13. Structured-Output/structured-output.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 十三. 结构化输出(Structured Output)\n",
    "\n",
    "有时候我们希望输出的内容不是普通的字符串文本，而是像 json 那样结构化的数据，更能快捷高效地交付给下游业务使用。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'The output should be a markdown code snippet formatted in the following schema, including the leading and trailing \"```json\" and \"```\":\\n\\n```json\\n{\\n\\t\"answer\": string  // answer to the user\\'s question\\n\\t\"source\": string  // source referred to answer the user\\'s question, should be a website.\\n}\\n```\\n Only markdown content, no other else!'"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from langchain.output_parsers import StructuredOutputParser, ResponseSchema\n",
    "from langchain.prompts import PromptTemplate\n",
    "from langchain import OpenAI\n",
    "llm = OpenAI(model_name=\"text-davinci-003\")\n",
    "\n",
    "# 告诉他我们生成的内容需要哪些字段，每个字段类型式啥\n",
    "# 定义响应的结构(JSON)，两个字段 answer和source。\n",
    "response_schemas = [\n",
    "    ResponseSchema(name=\"answer\", description=\"answer to the user's question\"),\n",
    "    ResponseSchema(name=\"source\", description=\"source referred to answer the user's question, should be a website.\")\n",
    "]\n",
    "\n",
    "# 初始化解析器\n",
    "output_parser = StructuredOutputParser.from_response_schemas(response_schemas)\n",
    "\n",
    "\n",
    "# 获取响应格式化的指令\n",
    "format_instructions = output_parser.get_format_instructions()\n",
    "format_instructions += \"\\n Only markdown content, no other else!\"\n",
    "format_instructions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "# partial_variables允许在代码中预填充提示此模版的部分变量。这类似于接口，抽象类之间的关系prompt = PromptTemplate(\n",
    "#template=\"answer the users question as best as possible.\\n{format_instructions}\\n{question}\",\n",
    "template = \"\"\"\n",
    "answer the users question as best as possible.\n",
    "\n",
    "{format_instructions}\n",
    "\n",
    "% QUESTION:\n",
    "{question}\n",
    "\n",
    "\"\"\"\n",
    "\n",
    "\n",
    "# 将我们的格式描述嵌入到 prompt 中去，告诉 llm 我们需要他输出什么样格式的内容\n",
    "prompt = PromptTemplate(\n",
    "    input_variables=[\"question\"],\n",
    "    partial_variables={\"format_instructions\": format_instructions},\n",
    "    template=template\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'answer': 'Paris', 'source': 'https://en.wikipedia.org/wiki/Paris'}"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "response = prompt.format_prompt(question=\"what's the capital of France?\")\n",
    "output = llm(response.to_string())\n",
    "output\n",
    "output_parser.parse(output)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "base",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.10.9"
  },
  "orig_nbformat": 4
 },
 "nbformat": 4,
 "nbformat_minor": 2
}