threshold<\/code> \u628a\u7ebf\u62c9\u9ad8\uff0c\u518d\u7559\u51e0\u6761\u786e\u5b9a\u6027\u65ad\u8a00\u515c\u5e95\u3002<\/p>\n<p><strong>\u7b2c\u4e09\u644a\uff0c\u7ea2\u961f\u626b\u63cf\uff0c\u8fd9\u662f\u5b83\u8fd1\u4e24\u5e74\u53d1\u529b\u7684\u65b9\u5411\u3002<\/strong> \u81ea\u52a8\u751f\u6210\u8d8a\u72f1\u3001\u63d0\u793a\u6ce8\u5165\u3001PII \u6cc4\u9732\u8fd9\u7c7b\u653b\u51fb\u8f93\u5165\u53bb\u6253\u4f60\u7684\u5e94\u7528\uff0c\u5b98\u65b9\u8bf4\u8986\u76d6 80 \u591a\u79cd\u653b\u51fb\u63d2\u4ef6\uff0c\u5bf9\u9f50 OWASP LLM Top 10 \u548c MITRE ATLAS\uff0c\u8fd8\u80fd\u6a21\u62df\u591a\u8f6e\u5347\u7ea7\u5f0f\u653b\u51fb\u3002\u65b9\u5411\u662f\u771f\u6709\u4ef7\u503c\u2014\u2014\u5b89\u5168\u8fd9\u4ef6\u4e8b\u9760\u4eba\u5de5\u60f3\u7528\u4f8b\uff0c\u8986\u76d6\u9762\u6ce8\u5b9a\u53ef\u601c\u3002\u4f46\u8981\u6cfc\u51b7\u6c34\u7684\u662f\uff1a\u653b\u51fb\u8d28\u91cf\u53c2\u5dee\uff0c\u4f4e\u5782\u7684\u679c\u5b9e\u5b83\u6458\u5f97\u52a8\uff0c\u9488\u5bf9\u4f60\u4e1a\u52a1\u903b\u8f91\u7684\u5b9a\u5411\u653b\u51fb\u8fd8\u5f97\u81ea\u5df1\u8865\uff1b\u800c\u4e14\u751f\u6210\u653b\u51fb\u548c\u7ed9\u653b\u51fb\u7ed3\u679c\u6253\u5206\u90fd\u8981\u8c03\u6a21\u578b\uff0c\u8dd1\u4e00\u8f6e\u662f\u6709\u5b9e\u6253\u5b9e\u7684 API \u8d26\u5355\u7684\u3002<\/p>\n<p><strong>\u7b2c\u56db\u644a\uff0cCI \u96c6\u6210\u548c\u4ee3\u7801\u626b\u63cf\u3002<\/strong> GitHub Action\u3001GitLab\u3001Jenkins \u90fd\u6709\u73b0\u6210\u63a5\u6cd5\uff0c\u6539\u4e86\u63d0\u793a\u8bcd\u81ea\u52a8\u56de\u5f52\uff0cPR \u91cc\u76f4\u63a5\u6302\u901a\u8fc7\u7387\uff1b\u8fd8\u6709\u4e2a\u9488\u5bf9\u4ed3\u5e93\u5185\u63d0\u793a\u8bcd\u6587\u4ef6\u505a\u5b89\u5168\u626b\u63cf\u7684 code scanning\uff0c\u5c5e\u4e8e\u6bd4\u8f83\u65b0\u7684\u4e1c\u897f\uff0c\u6211\u7528\u5f97\u8fd8\u4e0d\u6df1\u3002<\/p>\n<p>\u5766\u767d\u8bb2\uff0c\u7b2c\u4e00\u3001\u4e8c\u644a\u662f\u624e\u5b9e\u7684\uff0c\u7b2c\u4e09\u644a\u503c\u4e00\u534a\u7968\u4ef7\uff0c\u7b2c\u56db\u644a\u504f\u4f01\u4e1a\u53d9\u4e8b\u3002\u522b\u88ab\u5b98\u7f51\u90a3\u4e00\u957f\u4e32\u529f\u80fd\u6e05\u5355\u552c\u4f4f\uff0c\u6838\u5fc3\u5176\u5b9e\u5c31\u662f\u628a\u300c\u6539\u63d0\u793a\u8bcd\u300d\u8fd9\u4ef6\u4e8b\u53d8\u6210\u53ef\u56de\u5f52\u7684\u6d4b\u8bd5\u3002\u53e6\u5916\u4ea4\u4ee3\u4e00\u53e5\u6211\u7684\u9a8c\u8bc1\u8fb9\u754c\uff1a\u7ea2\u961f\u626b\u63cf\u90a3\u644a\u6211\u6ca1\u5b9e\u9645\u8dd1\u2014\u2014\u751f\u6210\u653b\u51fb\u8981\u6d88\u8d39\u6a21\u578b API\uff0c\u6211\u4e0d\u60f3\u4e3a\u8fd9\u7bc7\u4e1c\u897f\u70e7\u4e00\u7b14\u9884\u7b97\u2014\u2014\u4e0b\u9762\u6d89\u53ca\u7ea2\u961f\u7684\u5224\u65ad\u5168\u90e8\u6765\u81ea\u5b98\u65b9\u6587\u6863\u3001issue \u533a\u548c\u7b2c\u4e09\u65b9\u5b9e\u6d4b\u6587\u7ae0\u7684\u4ea4\u53c9\u6838\u5bf9\uff0c\u4e0d\u662f\u6211\u4eb2\u624b\u6253\u7684\u3002<\/p>\n<h2 id=\"_3\">\u58f0\u660e\u5f0f\u8fd9\u5957\u724c\uff0c\u4ee3\u4ef7\u4e5f\u660e\u6446\u7740<\/h2>\n<p>\u5b83\u4e3a\u4ec0\u4e48\u9009 YAML \u58f0\u660e\u5f0f\uff0c\u800c\u4e0d\u662f\u50cf\u522b\u7684\u6846\u67b6\u90a3\u6837\u8ba9\u4f60\u5199 Python \u4ee3\u7801\u7ec4\u7ec7\u6d4b\u8bd5\uff1f\u6211\u8ba4\u4e3a\u7406\u7531\u5f88\u786c\uff1a\u58f0\u660e\u5f0f\u914d\u7f6e\u624d\u80fd\u8fdb\u4ee3\u7801\u8bc4\u5ba1\u3001\u624d\u80fd\u8fdb