\u6211\u4eec\u5c06\u901a\u8fc7\u5b9e\u4f8b\u6f14\u793a\u5b8c\u6574\u7684\u5b9e\u73b0\u6d41\u7a0b\uff0c\u5305\u62ec\u6587\u6863\u5904\u7406\u3001\u5411\u91cf\u5b58\u50a8\u3001\u6a21\u578b\u8c03\u7528\u7b49\u5173\u952e\u6b65\u9aa4\u3002\u672c\u6559\u7a0b\u9009\u7528 DeepSeek-R1<\/a> 1.5B \u4f5c\u4e3a\u57fa\u7840\u8bed\u8a00\u6a21\u578b\u3002\u8003\u8651\u5230\u4e0d\u540c\u6a21\u578b\u5177\u6709\u5404\u81ea\u7684\u7279\u70b9\u548c\u6027\u80fd\u8868\u73b0\uff0c\u8bfb\u8005\u53ef\u4ee5\u6839\u636e\u5b9e\u9645\u9700\u6c42\u9009\u62e9\u5176\u4ed6\u5408\u9002\u7684\u6a21\u578b\u6765\u5b9e\u73b0 RAG<\/a> \u7cfb\u7edf\u3002<\/p>\n \u6ce8: \u672c\u6587\u6863\u5305\u542b\u6838\u5fc3\u4ee3\u7801\u7247\u6bb5\u548c\u8be6\u7ec6\u89e3\u91ca\u3002\u5b8c\u6574\u4ee3\u7801\u53ef\u89c1\u00a0notebook<\/a>\u00a0\u3002<\/p><\/blockquote>\n <\/p>\n \u9996\u5148\uff0c\u6211\u4eec\u9700\u8981\u4e0b\u8f7d Ollama \u4ee5\u53ca\u914d\u7f6e\u76f8\u5173\u73af\u5883\u3002<\/p>\n Ollama \u7684\u00a0GitHub\u4ed3\u5e93\u00a0\u4e2d\u63d0\u4f9b\u4e86\u8be6\u7ec6\u7684\u8bf4\u660e\uff0c\u7b80\u5355\u603b\u7ed3\u5982\u4e0b:<\/p>\n Step1\uff0c\u4e0b\u8f7d Ollama\u3002<\/p>\n \u4e0b\u8f7d<\/a>\u00a0\u5e76\u53cc\u51fb\u8fd0\u884c Ollama \u5e94\u7528\u7a0b\u5e8f\u3002<\/p>\n Step2\uff0c\u9a8c\u8bc1\u5b89\u88c5\u3002<\/p>\n \u5728\u547d\u4ee4\u884c\u8f93\u5165\u00a0 Step3\uff0c\u62c9\u53d6\u6a21\u578b\u3002<\/p>\n Step4\uff0c\u90e8\u7f72\u6a21\u578b\u3002<\/p>\n \u547d\u4ee4\u884c\u7a97\u53e3\u8fd0\u884c\u4ee5\u4e0b\u547d\u4ee4\uff0c\u90e8\u7f72\u6a21\u578b\u3002<\/p>\n \u4e5f\u53ef\u4ee5\u4ece\u547d\u4ee4\u884c\u76f4\u63a5\u8fd0\u884c\u90e8\u7f72\u6a21\u578b\uff0c\u4f8b\u5982\u00a0 \u6ce8\u610f\u5982\u679c\u53ea\u60f3\u4f7f\u7528 Ollama \u90e8\u7f72 DeepSeek R1 \u6a21\u578b\u5219\u65e0\u9700\u8fdb\u884c\u4ee5\u4e0b\u6b65\u9aa4\u3002<\/strong><\/p><\/blockquote>\n Step5\uff0c\u5b89\u88c5\u4f9d\u8d56\u3002<\/p>\n \u5b8c\u6210\u524d\u671f\u51c6\u5907\u5de5\u4f5c\u540e\uff0c\u8ba9\u6211\u4eec\u5f00\u59cb\u9010\u6b65\u6784\u5efa\u57fa\u4e8e LangChain\u3001Ollama \u548c DeepSeek R1<\/a> \u7684\u672c\u5730 RAG \u5e94\u7528\u3002\u4e0b\u9762\u5c06\u8be6\u7ec6\u4ecb\u7ecd\u5177\u4f53\u5b9e\u73b0\u6b65\u9aa4\u3002<\/p>\n <\/p>\n \u52a0\u8f7d PDF \u6587\u6863\u5e76\u5c06\u5176\u5207\u5206\u4e3a\u9002\u5f53\u5927\u5c0f\u7684\u6587\u672c\u5757\u3002<\/p>\n <\/p>\n \u4f7f\u7528 Chroma \u6570\u636e\u5e93\u5b58\u50a8\u6587\u6863\u5411\u91cf\uff0c\u5e76\u914d\u7f6e Ollama \u63d0\u4f9b\u7684\u5d4c\u5165\u6a21\u578b\u3002<\/p>\n <\/p>\n \u8bbe\u7f6e\u6a21\u578b\u548c\u63d0\u793a\u6a21\u677f\uff0c\u6784\u5efa\u5904\u7406\u94fe\u3002<\/p>\n <\/p>\n \u6574\u5408\u68c0\u7d22\u548c\u95ee\u7b54\u529f\u80fd\u3002<\/p>\n <\/p>\n \u672c\u6559\u7a0b\u8be6\u7ec6\u4ecb\u7ecd\u4e86\u5982\u4f55\u4f7f\u7528 DeepSeek R1 \u548c Ollama \u6784\u5efa\u672c\u5730\u5316\u7684 RAG \u5e94\u7528\u7cfb\u7edf\u3002\u6211\u4eec\u901a\u8fc7\u56db\u4e2a\u4e3b\u8981\u6b65\u9aa4\u5b9e\u73b0\u4e86\u5b8c\u6574\u7684\u529f\u80fd\uff1a<\/p>\n \u901a\u8fc7\u672c\u6559\u7a0b\uff0c\u53ef\u4ee5\u5feb\u901f\u642d\u5efa\u8d77\u81ea\u5df1\u7684\u672c\u5730 RAG \u7cfb\u7edf\uff0c\u5e76\u6839\u636e\u5b9e\u9645\u9700\u6c42\u8fdb\u884c\u5b9a\u5236\u5316\u6539\u8fdb\u3002\u5efa\u8bae\u5728\u5b9e\u8df5\u4e2d\u591a\u5c1d\u8bd5\u4e0d\u540c\u7684\u6a21\u578b\u548c\u53c2\u6570\u914d\u7f6e\uff0c\u4ee5\u83b7\u5f97\u6700\u4f73\u7684\u4f7f\u7528\u6548\u679c\u3002<\/p>\n \u6ce8: \u4f7f\u7528 streamlit \u6216 FastAPI \u7b49\u5de5\u5177\uff0c\u53ef\u4ee5\u5c06\u672c\u5730 RAG \u5e94\u7528\u90e8\u7f72\u4e3a Web \u670d\u52a1\uff0c\u5b9e\u73b0\u66f4\u5e7f\u6cdb\u7684\u5e94\u7528\u573a\u666f\u3002<\/p>\n \u4ed3\u5e93\u4e2d\u4e5f\u63d0\u4f9b\u4e86\u00a0 \u5bf9\u8bdd\u9875\u9762\u5982\u4e0b\uff1a<\/p>\n \u7b80\u4ecb \u672c\u6587\u6863\u8be6\u7ec6\u4ecb\u7ecd\u5982\u4f55\u5229\u7528 DeepSeek R1 \u548c Ollama \u6784\u5efa\u672c\u5730\u5316\u7684 RAG\uff08\u68c0\u7d22\u589e\u5f3a\u751f\u6210\uff09\u5e94\u7528\u3002\u540c\u65f6\u4e5f\u662f\u5bf9\u00a0\u4f7f\u7528 LangChain \u642d\u5efa\u672c\u5730 RAG \u5e94\u7528\u00a0\u7684\u8865\u5145\u3002 \u6211\u4eec\u5c06\u901a\u8fc7\u5b9e\u4f8b\u6f14\u793a\u5b8c\u6574\u7684\u5b9e\u73b0\u6d41\u7a0b\uff0c\u5305\u62ec\u6587\u6863\u5904\u7406\u3001\u5411\u91cf\u5b58…<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[182],"tags":[],"class_list":["post-28682","post","type-post","status-publish","format-standard","hentry","category-shicao"],"_links":{"self":[{"href":"https:\/\/www.kdjingpai.com\/ja\/wp-json\/wp\/v2\/posts\/28682","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.kdjingpai.com\/ja\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.kdjingpai.com\/ja\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.kdjingpai.com\/ja\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.kdjingpai.com\/ja\/wp-json\/wp\/v2\/comments?post=28682"}],"version-history":[{"count":0,"href":"https:\/\/www.kdjingpai.com\/ja\/wp-json\/wp\/v2\/posts\/28682\