{"id":5479,"date":"2024-02-02T23:37:56","date_gmt":"2024-02-02T15:37:56","guid":{"rendered":"https:\/\/aict.nkust.edu.tw\/digitrans\/?p=5479"},"modified":"2024-12-13T20:35:27","modified_gmt":"2024-12-13T12:35:27","slug":"langchain-%e6%80%8e%e9%ba%bc%e7%8e%a9%ef%bc%9f%e5%85%a5%e9%96%80%e6%95%99%e5%ad%b8%e7%af%87","status":"publish","type":"post","link":"https:\/\/aict.nkust.edu.tw\/digitrans\/?p=5479","title":{"rendered":"LangChain \u600e\u9ebc\u73a9\uff1f\u5165\u9580\u6559\u5b78\u7bc7"},"content":{"rendered":"\n<p>2024-02-02 | Amo Chen<\/p>\n\n\n\n<p>AI \u6642\u4ee3\uff0c\u6253\u4e0d\u8d0f\u5c31\u52a0\u5165\u5b83\uff01<\/p>\n\n\n\n<p>\u6240\u4ee5\u500b\u4eba\u8a8d\u70ba\u5b78\u6703 LangChain \u4e4b\u985e\u7684\u6846\u67b6\uff0c\u5728\u672a\u4f86\u53ef\u80fd\u6703\u662f\u6bcf\u500b\u7a0b\u5f0f\u8a2d\u8a08\u5e2b\u4e0d\u53ef\u6216\u7f3a\u7684\u6280\u8853\uff0c\u4e5f\u5c31\u662f\u8aaa\u9664\u4e86\u5beb\u7a0b\u5f0f\u4e4b\u5916\uff0c\u4f60\u53ef\u80fd\u9084\u9700\u8981\u7528 LangChain \u4e4b\u985e\u7684\u6846\u67b6\u505a\u51fa\u9069\u5408\u81ea\u5df1\u7684\u5de5\u5177\uff0c\u5e6b\u52a9\u63d0\u5347\u6548\u7387\u8207\u751f\u7522\u529b\uff0c\u85c9\u6b64\u589e\u52a0\u81ea\u8eab\u7684\u8077\u5834\u512a\u52e2\u3002<\/p>\n\n\n\n<p>\u672c\u6587\u5c07\u4ecb\u7d39 LangChain \u7d50\u5408 llama \u8a9e\u8a00\u6a21\u578b\u5982\u4f55\u4f7f\u7528\u7684\u5165\u9580\u6559\u5b78\u3002<\/p>\n\n\n\n<p>p.s. \u4f7f\u7528\u958b\u6e90\u8a9e\u8a00\u6a21\u578b\u7684 llama \u7684\u597d\u8655\u5728\u65bc\u4e0d\u7528\u4ed8\u8cbb\uff0c\u8f38\u51fa\u54c1\u8cea\u4e5f\u6709\u4e00\u5b9a\u4fdd\u8b49<\/p>\n\n\n\n<h3 class=\"wp-block-heading\" id=\"\u672c\u6587\u74b0\u5883\">\u672c\u6587\u74b0\u5883&nbsp;<a href=\"https:\/\/myapollo.com.tw\/blog\/langchain-tutorial-get-started\/#%e6%9c%ac%e6%96%87%e7%92%b0%e5%a2%83\"><\/a><\/h3>\n\n\n\n<ul class=\"wp-block-list\">\n<li>macOS<\/li>\n\n\n\n<li>Python 3<\/li>\n\n\n\n<li>LangChain<\/li>\n\n\n\n<li>Ollama<\/li>\n<\/ul>\n\n\n\n<p><code>$ pip install langchain<\/code><\/p>\n\n\n\n<p>\u672c\u6587\u9700\u8981 Ollama \u6307\u4ee4\u8207 Meta \u516c\u53f8\u63d0\u4f9b\u7684\u00a0<code>llama2<\/code>\u00a0\u6a21\u578b(model)\uff0c<code>ollama<\/code>\u00a0\u6307\u4ee4\u8acb\u81f3\u00a0Ollama\u00a0\u5b98\u65b9\u9801\u9762\u4e0b\u8f09\u5b89\u88dd\uff0c\u5b89\u88dd\u5b8c\u6210\u4e4b\u5f8c\u5373\u53ef\u57f7\u884c\u6307\u4ee4\u5b89\u88dd\u00a0<code>llama2<\/code>\u00a0\u6a21\u578b\uff0c\u5176\u5b89\u88dd\u6307\u4ee4\u5982\u4e0b\uff1a<code>$ ollama run llama2<\/code><\/p>\n\n\n\n<p>p.s. \u57f7\u884c\u4e0a\u8ff0\u6a21\u578b\u9700\u8981\u81f3\u5c11 8GB \u8a18\u61b6\u9ad4\uff0c 3.8G \u786c\u789f\u7a7a\u9593<\/p>\n\n\n\n<h3 class=\"wp-block-heading\" id=\"langchain-\u7c21\u4ecb\">LangChain \u7c21\u4ecb&nbsp;<a href=\"https:\/\/myapollo.com.tw\/blog\/langchain-tutorial-get-started\/#langchain-%e7%b0%a1%e4%bb%8b\"><\/a><\/h3>\n\n\n\n<p>LangChain\u00a0\u662f 1 \u5957\u5c08\u9580\u7528\u4f86\u958b\u767c\u8a9e\u8a00\u6a21\u578b(Language Model)\u76f8\u95dc\u61c9\u7528\u7684\u6846\u67b6(framework), \u7c21\u55ae\u4f86\u8aaa\u53ef\u4ee5\u8b93\u958b\u767c\u8005\u6574\u5408\u4e0d\u540c\u7684\u8a9e\u8a00\u6a21\u578b\uff0c\u958b\u767c\u50cf ChatGPT, \u804a\u5929\u6a5f\u5668\u4eba\u3001\u6578\u4f4d\u52a9\u7406\u7b49\u7b49\u61c9\u7528\u7a0b\u5f0f\u3002<\/p>\n\n\n\n<p>\u76f8\u4fe1\u5f88\u591a\u4eba\u5c0d LangChain \u9019\u500b\u540d\u5b57\u611f\u5230\u7591\u60d1\u3002<\/p>\n\n\n\n<p>\u5148\u4ecb\u7d39 Lang \u7684\u90e8\u5206\uff0c Lang \u4ee3\u8868 Language Model, \u4e5f\u662f\u547c\u61c9 LangChain \u5c08\u9580\u7528\u4ee5\u958b\u767c\u8a9e\u8a00\u6a21\u578b\u76f8\u95dc\u61c9\u7528\u7684\u7528\u9014\u3002<\/p>\n\n\n\n<p>Chain \u7684\u90e8\u5206\u5247\u662f LangChain \u7684\u529f\u80fd\uff0c\u5b83\u5c07\u5927\u8a9e\u8a00\u6a21\u578b\u7684\u61c9\u7528\u8996\u70ba 1 \u500b Chain, \u8209 ChatGPT \u804a\u5929\u61c9\u7528\u70ba\u4f8b\uff0c\u5b83\u81f3\u5c11\u9700\u8981\u7d50\u5408\u63d0\u793a\u8a5e(prompt)\u8207 1 \u500b\u5927\u8a9e\u8a00\u6a21\u578b\uff0c\u904e\u7a0b\u5f88\u50cf\u662f\u628a\u63d0\u793a\u8a5e\u4e1f\u7d66\u5927\u8a9e\u8a00\u6a21\u578b\u8655\u7406\uff0c\u4f8b\u5982\uff1a<code>\u63d0\u793a\u8a5e\uff1a\"\u4f60\u662f\u8cc7\u6df1 Python \u5de5\u7a0b\u5e2b\uff0c\u8acb\u5e6b\u6211\u5beb\u51fa 1 \u500b Python Hello World \u7bc4\u4f8b\" \u2b63\u2b63 \u5927\u8a9e\u8a00\u6a21\u578b<\/code><\/p>\n\n\n\n<p>\u9019\u6a23\u7684\u7d44\u5408\u884c\u70ba\u5728 LangChain \u88ab\u62bd\u8c61\u5316\u70ba 1 \u500b Chain, \u7a0b\u5f0f\u78bc\u7247\u6bb5\u5982\u4e0b\uff1a<code><strong>from<\/strong> langchain_openai <strong>import<\/strong> ChatOpenAI <strong>from<\/strong> langchain_core.prompts <strong>import<\/strong> ChatPromptTemplate llm = ChatOpenAI(openai_api_key=\"...\u7565...\") prompt = ChatPromptTemplate.from_messages([ (\"system\", \"\u4f60\u662f\u8cc7\u6df1 Python \u5de5\u7a0b\u5e2b\"), (\"user\", \"\u8acb\u5e6b\u6211\u5beb\u51fa 1 \u500b Python Hello World \u7bc4\u4f8b\") ]) chain = prompt | llm<\/code><\/p>\n\n\n\n<p>\u4e0a\u8ff0\u7bc4\u4f8b\u4e2d\u7684\u00a0<code>prompt | llm<\/code>\u00a0\u5c31\u662f 1 \u500b\u6700\u7c21\u55ae\u7684 Chain, \u6240\u4ee5\u958b\u767c LangChain \u61c9\u7528\u5176\u5be6\u591a\u534a\u662f\u5728\u7d44\u5408\/\u8655\u7406 Chain \u7684\u5de5\u4f5c\uff0c\u8b93 Chain \u80fd\u5920\u85c9\u7531\u8a9e\u8a00\u6a21\u578b\u7684\u529b\u91cf\u63d0\u4f9b\u670d\u52d9\u3002\u5662\uff0c\u9019\u500b\u7d44\u5408\u65b9\u5f0f\u7a31\u70ba\u00a0LCEL, LangChain