CI\u3001\u624d\u80fd\u8ba9\u300c\u63d0\u793a\u8bcd\u6539\u4e86\u4ec0\u4e48\u300d\u53d8\u6210\u4e00\u6b21 git diff\u3002\u63d0\u793a\u8bcd\u5c31\u662f\u8d44\u4ea7\uff0c\u8d44\u4ea7\u5c31\u8981\u6709\u7248\u672c\u3001\u6709\u57fa\u7ebf\u3001\u6709\u53d8\u66f4\u8bb0\u5f55\u2014\u2014\u8fd9\u662f\u5b83\u8ddf\u6d4b\u8bd5\u9a71\u52a8\u5f00\u53d1\u7684\u601d\u8def\u4e00\u8109\u76f8\u627f\u7684\u5730\u65b9\uff0c\u4e5f\u662f\u5b83\u8ddf\u4e00\u5806\u300c playground \u5957\u58f3\u300d\u62c9\u5f00\u5dee\u8ddd\u7684\u5206\u6c34\u5cad\u3002<\/p>\n<p>\u4f46\u4ee3\u4ef7\u540c\u6837\u660e\u6446\u7740\u3002\u7b2c\u4e00\uff0c\u7a0d\u5fae\u52a8\u6001\u4e00\u70b9\u7684\u6d41\u7a0b\u2014\u2014\u591a\u8f6e\u5bf9\u8bdd\u72b6\u6001\u3001\u6309\u8f93\u5165\u5206\u652f\u7684\u8c03\u7528\u94fe\u3001\u5e26\u5de5\u5177\u8c03\u7528\u7684\u7f16\u6392\u2014\u2014YAML \u8868\u8fbe\u8d77\u6765\u5c31\u5f88\u62e7\u5df4\uff0c\u5b98\u65b9\u7684\u89e3\u6cd5\u662f\u585e JavaScript \u7247\u6bb5\u6216\u8005\u5916\u6302\u81ea\u5df1\u7684 provider\uff0c\u5199\u591a\u4e86\u914d\u7f6e\u6587\u4ef6\u5c31\u6210\u4e86\u4e00\u5806\u7f1d\u8865\u7684\u8865\u4e01\u3002\u7b2c\u4e8c\uff0c\u914d\u7f6e\u6587\u4ef6\u4f1a\u81a8\u80c0\uff0c\u7528\u4f8b\u4e00\u591a\uff0c\u51e0\u767e\u884c\u7684 YAML \u7ef4\u62a4\u8d77\u6765\u5e76\u4e0d\u6bd4\u5199\u4ee3\u7801\u8f7b\u677e\uff0c\u800c\u4e14\u5b83\u6ca1\u6709\u5185\u7f6e\u7684\u63d0\u793a\u8bcd\u7248\u672c\u7ba1\u7406\uff0c\u5f97\u9760 git \u81ea\u5df1\u515c\u7740\u3002\u7b2c\u4e09\uff0c\u6574\u4e2a\u5de5\u5177\u662f\u4e2a Node \u5355\u4f53\uff0c\u88c5\u4e00\u6b21\u516d\u767e\u591a\u4e2a\u5305\uff0c\u5bf9\u4e60\u60ef\u4e86\u8f7b\u91cf CLI \u7684\u4eba\u6765\u8bf4\uff0c\u8fd9\u4e2a\u4f53\u79ef\u662f\u6709\u5fc3\u7406\u95e8\u69db\u7684\u3002<\/p>\n<p>\u8fd8\u6709\u4e00\u4e2a\u7ed5\u4e0d\u5f00\u7684\u70b9\uff1a\u6570\u636e\u672c\u5730\u5316\u3002\u8bc4\u4f30\u7ed3\u679c\u9ed8\u8ba4\u5b58\u5728\u672c\u5730\u7684 SQLite \u91cc\uff0c\u63d0\u793a\u8bcd\u548c\u6570\u636e\u4e0d\u51fa\u95e8\u2014\u2014\u8fd9\u5bf9\u5728\u610f\u654f\u611f\u6570\u636e\u7684\u56e2\u961f\u662f\u5b9e\u6253\u5b9e\u7684\u4f18\u70b9\uff0c\u6d4b\u8bd5\u7528\u4f8b\u91cc\u5e38\u6df7\u7740\u771f\u5b9e\u5ba2\u6237\u5bf9\u8bdd\uff0c\u8c01\u4e5f\u4e0d\u60f3\u628a\u8fd9\u4e9b\u4e1c\u897f\u9001\u51fa\u53bb\u3002\u4f46\u4ee3\u4ef7\u662f\u56e2\u961f\u534f\u4f5c\u8981\u4e48\u81ea\u6258\u7ba1\u4e00\u4e2a\u7ed3\u679c\u67e5\u770b\u670d\u52a1\uff0c\u8981\u4e48\u628a\u7ed3\u679c\u4f20\u5230\u5b83\u5b98\u65b9\u7684\u5206\u4eab\u7aef\u70b9\u3002\u81ea\u6258\u7ba1\u8fd9\u6761\u8def\uff0c\u5751\u8fd8\u4e0d\u7b97\u5c11\uff0c\u4e0b\u9762\u7ec6\u8bf4\u3002\u53e6\u5916\u5b83\u9ed8\u8ba4\u5f00\u9065\u6d4b\uff0c\u4ecb\u610f\u7684\u8bdd\u8bb0\u5f97\u8bbe <code>PROMPTFOO_DISABLE_TELEMETRY=1<\/code>\uff0c\u73af\u5883\u53d8\u91cf\u4e00\u8bbe\u5c31\u5b89\u9759\u4e86\u3002<\/p>\n<p>\u518d\u5f80\u6df1\u770b\u4e00\u5c42\uff0c\u5b83\u7684 provider \u62bd\u8c61\u5176\u5b9e\u5f88\u8584\uff1a\u4e0d\u7ba1\u63a5\u54ea\u5bb6\u6a21\u578b\uff0c\u6700\u7ec8\u90fd\u662f\u62fc HTTP \u8bf7\u6c42\u3001\u6536 JSON \u54cd\u5e94\uff0c\u518d\u6309\u7edf\u4e00\u7684\u683c\u5f0f\u5582\u7ed9\u65ad\u8a00\u5c42\u3002\u8fd9\u4e2a\u8bbe\u8ba1\u8ba9\u63a5\u5165\u65b0\u6a21\u578b\u7684\u6210\u672c\u6781\u4f4e\uff08\u5b98\u65b9\u4ed3\u5e93\u91cc