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.kdjingpai.com\/ja\/wp-json\/wp\/v2\/media?parent=28682"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.kdjingpai.com\/ja\/wp-json\/wp\/v2\/categories?post=28682"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.kdjingpai.com\/ja\/wp-json\/wp\/v2\/tags?post=28682"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}\u524d\u671f\u51c6\u5907<\/h2>\n
![\"-1\"](\"https:\/\/www.kdjingpai.com\/wp-content\/uploads\/2025\/03\/360fc850e4bf71d.png\" "\\"-1\\"")
<\/p>\nollama<\/code>\uff0c\u5982\u679c\u51fa\u73b0\u4ee5\u4e0b\u4fe1\u606f\uff0c\u8bf4\u660e Ollama \u5df2\u7ecf\u6210\u529f\u5b89\u88c5\u3002<\/p>\n![\"-2\"](\"https:\/\/www.kdjingpai.com\/wp-content\/uploads\/2025\/03\/3627b0e94a6f4bb.png\" "\\"-2\\"")
<\/p>\n\n
deepseek-r1:1.5b<\/code>\u00a0\u548c\u00a0nomic-embed-text<\/code>\u00a0\u4e3a\u4f8b:\n\n
ollama pull deepseek-r1:1.5b<\/code>\uff0c\u62c9\u53d6\u901a\u7528\u7684\u5f00\u6e90\u5927\u8bed\u8a00\u6a21\u578b\u00a0deepseek-r1:1.5b<\/code>\uff1b\uff08\u62c9\u53d6\u6a21\u578b\u65f6\uff0c\u53ef\u80fd\u6bd4\u8f83\u7f13\u6162\u3002\u5982\u679c\u51fa\u73b0\u62c9\u53d6\u9519\u8bef\uff0c\u53ef\u4ee5\u91cd\u65b0\u8f93\u5165\u6307\u4ee4\u62c9\u53d6\uff09<\/li>\nollama pull nomic-embed-text<\/code>\u00a0\u62c9\u53d6\u00a0\u6587\u672c\u5d4c\u5165\u6a21\u578b<\/a>\u00a0nomic-embed-text<\/code>\u3002<\/li>\n<\/ul>\n<\/li>\nlocalhost:11434<\/code>\u00a0\u4e0a\u542f\u52a8\u3002<\/li>\nCPU Memory > 8GB<\/code>\u3002<\/li>\n<\/ul>\nollama run deepseek-r1:1.5b\r\n<\/code><\/pre>\n![\"-3\"](\"https:\/\/www.kdjingpai.com\/wp-content\/uploads\/2025\/03\/8d4dca459d10339.png\" "\\"-3\\"")
<\/p>\nollama run deepseek-r1:1.5b<\/code>\u3002<\/p><\/blockquote>\n![\"-4\"](\"https:\/\/www.kdjingpai.com\/wp-content\/uploads\/2025\/03\/f56eaf302132434.png\" "\\"-4\\"")
<\/p>\n# langchain_community\r\npip install langchain langchain_community\r\n\r\n# Chroma\r\npip install langchain_chroma\r\n\r\n# Ollama\r\npip install langchain_ollama\r\n<\/code><\/pre>\n1. \u6587\u6863\u52a0\u8f7d<\/h2>\n
from langchain_community.document_loaders import PDFPlumberLoader\r\nfile = \"DeepSeek_R1.pdf\"\r\n# Load the PDF\r\nloader = PDFPlumberLoader(file)\r\ndocs = loader.load()\r\nfrom langchain.text_splitter import RecursiveCharacterTextSplitter\r\ntext_splitter = RecursiveCharacterTextSplitter(chunk_size=500, chunk_overlap=0)\r\nall_splits = text_splitter.split_documents(docs)\r\n<\/code><\/pre>\n2. \u521d\u59cb\u5316\u5411\u91cf\u5b58\u50a8<\/h2>\n
