rag.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import { load } from \"dotenv\";\n",
    "const env = await load();\n",
    "\n",
    "const process = {\n",
    "    env\n",
    "}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import { TextLoader } from \"langchain/document_loaders/fs/text\";\n",
    "const loader = new TextLoader(\"data/qiu.txt\");\n",
    "const docs = await loader.load();"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import { RecursiveCharacterTextSplitter } from \"langchain/text_splitter\";\n",
    "\n",
    "const splitter = new RecursiveCharacterTextSplitter({\n",
    "    chunkSize: 500,\n",
    "    chunkOverlap: 100,\n",
    "  });\n",
    "\n",
    "const splitDocs = await splitter.splitDocuments(docs);\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "console.log(splitDocs[4])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "console.log(splitDocs[4].pageContent)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import { OpenAIEmbeddings } from \"@langchain/openai\";\n",
    "\n",
    "\n",
    "const embeddings = new OpenAIEmbeddings();"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import { MemoryVectorStore } from \"langchain/vectorstores/memory\";\n",
    "\n",
    "const vectorstore = new MemoryVectorStore(embeddings);\n",
    "await vectorstore.addDocuments(splitDocs);"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "const retriever = vectorstore.asRetriever(2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "const res = await retriever.invoke(\"原文中，谁提出了宏原子的假设？并详细介绍给我宏原子假设的理论\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "res"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import { RunnableSequence } from \"@langchain/core/runnables\";\n",
    "import { Document } from \"@langchain/core/documents\";\n",
    "\n",
    "const convertDocsToString = (documents: Document[]): string => {\n",
    "     return documents.map((document) => document.pageContent).join(\"\\n\")\n",
    "    }\n",
    "const contextRetriverChain = RunnableSequence.from([\n",
    "    (input) => input.question,\n",
    "    retriever,\n",
    "    convertDocsToString\n",
    "])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "const result = await contextRetriverChain.invoke({ question: \"原文中，谁提出了宏原子的假设？并详细介绍给我宏原子假设的理论\"})\n",
    "\n",
    "console.log(result)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import { ChatPromptTemplate } from \"@langchain/core/prompts\";\n",
    "\n",
    "const TEMPLATE = `\n",
    "你是一个熟读刘慈欣的《球状闪电》的终极原著党，精通根据作品原文详细解释和回答问题，你在回答时会引用作品原文。\n",
    "并且回答时仅根据原文，尽可能回答用户问题，如果原文中没有相关内容，你可以回答“原文中没有相关内容”，\n",
    "\n",
    "以下是原文中跟用户回答相关的内容：\n",
    "{context}\n",
    "\n",
    "现在，你需要基于原文，回答以下问题：\n",
    "{question}`;\n",
    "\n",
    "const prompt = ChatPromptTemplate.fromTemplate(\n",
    "    TEMPLATE\n",
    ");"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import { ChatOpenAI } from \"@langchain/openai\";\n",
    "\n",
    "const model = new ChatOpenAI();"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import { StringOutputParser } from \"@langchain/core/output_parsers\";\n",
    "\n",
    "const ragChain = RunnableSequence.from([\n",
    "    {\n",
    "        context: contextRetriverChain,\n",
    "        question: (input) => input.question,\n",
    "    },\n",
    "    prompt,\n",
    "    model,\n",
    "    new StringOutputParser()\n",
    "])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "const answer = await ragChain.invoke({\n",
    "    question: \"什么是球状闪电\"\n",
    "  });\n",
    "  \n",
    "  console.log(answer);"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "const answer = await ragChain.invoke({\n",
    "    question: \"详细描述原文中有什么跟直升机相关的场景\"\n",
    "  });\n",
    "\n",
    "console.log(answer);"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Deno",
   "language": "typescript",
   "name": "deno"
  },
  "language_info": {
   "file_extension": ".ts",
   "mimetype": "text/x.typescript",
   "name": "typescript",
   "nb_converter": "script",
   "pygments_lexer": "typescript",
   "version": "5.3.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}