provider \u76ee\u5f55\u662f\u8d21\u732e\u6700\u6d3b\u8dc3\u7684\u5730\u65b9\u4e4b\u4e00\uff09\uff0c\u4f46\u4e5f\u610f\u5473\u7740\u6bcf\u5bb6\u7684\u7279\u6b8a\u80fd\u529b\u2014\u2014\u7f13\u5b58\u7b56\u7565\u3001\u91cd\u8bd5\u8bed\u4e49\u3001\u8d85\u65f6\u7ec6\u8282\u2014\u2014\u90fd\u5f97\u5728\u5404\u81ea\u7684 provider \u5b9e\u73b0\u91cc\u5bf9\u9f50\uff0c\u9002\u914d\u8d28\u91cf\u5c31\u53c2\u5dee\u4e86\u3002\u8fd9\u89e3\u91ca\u4e86\u4e3a\u4ec0\u4e48 issue \u533a\u91cc provider \u7c7b\u95ee\u9898\u6c38\u8fdc\u4fee\u4e0d\u5b8c\uff1a\u4e0d\u662f\u5b83\u505a\u5f97\u5dee\uff0c\u662f\u5e95\u4e0b\u5404\u5bb6\u63a5\u53e3\u5b9e\u5728\u4e94\u82b1\u516b\u95e8\u3002<\/p>\n<h2 id=\"node-2222\">\u88c5\u4e00\u904d\uff1a\u95e8\u69db\u5c31\u4e00\u6761\uff0cNode 22.22<\/h2>\n<p>\u5b89\u88c5\u65b9\u5f0f\u5b98\u65b9\u7ed9\u4e86\u56db\u79cd\uff1a<code>npm install -g promptfoo<\/code>\u3001<code>brew install promptfoo<\/code>\u3001<code>pip install promptfoo<\/code>\u3001\u6216\u8005 <code>npx promptfoo@latest<\/code> \u514d\u5b89\u88c5\u76f4\u63a5\u8dd1\u3002\u6211\u7684\u5efa\u8bae\u662f\u957f\u671f\u7528\u5c31\u5168\u5c40\u88c5\uff0c<code>npx<\/code> \u6bcf\u6b21\u89e3\u6790\u4f9d\u8d56\u4f1a\u62d6\u6162\u542f\u52a8\uff0c\u5076\u5c14\u8bd5\u4e00\u4e0b\u624d\u503c\u5f97\u7528\u5b83\u3002<\/p>\n<p>\u771f\u6b63\u7684\u786c\u95e8\u69db\u5728 Node \u7248\u672c\u4e0a\uff0c<code>package.json<\/code> \u7684 engines \u5199\u5f97\u660e\u660e\u767d\u767d\uff1a<strong><code>node &gt;= 22.22.0<\/code><\/strong>\u3002\u8fd9\u4e2a\u7248\u672c\u53f7\u76f8\u5f53\u65b0\uff0c\u6211\u672c\u673a\u662f v22.22.2\uff0c\u5c5e\u4e8e\u8e29\u7ebf\u8fc7\u5173\u3002\u65e7\u4e00\u70b9\u7684 Node \u76f4\u63a5\u88c5\u4e0d\u4e0a\uff0c\u52a8\u624b\u524d\u5148 <code>node --version<\/code> \u770b\u4e00\u773c\uff0c\u522b\u7b49\u516d\u767e\u4e2a\u5305\u88c5\u5b8c\u624d\u62a5\u9519\u3002<\/p>\n<p>\u786c\u4ef6\u8fd9\u5757\u5f97\u8bf4\u5b9e\u8bdd\uff1a\u5b98\u65b9 README \u548c\u6587\u6863\u90fd\u6ca1\u7ed9\u6574\u673a\u914d\u7f6e\u8868\uff0c\u6211\u6309\u5b83\u7684\u5de5\u4f5c\u539f\u7406\u548c\u793e\u533a\u53cd\u9988\u4f30\u7684\u2014\u2014\u5b83\u672c\u8d28\u662f\u4e2a\u8c03\u5ea6\u5668\uff0c\u91cd\u6d3b\u662f\u8c03 API \u8ba9\u6a21\u578b\u5e72\uff0c\u672c\u5730\u53ea\u505a\u7f16\u6392\u3001\u65ad\u8a00\u548c\u5b58\u7ed3\u679c\uff1a<\/p>\n<table>\n<thead>\n<tr>\n<th>\u6863\u4f4d<\/th>\n<th>\u914d\u7f6e<\/th>\n<th>\u6211\u7684\u4f9d\u636e<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>\u6700\u4f4e\u53ef\u7528<\/td>\n<td>2 \u6838 CPU\u30014GB \u5185\u5b58\u30011GB \u78c1\u76d8<\/td>\n<td>\u8c03\u7528\u4e91\u7aef API \u7684\u573a\u666f\uff0c\u672c\u5730\u8d1f\u8f7d\u5f88\u8f7b<\/td>\n<\/tr>\n<tr>\n<td>\u5efa\u8bae<\/td>\n<td>4 \u6838 CPU\u30018GB \u5185\u5b58\u30012GB \u78c1\u76d8<\/td>\n<td>\u81ea\u6258\u7ba1\u7ed3\u679c\u670d\u52a1\u3001\u6216\u672c\u5730 provider\uff08Ollama \u4e4b\u7c7b\uff09\u540c\u673a\u8dd1\u65f6\u7559\u4f59\u91cf<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p>\u8fd9\u4e24\u6863\u662f\u6211\u6309\u88c5\u673a\u548c\u8dd1\u901a\u7684\u5b9e\u6d4b\u4f53\u611f\u4f30\u7684\uff0c\u4e0d\u662f\u5b98\u65b9\u6570\u5b57\uff0c\u5b98\u65b9\u6ca1\u7ed9\uff0c\u6211\u4e0d\u66ff\u5b83\u80cc\u4e66\u3002\u78c1\u76d8\u5360\u7528\u522b\u53ea\u770b\u5b83\u672c\u4f53\uff1a\u5168\u5c40\u5b89\u88c5\u8fde\u4f9d\u8d56\u5360\u6389\u5927\u51e0\u767e