from langchain_chroma import Chroma\r\nfrom langchain_ollama import OllamaEmbeddings\r\nlocal_embeddings = OllamaEmbeddings(model=\"nomic-embed-text\")\r\nvectorstore = Chroma.from_documents(documents=all_splits, embedding=local_embeddings)\r\n<\/code><\/pre>\n3. \u6784\u5efa Chain \u8868\u8fbe\u5f0f<\/h2>\n
from langchain_core.output_parsers import StrOutputParser\r\nfrom langchain_core.prompts import ChatPromptTemplate\r\nfrom langchain_ollama import ChatOllama<\/a>\r\nmodel = ChatOllama(\r\nmodel=\"deepseek-r1:1.5b\",\r\n)\r\nprompt = ChatPromptTemplate.from_template(\r\n\"Summarize the main themes in these retrieved docs: {docs}\"\r\n)\r\n# \u5c06\u4f20\u5165\u7684\u6587\u6863\u8f6c\u6362\u6210\u5b57\u7b26\u4e32\u7684\u5f62\u5f0f\r\ndef format_docs(docs):\r\nreturn \"\\n\\n\".join(doc.page_content for doc in docs)\r\nchain = {\"docs\": format_docs} | prompt | model | StrOutputParser()\r\nquestion = \"What is the purpose of the DeepSeek project?\"\r\ndocs = vectorstore.similarity_search(question)\r\nchain.invoke(docs)\r\n<\/code><\/pre>\n4. \u5e26\u6709\u68c0\u7d22\u7684 QA<\/h2>\n
from langchain_core.runnables import RunnablePassthrough\r\nRAG_TEMPLATE = \"\"\"\r\nYou are an assistant for question-answering tasks. Use the following pieces<\/a> of retrieved context to answer the question. If you don't know the answer, just say that you don't know. Use three sentences maximum and keep the answer concise.\r\n<context>\r\n{context}\r\n<\/context>\r\nAnswer the following question:\r\n{question}\"\"\"\r\nrag_prompt = ChatPromptTemplate.from_template(RAG_TEMPLATE)\r\nretriever = vectorstore.as_retriever()\r\nqa_chain = (\r\n{\"context\": retriever | format_docs, \"question\": RunnablePassthrough()}\r\n| rag_prompt\r\n| model\r\n| StrOutputParser()\r\n)\r\nquestion = \"What is the purpose of the DeepSeek project?\"\r\n# Run\r\nqa_chain.invoke(question)\r\n<\/code><\/pre>\n\u603b\u7ed3<\/h2>\n
\n
app.py<\/code><\/a>\u00a0\u6587\u4ef6\uff0c\u53ef\u4ee5\u76f4\u63a5\u8fd0\u884c\u8be5\u6587\u4ef6\uff0c\u542f\u52a8 Web \u670d\u52a1\u3002 \u53c2\u8003\u6587\u6863\u00a0Build a RAG System with DeepSeek R1 & Ollama<\/a>\u3002 \u6ce8\u610f\uff1a\u8fd0\u884c\u8be5\u4ee3\u7801\u524d\uff0c\u8981\u63d0\u524d\u8fd0\u884c Ollama \u670d\u52a1\u3002<\/p><\/blockquote>\n![\"-5\"](\"https:\/\/www.kdjingpai.com\/wp-content\/uploads\/2025\/03\/c18abc0844c5225.png\" "\\"-5\\"")
<\/p>\n","protected":false},"excerpt":{"rendered":"