Expression Language\u00a0\u3002<\/p>\n\n\n\n<p>\u76ee\u524d LangChain \u5305\u542b\u591a\u500b\u90e8\u4efd\uff1a<\/p>\n\n\n\n<ol class=\"wp-block-list\">\n<li>LangChain Libraries LangChain \u76f8\u95dc\u7684\u529f\u80fd\u8207 API \u7b49\u7b49\uff0c\u652f\u63f4 Python \u8207 JavaScript<\/li>\n\n\n\n<li>LangChain Templates\u00a0\u4e00\u4e9b\u5e38\u898b\u7684 LangChain \u6a23\u677f(templates)\u96c6\u5408\uff0c\u8b93\u958b\u767c\u8005\u53ef\u4ee5\u5957\u7528\uff0c\u7701\u53bb\u91cd\u65b0\u958b\u767c\u7684\u6210\u672c<\/li>\n\n\n\n<li>LangServe\u00a0\u53ef\u4ee5\u5c07 LangChain \u7684 chain \u958b\u767c\u6210 REST API \u7684\u90e8\u4ef6<\/li>\n\n\n\n<li>LangSmith\u00a0\u53ef\u4ee5\u8b93\u958b\u767c\u8005\u5c0d\u8a9e\u8a00\u6a21\u578b\u9032\u884c\u6e2c\u8a66\u3001\u8a55\u4f30\u3001\u9664\u932f\u3001\u76e3\u63a7\u7b49\u884c\u70ba\u7684 1 \u500b\u5e73\u53f0<\/li>\n<\/ol>\n\n\n<div class=\"wp-block-image\">\n<figure class=\"aligncenter size-full\"><img data-recalc-dims=\"1\" decoding=\"async\" src=\"https:\/\/i0.wp.com\/aict.nkust.edu.tw\/digitrans\/wp-content\/uploads\/2024\/03\/image-2.png?w=640&#038;ssl=1\" alt=\"\" class=\"wp-image-5482\"\/><\/figure>\n<\/div>\n\n\n<p class=\"has-text-align-center\">\u00a0\u5716\u7247\u5f15\u7528\u4f86\u6e90 &#8211; LangCain \u5b98\u7db2<\/p>\n\n\n\n<p>\u672c\u6587\u5c07\u6703\u8457\u91cd\u5728 LangChain Libraries \u7684\u90e8\u5206\u3002<\/p>\n\n\n\n<h3 class=\"wp-block-heading\" id=\"\u725b\u5200\u5c0f\u8a66--\u554f\u5019\u804a\u5929\">\u725b\u5200\u5c0f\u8a66 \u2014 \u554f\u5019\u3001\u804a\u5929&nbsp;<a href=\"https:\/\/myapollo.com.tw\/blog\/langchain-tutorial-get-started\/#%e7%89%9b%e5%88%80%e5%b0%8f%e8%a9%a6--%e5%95%8f%e5%80%99%e8%81%8a%e5%a4%a9\"><\/a><\/h3>\n\n\n\n<p>\u5c0d LangChain \u6709\u57fa\u672c\u8a8d\u8b58\u4e4b\u5f8c\uff0c\u63a5\u8457\u9032\u884c\u7b2c 1 \u500b\u7bc4\u4f8b\uff0c\u8b93\u6211\u5011\u8ddf\u8a9e\u8a00\u6a21\u578b\u804a\u5929\u5427\uff01<\/p>\n\n\n\n<p>\u4ee5\u4e0b\u662f\u7b2c 1 \u500b\u7bc4\u4f8b\uff0c\u8a72\u7a0b\u5f0f\u78bc\u4ee5\u7c21\u55ae\u4e00\u53e5\u554f\u5019&nbsp;<code>Hi, how are you today?<\/code>&nbsp;\u8207\u8a9e\u8a00\u6a21\u578b&nbsp;<code>Ollama<\/code>&nbsp;\u9032\u884c\u4e92\u52d5\uff1a<code><strong>from<\/strong> langchain_community.llms <strong>import<\/strong> Ollama llm = Ollama(model='llama2') print(llm.invoke('Hi, how are you today?'))<\/code><\/p>\n\n\n\n<p>\u4e0a\u8ff0\u7a0b\u5f0f\u57f7\u884c\u4e4b\u5f8c\uff0c\u5f88\u9ad8\u6a5f\u7387\u6703\u56de\u61c9\u985e\u4f3c\u4ee5\u4e0b\u7d50\u679c\uff1a<code>I'm just an AI, I don't have emotions <strong>or<\/strong> feelings, so I can't really experience the world in the same way that humans do. However, I'm here to help answer any questions you may have <strong>or<\/strong> provide assistance <strong>with<\/strong> any tasks you may have, so feel free to ask me anything!