MB\uff0c\u8bc4\u4f30\u8dd1\u591a\u4e86\u672c\u5730 SQLite \u4e5f\u4f1a\u6162\u6162\u957f\uff0c\u4f46\u6b63\u5e38\u4f7f\u7528\u91cf\u7ea7\u8fdc\u8c08\u4e0d\u4e0a\u538b\u529b\u3002<\/p>\n<p>\u6211\u5b9e\u9645\u88c5\u7684\u8fc7\u7a0b\uff1a\u9ed8\u8ba4\u6e90\u4e0b\u8f7d\u6162\u5f97\u79bb\u8c31\uff0c\u6362 npmmirror \u955c\u50cf\u540e 606 \u4e2a\u5305 4 \u5206\u949f\u88c5\u5b8c\uff0c<code>promptfoo --version<\/code> \u56de\u663e <code>0.125.1<\/code>\uff082026 \u5e74 10 \u6708 25 \u65e5\u53d1\u5e03\u7684\u7248\u672c\uff09\uff0c\u4e00\u6b21\u6210\u529f\uff0c\u6ca1\u6709\u9047\u5230\u7f3a\u4f9d\u8d56\u6216\u8005\u7f16\u8bd1\u5931\u8d25\u7684\u95ee\u9898\u3002<\/p>\n<h2 id=\"key\">\u96f6 key \u79bb\u7ebf\u8dd1\u901a\u7b2c\u4e00\u4e2a\u77e9\u9635<\/h2>\n<p>\u88c5\u5b8c\u6211\u505a\u7684\u7b2c\u4e00\u4ef6\u4e8b\u4e0d\u662f\u914d OpenAI \u7684 key\uff0c\u800c\u662f\u9a8c\u8bc1\u4e00\u4e2a\u66f4\u6709\u610f\u601d\u7684\u70b9\uff1a<strong>\u5b83\u5185\u7f6e\u7684 <code>echo<\/code> provider \u53ef\u4ee5\u5b8c\u5168\u79bb\u7ebf\u8dd1\u901a\u6574\u6761\u8bc4\u4f30\u94fe\u8def<\/strong>\uff0c\u65ad\u8a00\u3001\u77e9\u9635\u3001\u7ed3\u679c\u843d\u5e93\u5168\u90fd\u80fd\u8d70\u4e00\u904d\u3002\u8fd9\u5bf9\u6211\u8fd9\u79cd\u5148\u9a8c\u8d27\u518d\u638f\u94b1\u7684\u4eba\u592a\u53cb\u597d\u4e86\u3002<\/p>\n<p>\u65b0\u5efa\u4e00\u4e2a\u76ee\u5f55\uff0c\u653e\u4e00\u4e2a <code>promptfooconfig.yaml<\/code>\uff1a<\/p>\n<pre><code class=\"language-yaml\">description: &quot;\u5ba2\u670d\u8bdd\u672f\u6539\u5199 - \u79bb\u7ebf\u5192\u70df\u6d4b\u8bd5&quot;\n\nprompts:\n  - &quot;\u8bf7\u628a\u4e0b\u9762\u8fd9\u53e5\u8bdd\u6539\u5199\u6210\u793c\u8c8c\u7684\u5ba2\u670d\u53e3\u543b\uff1a{{customer_text}}&quot;\n  - &quot;\u4f60\u662f\u8d44\u6df1\u5ba2\u670d\u3002\u8bf7\u628a\u4e0b\u9762\u8fd9\u53e5\u8bdd\u6539\u5199\u5f97\u66f4\u793c\u8c8c\uff0c\u53ea\u8f93\u51fa\u6539\u5199\u7ed3\u679c\uff1a{{customer_text}}&quot;\n\nproviders:\n  - echo\n\ntests:\n  - vars:\n      customer_text: \u4e1c\u897f\u574f\u4e86\u8d76\u7d27\u7ed9\u6211\u9000\u6b3e\n    assert:\n      - type: contains\n        value: \u9000\u6b3e\n      - type: not-contains\n        value: \u8d76\u7d27\n  - vars:\n      customer_text: \u4f60\u4eec\u8fd9\u4ec0\u4e48\u7834\u4ea7\u54c1\n    assert:\n      - type: javascript\n        value: output.length &gt; 0\n<\/code><\/pre>\n<p>\u7136\u540e\u8dd1 <code>promptfoo eval --no-cache<\/code>\uff0c\u7ec8\u7aef\u76f4\u63a5\u51fa\u77e9\u9635\uff1a<\/p>\n<pre><code>\u250c\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u252c\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u252c\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2510\n\u2502 customer_text            \u2502 [echo]   \u2502 [echo]   \u2502\n\u2502 \u4e1c\u897f\u574f\u4e86\u8d76\u7d27\u7ed9\u6211\u9000\u6b3e      \u2502 [FAIL]   \u2502 [FAIL]   \u2502\n\u2502 \u4f60\u4eec\u8fd9\u4ec0\u4e48\u7834\u4ea7\u54c1          \u2502 [PASS]   \u2502 [PASS]   \u2502\n\u2514\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2534\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2534\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2518\n\nResults:\n  \u2713 2 passed (50.00%)\n  \u2717 2 failed (50.00%)\n  0 errors (0%)\nDuration: 0s (concurrency: 4)\n<\/code><\/pre>\n<p>\u4e24\u4e2a FAIL \u4e0d\u662f\u51fa\u9519\u4e86\uff0c\u6070\u6070\u662f\u65ad\u8a00\u5728\u5e72\u6d3b\uff1a<code>echo<\/code> \u4f1a\u628a\u63d0\u793a\u8bcd\u539f\u6837\u56de\u663e\uff0c<code>customer_text<\/code> \u91cc\u7684\u300c\u8d76\u7d27\u300d\u81ea\u7136\u649e\u4e0a <code>not-contains<\/code>\u3002\u6362\u6210\u771f\u6a21\u578b\uff0c\u8fd9\u4e24\u4e2a\u683c\u5b50\u5c31\u662f\u5224\u65ad\u6a21\u578b\u6539\u5199\u8d28\u91cf\u7684\u5730\u65b9\u2014\u2014<code>contains<\/code> \u67e5\u5173\u952e\u8bcd\u5728\u4e0d\u5728\uff0c<code>javascript<\/code> \u5199\u81ea\u5b9a\u4e49\u89c4\u5219\uff0c\u60f3\u4e0a <code>llm-rubric<\/code> \u5c31\u914d\u4e00\u4e2a\u88c1\u5224\u6a21\u578b\u7684 key\uff0c\u914d\u7f6e\u957f\u8fd9\u6837\uff1a<\/p>\n<pre><code class=\"language-yaml\">defaultTest:\n  options:\n    provider: openai:gpt-4.1\ntests:\n  - vars:\n      customer_text: \u4e1c\u897f\u574f\u4e86\u8d76\u7d27\u7ed9\u6211\u9000\u6b3e\n    assert:\n      - type: llm-rubric\n        value: \u56de\u590d\u662f\u5426\u793c\u8c8c\u3001\u5b8c\u6574\uff0c\u4e14\u660e\u786e\u63d0\u5230\u9000\u6b3e\u5904\u7406\u65b9\u5f0f\n<\/code><\/pre>\n<p>\u8dd1\u5b8c\u63a5\u4e00\u53e5 <code>promptfoo view<\/code>\uff0c\u6d4f\u89c8\u5668\u91cc\u80fd\u770b\u5230\u6bcf\u6761\u7528\u4f8b\u7684\u539f\u59cb\u8f93\u51fa\u3001\u6253\u5206\u7406\u7531\u548c\u8017\u65f6\uff0c\u6bd4\u7ec8\u7aef\u8868\u683c\u76f4\u89c2\u5f97\u591a\u3002\u8f93\u5165\u3001\u6267\u884c\u3001\u65ad\u8a00\u3001\u62a5\u544a\uff0c\u4e00\u6761\u94fe\u8def\u5168\u901a\uff0c\u5168\u7a0b\u6ca1\u82b1\u4e00\u5206\u94b1\u3002<\/p>\n<h2 id=\"_4\">\u522b\u4eba\u7684\u5751\uff0c\u6211\u6361\u4e86\u516b\u4e2a<\/h2>\n<p>\u6211\u628a\u5b83\u7684 issue \u533a\u7ffb\u4e86\u4e2a\u5e95\u671d\u5929\u3002\u7eaf issue \u53ea\u6709 153 \u4e2a\u5f00\u653e\u7740\uff08\u5f88\u591a\u95ee\u9898\u76f4\u63a5\u8d70 PR \u5173\u6389\uff0c\u8fd9\u70b9\u793e\u533a\u7684\u5de5\u7a0b\u5316\u7a0b\u5ea6\u786e\u5b9e\u9ad8\uff09\uff0c\u5386\u53f2\u9ad8\u8ba8\u8bba\u5ea6\u7684\u95ee\u9898\u96c6\u4e2d\u5728\u51e0\u5757\uff1a<\/p>\n<p><strong>\u81ea\u6258\u7ba1\u662f\u4e00\u53f7\u91cd\u707e\u533a\u3002<\/strong> \u60f3\u81ea\u5df1\u642d\u4e00\u4e2a\u56e2\u961f\u5171\u4eab\u7684\u7ed3\u679c\u67e5\u770b\u670d\u52a1\uff1f\u5386\u53f2\u4e0a\u7684\u5751\u6392\u7740\u961f\uff1aDocker \u955c\u50cf\u7f3a\u6587\u4ef6\uff08<a href=\"https:\/\/github.com\/promptfoo\/promptfoo\/issues\/412\">#412<\/a>\uff09\u3001\u90e8\u7f72\u5230 K8s \u4e00\u8def\u6478\u7d22\uff08<a href=\"https:\/\/github.com\/promptfoo\/promptfoo\/issues\/1252\">#1252<\/a>\uff09\u3001\u81ea\u5b9a\u4e49\u5b50\u8def\u5f84\u4e0d\u652f\u6301\u8981\u81ea\u5df1\u6539\uff08<a href=\"https:\/\/github.com\/promptfoo\/promptfoo\/issues\/1611\">#1611<\/a>\uff09\u3001\u90e8\u7f72\u5230 Railway \u4e5f\u8981\u8e29\u4e00\u8f6e\uff08<a