<\/code><\/p>\n\n\n\n<p>\u610f\u601d\u662f\u300c\u6211\u662f AI, \u6211\u6c92\u6709\u4efb\u4f55\u60c5\u7dd2\u8207\u611f\u89ba\uff0c\u6211\u7121\u6cd5\u8207\u4eba\u985e\u822c\u9ad4\u9a57\u4e16\u754c\u3002\u4e0d\u904e\uff0c\u6211\u9084\u662f\u80fd\u5920\u56de\u7b54\u4f60\u7684\u4efb\u4f55\u554f\u984c\u8207\u63d0\u4f9b\u5e6b\u52a9\uff0c\u6b61\u8fce\u554f\u6211\u4efb\u4f55\u554f\u984c\u3002\u300d<\/p>\n\n\n\n<p>\u7c21\u55ae 3 \u884c\u7a0b\u5f0f\u78bc\uff0c\u5c31\u80fd\u5920\u8207\u8a9e\u8a00\u6a21\u578b\u9032\u884c\u4e92\u52d5\uff0c\u8db3\u898b LangChain \u5f88\u5bb9\u6613\u4f7f\u7528\u3002<\/p>\n\n\n\n<p>\u9019 3 \u884c\u7a0b\u5f0f\u78bc\u4e3b\u8981\u662f import&nbsp;<code>Ollama<\/code>&nbsp;\u6a21\u7d44\uff0c\u4e26\u5229\u7528&nbsp;<code>Ollama<\/code>&nbsp;\u6a21\u7d44\u8f09\u5165&nbsp;<code>llama2<\/code>&nbsp;\u8a9e\u8a00\u6a21\u578b\uff0c\u6700\u5f8c\u4f7f\u7528\u8a9e\u8a00\u6a21\u578b\u7684&nbsp;<code>invoke()<\/code>&nbsp;\u65b9\u6cd5\uff0c\u50b3\u5165\u6211\u5011\u60f3\u8207\u4e4b\u4e92\u52d5\u7684\u5b57\u4e32\uff0c\u6700\u5f8c\u5c07\u7d50\u679c\u5217\u5370\u51fa\u4f86\u3002<\/p>\n\n\n\n<h3 class=\"wp-block-heading\" id=\"\u8b8a\u6210-chain-\u5427\">\u8b8a\u6210 Chain \u5427\uff01&nbsp;<a href=\"https:\/\/myapollo.com.tw\/blog\/langchain-tutorial-get-started\/#%e8%ae%8a%e6%88%90-chain-%e5%90%a7\"><\/a><\/h3>\n\n\n\n<p>\u5b78\u7fd2 LangChain \u6700\u91cd\u8981\u7684\u5c31\u662f\u5c07\u8a9e\u8a00\u6a21\u578b\u7684\u64cd\u4f5c\u8b8a\u6210 Chain \u3002<\/p>\n\n\n\n<p>Chain \u7684\u7d44\u5408\u9700\u8981\u4f7f\u7528&nbsp;<code>|<\/code>&nbsp;\u904b\u7b97\u5b50\uff0c\u8ddf Linux \u6307\u4ee4\u7684\u7ba1\u9053(pipe)\u7b26\u865f\u5f88\u50cf\uff0c\u90fd\u662f\u5c07\u524d 1 \u500b\u6307\u4ee4\u7684\u7d50\u679c\u900f\u904e\u7ba1\u9053\u50b3\u7d66\u4e0b 1 \u500b\u6307\u4ee4\u7684\u4f5c\u7528\u3002<\/p>\n\n\n\n<p>\u56e0\u6b64\u6700\u7c21\u55ae\u7684 Chain \u662f\u4ee5\u4e0b\u5f62\u5f0f\uff1a<code>chain = prompt | llm<\/code><\/p>\n\n\n\n<p>\u63a5\u4e0b\u4f86\uff0c\u6211\u5011\u5c07\u7b2c 1 \u500b\u7bc4\u4f8b\u8f49\u70ba Chain \u5f62\u5f0f\uff1a<code><strong>from<\/strong> langchain_community.llms <strong>import<\/strong> Ollama <strong>from<\/strong> langchain_core.prompts <strong>import<\/strong> ChatPromptTemplate llm = Ollama(model='llama2') prompt = ChatPromptTemplate.from_messages([ (\"user\", \"{input}\"), ]) chain = prompt | llm print(chain.invoke({\"input\": \"Hi, how are you today?