href=\"https:\/\/github.com\/promptfoo\/promptfoo\/issues\/595\">#595<\/a>\uff09\u3002\u7ed3\u8bba\u5f88\u76f4\u63a5\uff1a\u4e2a\u4eba\u7528\u6ca1\u4e8b\uff0c\u56e2\u961f\u534f\u4f5c\u7684\u81ea\u6258\u7ba1\u8981\u6709\u5fc3\u7406\u51c6\u5907\uff0c\u5b98\u65b9\u7684\u7cbe\u529b\u660e\u663e\u5728\u4e91\u670d\u52a1\u90a3\u8fb9\u3002<\/p>\n<p><strong>Web \u67e5\u770b\u5668\u6709\u8fc7\u7a33\u5b9a\u6027\u524d\u79d1\u3002<\/strong> <code>e.score.toFixed is not a function<\/code>\uff08<a href=\"https:\/\/github.com\/promptfoo\/promptfoo\/issues\/545\">#545<\/a>\uff09\u3001\u8bc4\u4f30\u9875\u6253\u5f00\u4e00\u7247\u7a7a\u767d\uff08<a href=\"https:\/\/github.com\/promptfoo\/promptfoo\/issues\/1873\">#1873<\/a>\uff09\uff0c\u90fd\u662f\u5386\u53f2\u4e0a\u771f\u5b9e\u5361\u8fc7\u4eba\u7684\u3002\u73b0\u5728\u7684\u7248\u672c\u597d\u4e86\u5f88\u591a\uff0c\u4f46\u7ed3\u679c\u9875\u6253\u4e0d\u5f00\u65f6\u5148\u60f3\u60f3\u662f\u4e0d\u662f\u65e7\u7248\u672c\u6570\u636e\u683c\u5f0f\u6ca1\u8fc1\u79fb\u3002<\/p>\n<p><strong>Provider \u9002\u914d\u662f\u957f\u5c3e\u95ee\u9898\u3002<\/strong> \u6bd4\u5982 Gemini \u67d0\u7248\u672c\u63a5\u53e3\u53d8\u52a8\u5bfc\u81f4 Google provider \u76f4\u63a5\u6302\uff08<a href=\"https:\/\/github.com\/promptfoo\/promptfoo\/issues\/1876\">#1876<\/a>\uff09\u3002\u6a21\u578b\u5382\u5546\u63a5\u53e3\u4e00\u53d8\uff0c\u5b83\u5c31\u5f97\u8ffd\uff0c\u7528\u975e\u4e3b\u6d41 provider \u524d\u5148\u641c\u4e00\u4e0b issue\uff0c\u8fd9\u662f\u7528\u4efb\u4f55\u805a\u5408\u7c7b\u5de5\u5177\u5171\u540c\u7684\u5bbf\u547d\u3002<\/p>\n<p><strong>\u6253\u5206\u65ad\u8a00\u8fd8\u6709\u65b0\u5751\u5728\u5192\u5934\u3002<\/strong> \u5f00\u653e issue \u91cc\u6700\u8fd1\u7684\u4e24\u6761\u90fd\u8ddf\u6253\u5206\u6709\u5173\uff1a\u88c1\u5224\u6a21\u578b\u628a JSON \u6392\u7248\u4e86\u5c31\u89e3\u6790\u5931\u8d25\uff08<a href=\"https:\/\/github.com\/promptfoo\/promptfoo\/issues\/12061\">#12061<\/a>\uff09\u3001<code>select-best<\/code> \u7684\u5224\u636e\u6ca1\u6709\u6b63\u786e\u5957\u7528\u6d4b\u8bd5\u53d8\u91cf\uff08<a href=\"https:\/\/github.com\/promptfoo\/promptfoo\/issues\/12077\">#12077<\/a>\uff09\u3002\u90fd\u662f\u6211\u5728\u5199\u8fd9\u7bc7\u7684\u65f6\u5019\u521a\u786e\u8ba4\u8fc7\u7684\u6d3b\u8dc3\u95ee\u9898\uff0c\u8fd8\u6ca1\u4fee\u3002\u5982\u679c\u91cd\u5ea6\u4f9d\u8d56\u6a21\u578b\u6253\u5206\uff0c\u8fd9\u4e24\u6761\u5efa\u8bae\u76ef\u7d27\u3002<\/p>\n<p><strong>\u7b2c\u4e09\u65b9\u53e3\u7891\u4e24\u9762\u90fd\u6709\u3002<\/strong> \u4e00\u4e2a\u7b2c\u4e09\u65b9\u70b9\u8bc4\u7ad9\u7ed9\u5b83 65 \u5206\uff0c\u5410\u69fd\u96c6\u4e2d\u5728 Web UI \u4e0d\u5982 LangSmith \u7cbe\u81f4\u3001\u7f3a\u9ad8\u7ea7\u5206\u6790\u9762\u677f\u3001\u6ca1\u6709\u5185\u7f6e\u7684\u63d0\u793a\u8bcd\u7248\u672c\u7ba1\u7406\uff08<a href=\"https:\/\/openalt.pro\/zh\/tools\/promptfoo-baf9cbcd\">OpenAlt<\/a>\uff09\uff1b\u53e6\u4e00\u4e2a\u7ed9 8.2 \u5206\uff0c\u63d0\u7684\u662f\u4f01\u4e1a\u7248\u4ef7\u683c\u53ea\u63a5\u53d7\u8be2\u4ef7\u3001YAML \u5bf9\u975e\u5de5\u7a0b\u5e08\u4e0d\u53cb\u597d\u3001\u624b\u673a\u7aef\u4f53\u9a8c\u57fa\u672c\u4e3a\u96f6\uff08<a href=\"https:\/\/topreviewed.ai\/products\/promptfoo\">TopReviewed<\/a>\uff09\u3002\u7ea2\u961f\u514d\u8d39\u989d\u5ea6\u300c\u6bcf\u6708 10,000 probes\u300d\u8fd9\u4e2a\u6570\u5b57\u53ea\u6709\u7b2c\u4e09\u65b9\u5728\u4f20\uff0c\u5b98\u65b9\u5b9a\u4ef7\u9875\u6211\u6ca1\u6838\u5230\u539f\u8bdd\uff0c\u59d1\u4e14\u542c\u4e4b\u3002\u4e2d\u6587\u793e\u533a\u8ba8\u8bba\u4e0d\u7b97\u591a\uff0c\u4e0a\u9762\u8fd9\u4e9b\u5c31\u662f\u6211\u7ffb