\"}))<\/code><\/p>\n\n\n\n<p>\u4e0a\u8ff0\u7bc4\u4f8b\u8207\u7b2c 1 \u500b\u7bc4\u4f8b\u6700\u5927\u7684\u4e0d\u540c\uff0c\u662f\u6211\u5011\u4f7f\u7528\u00a0<code>langchain_core.prompts<\/code>\u00a0\u6a21\u7d44\u4e2d\u7684\u00a0<code>ChatPromptTemplate<\/code>\u00a0\u5b9a\u7fa9 prompt, \u9019\u7d44 prompt \u53ea\u63a5\u53d7 1 \u500b\u4f7f\u7528\u8005\u8f38\u5165\u7684\u53c3\u6578\u00a0<code>{input}<\/code>\u00a0(<code>input<\/code>\u00a0\u53ef\u4ee5\u96a8\u559c\u597d\u547d\u540d\uff0c\u6b64\u8655\u662f Python format string \u7684\u7528\u6cd5\uff0c\u5982\u679c\u4e0d\u4e86\u89e3\u7684\u8a71\u8acb\u53c3\u8003\u00a0\u5236\u9738 Python f-string \u5404\u7a2e\u683c\u5f0f\u4f7f\u7528\u65b9\u6cd5)\uff0c\u63a5\u8457\u6211\u5011\u5c07\u00a0<code>prompt<\/code>\u00a0\u8207\u00a0<code>llm<\/code>\u00a0\u900f\u904e\u00a0<code>|<\/code>\u00a0\u904b\u7b97\u5b50\u7d44\u5408\u6210\u00a0<code>chain<\/code>\u00a0\uff0c\u6700\u5f8c\u900f\u904e\u00a0<code>chain<\/code>\u00a0\u7684\u00a0<code>invoke<\/code>\u00a0\u65b9\u6cd5\uff0c\u5c07\u00a0<code>input<\/code>\u00a0\u53c3\u6578\u4ee3\u5165\u3002<\/p>\n\n\n\n<p>\u4e0a\u8ff0\u7bc4\u4f8b\u7684\u57f7\u884c\u7d50\u679c\u5982\u4e0b\uff0c\u773c\u5c16\u7684\u4eba\u61c9\u8a72\u6703\u767c\u73fe\u591a\u51fa&nbsp;<code>Assistant:<\/code>&nbsp;\u5b57\u4e32\uff1a<code>Assistant: Hello! I'm just an AI, I don't have feelings <strong>or<\/strong> emotions like humans do, so I can't really experience the world in the same way that you do. However, I'm here to help you <strong>with<\/strong> any questions <strong>or<\/strong> tasks you may have, so feel free to ask me anything!<\/code><\/p>\n\n\n\n<p>\u5f88\u591a\u6642\u5019\uff0c\u6211\u5011\u53ef\u80fd\u9700\u8981\u5c0d\u8a9e\u8a00\u6a21\u578b\u7684\u8f38\u51fa\u505a\u984d\u5916\u7684\u683c\u5f0f\u8655\u7406\uff0c\u4f8b\u5982\u53bb\u6389&nbsp;<code>Assistant:<\/code>&nbsp;\u6216\u8005\u8f49\u70ba\u5176\u4ed6\u683c\u5f0f\u7b49\u7b49\uff0c\u9019\u500b\u4e5f\u80fd\u5920\u900f\u904e\u5728 Chain \u7684\u5c3e\u7aef\u52a0\u5165\u8655\u7406\u8f38\u51fa\u7684\u65b9\u5f0f\u9054\u6210\uff1a<code>chain = prompt | llm | &lt;\u8655\u7406\u8f38\u51fa\u7684\u7a0b\u5f0f&gt;<\/code><\/p>\n\n\n\n<p><code>&lt;\u8655\u7406\u8f38\u51fa\u7684\u7a0b\u5f0f><\/code>\u00a0\u5728 LangChain \u7a31\u70ba output parser, \u540c\u6642 LangChain \u4e5f\u6709\u5167\u5efa 1 \u500b class \u7a31\u70ba\u00a0StrOutputParser\u00a0\u53ef\u4ee5\u4f7f\u7528(\u66f4\u591a\u6a21\u7d44\u8acb\u53c3\u8003\u00a0langchain_core.output_parsers\u00a0)\uff0c\u56e0\u6b64\u53ef\u4ee5\u900f\u904e\u7e7c\u627f\u7684\u65b9\u5f0f\u505a\u5ba2\u88fd\u5316\u7684\u8f38\u51fa\u8655\u7406\uff0c\u6700\u7c21\u55ae\u7684\u5ba2\u88fd\u5316\u7684\u7a0b\u5f0f\u78bc\u5982\u4e0b\u6240\u793a\uff1a<code><strong>class<\/strong> MyOutputParser(StrOutputParser): <strong>def<\/strong> parse(self, text): <strong>return<\/strong> text.replace('Assistant:', '')<\/code><\/p>\n\n\n\n<p>\u4e0a\u8ff0\u7bc4\u4f8b\u900f\u904e