issue \u52a0\u4ea4\u53c9\u641c\u5916\u7f51\u5f52\u7eb3\u7684\u5168\u90e8\uff0c\u6ca1\u6709\u4e00\u6761\u662f\u6211\u7f16\u7684\u3002<\/p>\n<h2 id=\"_5\">\u8981\u4e0d\u8981\u4e0a\uff0c\u6211\u7684\u8d26\u662f\u8fd9\u4e48\u7b97\u7684<\/h2>\n<p>\u6211\u7684\u7ed3\u8bba\uff1a<strong>\u53ea\u8981\u4f60\u7684\u63d0\u793a\u8bcd\u6539\u52a8\u9891\u7387\u8d85\u8fc7\u4e00\u5468\u4e00\u6b21\uff0c\u5b83\u5c31\u503c\u5f97\u4e0a\uff0c\u800c\u4e14\u5e94\u8be5\u76f4\u63a5\u8fdb CI\u3002<\/strong> \u5b83\u89e3\u51b3\u7684\u95ee\u9898\u4e0d\u662f\u300c\u5e2e\u4f60\u5199\u63d0\u793a\u8bcd\u300d\uff0c\u800c\u662f\u300c\u8ba9\u4f60\u6539\u63d0\u793a\u8bcd\u7684\u65f6\u5019\u5fc3\u91cc\u6709\u6570\u300d\u2014\u2014\u8fd9\u4ef6\u4e8b\u7684\u4ef7\u503c\u4f1a\u968f\u7740\u65f6\u95f4\u590d\u5229\uff0c\u6d4b\u8bd5\u77e9\u9635\u6512\u5f97\u8d8a\u4e45\u8d8a\u503c\u94b1\uff0c\u4e09\u4e2a\u6708\u540e\u4f60\u56de\u770b\u6bcf\u6b21\u6539\u52a8\u7684\u901a\u8fc7\u7387\u66f2\u7ebf\uff0c\u4f1a\u6bd4\u4efb\u4f55\u6587\u6863\u90fd\u80fd\u8bf4\u660e\u8fd9\u5957\u63d0\u793a\u8bcd\u7684\u6f14\u5316\u8def\u5f84\u3002<\/p>\n<p>\u4e0d\u9002\u5408\u8c01\uff1f\u4e24\u7c7b\u4eba\u3002\u4e00\u662f\u5076\u5c14\u624b\u6413\u51e0\u6761 prompt \u7684\u8f7b\u5ea6\u73a9\u5bb6\uff0c\u4e3a\u8fd9\u4e2a\u88c5 606 \u4e2a\u5305\u7eaf\u5c5e\u6740\u9e21\u7528\u725b\u5200\uff0c\u4e00\u4e2a <code>curl<\/code> \u811a\u672c\u52a0\u8089\u773c\u770b\u5c31\u591f\uff1b\u4e8c\u662f\u60f3\u8981\u6258\u7ba1\u5f0f\u5f00\u7bb1\u5373\u7528\u5e73\u53f0\u7684\u56e2\u961f\uff0c\u5b83\u7684\u81ea\u6258\u7ba1\u4f53\u9a8c\u548c\u5386\u53f2\u5751\u4f1a\u78e8\u6389\u4f60\u7684\u8010\u5fc3\uff0c\u975e\u5de5\u7a0b\u5e08\u89d2\u8272\uff08\u4ea7\u54c1\u3001\u8fd0\u8425\uff09\u5728 YAML \u9762\u524d\u57fa\u672c\u5bf8\u6b65\u96be\u884c\uff0c\u7ed3\u679c\u8fd8\u5f97\u5de5\u7a0b\u5e08\u4ee3\u770b\u3002<\/p>\n<p>\u66ff\u4ee3\u65b9\u6848\u70b9\u540d\u4e24\u4e2a\u3002<strong>DeepEval<\/strong>\uff0819,231 star\uff0cApache-2.0\uff09\uff1aPython \u539f\u751f\uff0cpytest \u98ce\u683c\uff0c\u9002\u5408\u6280\u672f\u6808\u5728 Python \u91cc\u7684\u56e2\u961f\uff0c\u6d4b\u8bd5\u5199\u8d77\u6765\u662f\u4ee3\u7801\u4e0d\u662f\u914d\u7f6e\uff0c\u7075\u6d3b\u5ea6\u9ad8\uff0c\u8ddf\u73b0\u6709\u5355\u5143\u6d4b\u8bd5\u4f4f\u5728\u540c\u4e00\u4e2a\u4ed3\u5e93\u91cc\u6beb\u65e0\u8fdd\u548c\uff0c\u4f46\u6ca1\u6709 promptfoo \u8fd9\u79cd\u4e00\u773c\u770b\u7a7f\u5168\u77e9\u9635\u7684 CLI \u4f53\u9a8c\uff0c\u7ea2\u961f\u6df1\u5ea6\u4e5f\u5dee\u4e00\u622a\u3002<strong>Giskard<\/strong>\uff1a\u4e3b\u6253\u6d4b\u8bd5\u52a0\u5408\u89c4\u5bf9\u9f50\uff0c\u5982\u679c\u4f60\u6709\u6b27\u76df\u6cd5\u6848\u4e4b\u7c7b\u7684\u786c\u5408\u89c4\u9700\u6c42\u5b83\u66f4\u5bf9\u53e3\uff0c\u7eaf\u6d4b\u8bd5\u4f53\u9a8c\u4e0a\u53cd\u800c\u4e0d\u53ca promptfoo \u987a\u624b\u3002\u6211\u7684\u9009\u62e9\u6807\u51c6\u5c31\u4e00\u6761\uff1a\u56e2\u961f\u4e3b\u8bed\u8a00\u662f TypeScript\u3001\u6216\u8005\u90e8\u7f72\u5f62\u6001\u662f CI \u6d41\u6c34\u7ebf\uff0c\u9009 promptfoo\uff1b\u4e3b\u8bed\u8a00\u662f Python\u3001\u6d4b\u8bd5\u60f3\u8ddf\u5355\u5143\u6d4b\u8bd5\u4f4f\u4e00\u8d77\uff0c\u9009 