override&nbsp;<code>parse()<\/code>&nbsp;\u65b9\u6cd5\uff0c\u9054\u5230\u53bb\u6389&nbsp;<code>Assistant:<\/code>&nbsp;\u5b57\u4e32\u7684\u76ee\u7684\u3002<\/p>\n\n\n\n<p>p.s. LangChain \u7684 output parser \u4e5f\u6709\u652f\u63f4 async IO, \u6709\u8208\u8da3\u7684\u8a71\u53ef\u4ee5\u7ffb\u95b1\u6587\u4ef6<\/p>\n\n\n\n<p>\u63a5\u8457\uff0c\u53ea\u8981\u628a\u4e0a\u8ff0\u7684 output parser \u4e32\u5728 chain \u7684\u5c3e\u7aef\u5373\u53ef\uff1a<code><strong>from<\/strong> langchain_community.llms <strong>import<\/strong> Ollama <strong>from<\/strong> langchain_core.output_parsers <strong>import<\/strong> StrOutputParser <strong>from<\/strong> langchain_core.prompts <strong>import<\/strong> ChatPromptTemplate <strong>class<\/strong> MyOutputParser(StrOutputParser): <strong>def<\/strong> parse(self, text): <strong>return<\/strong> text.replace('Assistant: ', '') output_parser = MyOutputParser() llm = Ollama(model='llama2') prompt = ChatPromptTemplate.from_messages([ (\"user\", \"{input}\"), ]) chain = prompt | llm | output_parser print(chain.invoke({\"input\": \"Hi, how are you today?\"}))<\/code><\/p>\n\n\n\n<p>\u4e0a\u8ff0\u7684\u57f7\u884c\u7d50\u679c\u6703\u767c\u73fe&nbsp;<code>Assistant:<\/code>&nbsp;\u4e0d\u898b\u4e86\uff1a<code>Hello! I'm just an AI, I don't have feelings <strong>or<\/strong> emotions like humans do, so I can't really experience the world in the same way that you do. However, I'm here to help you <strong>with<\/strong> any questions <strong>or<\/strong> tasks you may have, so feel free to ask me anything!<\/code><\/p>\n\n\n\n<p>\u76ee\u524d\u70ba\u6b62\uff0c\u5927\u5bb6\u61c9\u8a72\u80fd\u5920\u7a0d\u5fae\u9ad4\u6703\u5230 LangChain \u7684\u529b\u91cf\u3002<\/p>\n\n\n\n<h3 class=\"wp-block-heading\" id=\"\u8ce6\u4e88\u8a9e\u8a00\u6a21\u578b\u7cfb\u7d71\u6307\u4ee4\">\u8ce6\u4e88\u8a9e\u8a00\u6a21\u578b\u7cfb\u7d71\u6307\u4ee4&nbsp;<a href=\"https:\/\/myapollo.com.tw\/blog\/langchain-tutorial-get-started\/#%e8%b3%a6%e4%ba%88%e8%aa%9e%e8%a8%80%e6%a8%a1%e5%9e%8b%e7%b3%bb%e7%b5%b1%e6%8c%87%e4%bb%a4\"><\/a><\/h3>\n\n\n\n<p>\u5b78\u6703 Chain \u7684\u7528\u6cd5\u4e4b\u5f8c\uff0c\u5c31\u53ef\u4ee5\u70ba\u8a9e\u8a00\u6a21\u578b\u505a\u4e9b\u52a0\u503c\u61c9\u7528\uff0c\u4f8b\u5982\u8ce6\u4e88\u5b83\u7cfb\u7d71\u6307\u4ee4\uff0c\u4f8b\u5982\u8acb\u5b83\u626e\u6f14 1 \u500b\u5177\u6709 SEO \u77e5\u8b58\u7684\u5c08\u696d\u5beb\u624b\uff0c\u5e6b\u52a9\u6211\u5011\u5beb\u4e00\u4e9b\u6587\u7ae0\u3002<\/p>\n\n\n\n<p>\u8981\u8ce6\u4e88\u5b83\u7cfb\u7d71\u6307\u4ee4\uff0c\u5c31\u9700\u8981\u5728 prompt \u8457\u624b\uff0c\u52a0\u4e0a\u7cfb\u7d71\u6307\u4ee4\uff0c\u7cfb\u7d71\u6307\u4ee4\u7684\u683c\u5f0f\u70ba\uff1a<code>(\"system\", \"\u7cfb\u7d71\u6307\u4ee4\")<\/code><\/p>\n\n\n\n<p>\u4ee5\u525b\u525b\u7684 SEO \u5c08\u696d\u5beb\u624b\u70ba\u4f8b\uff0c\u7cfb\u7d71\u6307\u4ee4\u53ef\u4ee5\u5beb\u6210\uff1a<code>(\"system\", \"You are a content manager with extensive SEO knowledge. Your task is to write an article based on a given title.