DeepEval\u3002<\/p>\n<p>\u6700\u540e\u8bf4\u53e5\u5b9e\u5728\u7684\u3002OpenAI \u6536\u7f16\u8fd9\u4e2a\u4e8b\uff0c\u77ed\u671f\u770b\u662f\u5229\u597d\u2014\u2014\u5b98\u65b9\u516c\u544a\u767d\u7eb8\u9ed1\u5b57\u627f\u8bfa\u7ee7\u7eed\u5efa\u8bbe\u5f00\u6e90\u9879\u76ee\uff0cMIT \u534f\u8bae\u4e00\u5206\u6ca1\u52a8\uff0c\u73b0\u5728\u4ed3\u5e93\u8fd8\u6709 26,708 \u4e2a star\uff0c\u6700\u65b0\u7248 0.125.1 \u5341\u6708\u5e95\u521a\u53d1\uff0c\u8fd1\u56db\u5468\u63d0\u4ea4 441 \u6b21\uff0c\u8282\u594f\u4e00\u70b9\u6ca1\u6162\u3002\u4f46\u957f\u671f\u561b\uff0c\u4f01\u4e1a\u80fd\u529b\u5f80 Frontier \u91cc\u5e76\u3001\u5f00\u6e90\u7248\u8ddf\u4e91\u7248\u7684\u5dee\u8ddd\u4f1a\u4e0d\u4f1a\u8d8a\u62c9\u8d8a\u5927\uff0c\u8fd9\u624d\u662f\u8981\u76ef\u7740\u7684\u7ebf\u3002\u53cd\u6b63\u4ee3\u7801\u5728 MIT \u624b\u91cc\uff0c\u771f\u5230\u90a3\u5929\uff0cfork \u4e00\u4efd\u81ea\u5df1\u517b\u4e5f\u4e0d\u662f\u4e0d\u884c\u2014\u2014\u8fd9\u4e5f\u662f\u6211\u6562\u73b0\u5728\u5c31\u628a\u5b83\u5199\u8fdb\u6d41\u6c34\u7ebf\u7684\u5e95\u6c14\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u6539\u4e00\u7248\u7cfb\u7edf\u63d0\u793a\u8bcd\u5c31\u4e0a\u7ebf\uff0c\u7ffb\u8f66\u4e86\u624d\u53d1\u73b0\u6ca1\u6709\u57fa\u7ebf\u3002promptfoo \u7528 YAML \u628a\u6d4b\u8bd5\u77e9\u9635\u56fa\u5316\u8fdb CI\uff0c\u6211\u88c5\u5b8c\u7528\u5185\u7f6e echo provider \u96f6 key \u8dd1\u901a\u4e86\u7b2c\u4e00\u5f20\u77e9\u9635\u3002\u88ab OpenAI \u6536\u7f16\u540e\u5b83\u4ecd\u7ef4\u6301 MIT \u5f00\u6e90\uff0c26,708 \u4e2a star\uff0c\u8fd1\u56db\u5468 441 \u6b21\u63d0\u4ea4\u3002<\/p>\n","protected":false},"author":1,"featured_media":10021532,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_jetpack_memberships_contains_paid_content":false,"footnotes":""},"categories":[4],"tags":[],"class_list":["post-10021533","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-share"],"jetpack_sharing_enabled":true,"jetpack_featured_media_url":"https:\/\/itmax.cn\/wp-content\/uploads\/2026\/11\/promptfoo-cover.png","_links":{"self":[{"href":"https:\/\/itmax.cn\/index.php?rest_route=\/wp\/v2\/posts\/10021533","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/itmax.cn\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/itmax.cn\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/itmax.cn\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/itmax.cn\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=10021533"}],"version-history":[{"count":1,"href":"https:\/\/itmax.cn\/index.php?rest_route=\/wp\/v2\/posts\/10021533\/revisions"}],"predecessor-version":[{"id":10021534,"href":"https:\/\/itmax.cn\/index.php?rest_route=\/wp\/v2\/posts\/10021533\/revisions\/10021534"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/itmax.cn\/index.php?rest_route=\/wp\/v2\/media\/10021532"}],"wp:attachment":[{"href":"https:\/\/itmax.cn\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=10021533"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/itmax.cn\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=10021533"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/itmax.cn\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=10021533"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}