\")<\/code><\/p>\n\n\n\n<p>\u63a5\u8457\uff0c\u5c07\u7cfb\u7d71\u6307\u4ee4\u653e\u5230 prompt \u4e2d\uff1a<code><strong>from<\/strong> langchain_community.llms <strong>import<\/strong> Ollama <strong>from<\/strong> langchain_core.output_parsers <strong>import<\/strong> StrOutputParser <strong>from<\/strong> langchain_core.prompts <strong>import<\/strong> ChatPromptTemplate <strong>class<\/strong> MyOutputParser(StrOutputParser): <strong>def<\/strong> parse(self, text): <strong>return<\/strong> text.replace('Assistant:', '') output_parser = MyOutputParser() llm = Ollama(model='llama2') prompt = ChatPromptTemplate.from_messages([ (\"system\", \"You are a content manager with extensive SEO knowledge. Your task is to write an article based on a given title.\"), (\"user\", \"{input}\"), ]) chain = prompt | llm | output_parser print(chain.invoke({\"input\": \"How does software change the world?\"}))<\/code><\/p>\n\n\n\n<p>\u4e0a\u8ff0\u7bc4\u4f8b\u57f7\u884c\u7d50\u679c\u6703\u7522\u751f 1 \u7bc7\u82f1\u6587\u6587\u7ae0\uff08\u57f7\u884c\u6642\u9593\u53ef\u80fd\u6703\u6709\u9ede\u4e45\uff09\uff0c\u5c31\u4e0d\u7279\u5730\u5c07\u7d50\u679c\u653e\u5728\u6587\u7ae0\u4e2d\u3002<\/p>\n\n\n\n<p>\u81f3\u6b64\uff0c\u4f60\u5df2\u7d93\u5b78\u6703\u4f7f\u7528\u7cfb\u7d71\u6307\u4ee4\uff0c\u53ef\u4ee5\u958b\u59cb\u6253\u9020\u9069\u5408\u81ea\u5df1\u7684\u5de5\u5177\u4e86\uff01<\/p>\n\n\n\n<h3 class=\"wp-block-heading\" id=\"\u7e3d\u7d50\">\u7e3d\u7d50&nbsp;<a href=\"https:\/\/myapollo.com.tw\/blog\/langchain-tutorial-get-started\/#%e7%b8%bd%e7%b5%90\"><\/a><\/h3>\n\n\n\n<p>\u622a\u81f3\u76ee\u524d\u70ba\u6b62\uff0c\u672c\u6587\u7684\u5167\u5bb9\u8b93\u5927\u5bb6\u8a8d\u8b58 LangChain \u7684\u5927\u81f4\u6a23\u8c8c\uff0c\u4e26\u4e14\u5b78\u6703\u5982\u4f55\u4f7f\u7528 LangChain \u4ee5\u53ca\u7d44\u5408\u51fa Chain \u8ddf\u4f7f\u7528\u7cfb\u7d71\u6307\u4ee4\uff0c\u5404\u4f4d\u73fe\u5728\u61c9\u8a72\u90fd\u6709\u80fd\u529b\u80fd\u5920\u505a\u51fa\u50c5\u4f9b\u81ea\u5df1\u4f7f\u7528\u7684 AI \u5de5\u5177\u3002<\/p>\n\n\n\n<p>\u7919\u65bc\u7bc7\u5e45\u592a\u9577\u5bb9\u6613\u6d88\u5316\u4e0d\u826f\uff0c\u6240\u4ee5\u672c\u7cfb\u5217\u6587\u7ae0\u5c07\u62c6\u6210\u591a\u7bc7\u9032\u884c\uff01\u5f8c\u7e8c\u5c07\u4ecb\u7d39\u66f4\u591a\u3001\u66f4\u5b8c\u6574\u7684 AI \u61c9\u7528\u8207 LangChain \u529f\u80fd\uff01<\/p>\n\n\n\n<p>\u4ee5\u4e0a\uff01<\/p>\n\n\n\n<p>Enjoy!<\/p>\n\n\n\n<p>\u8cc7\u6599\u4f86\u6e90:<a href=\"https:\/\/myapollo.com.tw\/blog\/langchain-tutorial-get-started\/\" data-type=\"link\" data-id=\"https:\/\/myapollo.com.tw\/blog\/langchain-tutorial-get-started\/\">https:\/\/myapollo.com.tw\/blog\/langchain-tutorial-get-started\/<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>2024-02-02 | Amo Chen AI \u6642\u4ee3\uff0c\u6253\u4e0d\u8d0f\u5c31\u52a0\u5165\u5b83\uff01 \u6240\u4ee5\u500b\u4eba\u8a8d\u70ba\u5b78\u6703 LangChain \u4e4b\u985e\u7684\u6846\u67b6\uff0c\u5728\u672a\u4f86\u53ef\u80fd\u6703\u662f\u6bcf\u500b\u7a0b\u5f0f\u8a2d\u8a08\u5e2b\u4e0d\u53ef\u6216\u7f3a\u7684\u6280\u8853\uff0c\u4e5f\u5c31\u662f\u8aaa\u9664\u4e86&hellip;<\/p>\n","protected":false},"author":9,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"jetpack_post_was_ever_published":false,"_jetpack_newsletter_access":"","_jetpack_dont_email_post_to_subs":false,"_jetpack_newsletter_tier_id":0,"_jetpack_memberships_contains_paywalled_content":false,"_jetpack_memberships_contains_paid_content":false,"footnotes":""},"categories":[579,4],"tags":[26,40],"class_list":["post-5479","post","type-post","status-publish","format-standard","hentry","category-579","category-industry-news","tag-ai","tag-40"],"gutentor_comment":0,"jetpack_featured_media_url":"","jetpack-related-posts":[],"jetpack_sharing_enabled":true,"_links":{"self":[{"href":"https:\/\/aict.nkust.edu.tw\/digitrans\/index.php?rest_route=\/wp\/v2\/posts\/5479","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/aict.nkust.edu.tw\/digitrans\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/aict.nkust.edu.tw\/digitrans\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/aict.nkust.edu.tw\/digitrans\/index.php?rest_route=\/wp\/v2\/users\/9"}],"replies":[{"embeddable":true,"href":"https:\/\/aict.nkust.edu.tw\/digitrans\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=5479"}],"version-history":[{"count":1,"href":"https:\/\/aict.nkust.edu.tw\/digitrans\/index.php?rest_route=\/wp\/v2\/posts\/5479\/revisions"}],"predecessor-version":[{"id":5483,"href":"https:\/\/aict.nkust.edu.tw\/digitrans\/index.php?rest_route=\/wp\/v2\/posts\/5479\/revisions\/5483"}],"wp:attachment":[{"href":"https:\/\/aict.nkust.edu.tw\/digitrans\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=5479"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/aict.nkust.edu.tw\/digitrans\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=5479"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/aict.nkust.edu.tw\/digitrans\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=5479"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}