\n", " OPENAI_API_KEY=\"your_api_key\" \n", "
\n", " \n", " \u66ff\u6362\"your_api_key\"\u4e3a\u4f60\u81ea\u5df1\u7684 API Key"]}, {"cell_type": "code", "execution_count": 1, "metadata": {}, "outputs": [], "source": ["# \u4e0b\u8f7d\u9700\u8981\u7684\u5305python-dotenv\u548copenai\n", "# \u5982\u679c\u4f60\u9700\u8981\u67e5\u770b\u5b89\u88c5\u8fc7\u7a0b\u65e5\u5fd7\uff0c\u53ef\u5220\u9664 -q \n", "!pip install -q python-dotenv\n", "!pip install -q openai"]}, {"cell_type": "code", "execution_count": 2, "metadata": {"tags": []}, "outputs": [], "source": ["import os\n", "import openai\n", "from dotenv import load_dotenv, find_dotenv\n", "\n", "# \u8bfb\u53d6\u672c\u5730/\u9879\u76ee\u7684\u73af\u5883\u53d8\u91cf\u3002\n", "\n", "# find_dotenv()\u5bfb\u627e\u5e76\u5b9a\u4f4d.env\u6587\u4ef6\u7684\u8def\u5f84\n", "# load_dotenv()\u8bfb\u53d6\u8be5.env\u6587\u4ef6\uff0c\u5e76\u5c06\u5176\u4e2d\u7684\u73af\u5883\u53d8\u91cf\u52a0\u8f7d\u5230\u5f53\u524d\u7684\u8fd0\u884c\u73af\u5883\u4e2d \n", "# \u5982\u679c\u4f60\u8bbe\u7f6e\u7684\u662f\u5168\u5c40\u7684\u73af\u5883\u53d8\u91cf\uff0c\u8fd9\u884c\u4ee3\u7801\u5219\u6ca1\u6709\u4efb\u4f55\u4f5c\u7528\u3002\n", "_ = load_dotenv(find_dotenv())\n", "\n", "# \u83b7\u53d6\u73af\u5883\u53d8\u91cf OPENAI_API_KEY\n", "openai.api_key = os.environ['OPENAI_API_KEY'] "]}, {"cell_type": "markdown", "metadata": {"tags": []}, "source": ["## \u4e8c\u3001\u76f4\u63a5\u4f7f\u7528OpenAI\n", "\n", "\u6211\u4eec\u5148\u4ece\u76f4\u63a5\u8c03\u7528OpenAI\u7684API\u5f00\u59cb\u3002\n", "\n", "`get_completion`\u51fd\u6570\u662f\u57fa\u4e8e`openai`\u7684\u5c01\u88c5\u51fd\u6570\uff0c\u5bf9\u4e8e\u7ed9\u5b9a\u63d0\u793a\uff08prompt\uff09\u8f93\u51fa\u76f8\u5e94\u7684\u56de\u7b54\u3002\u5176\u5305\u542b\u4e24\u4e2a\u53c2\u6570\n", " \n", " - `prompt` \u5fc5\u9700\u8f93\u5165\u53c2\u6570\u3002 \u4f60\u7ed9\u6a21\u578b\u7684**\u63d0\u793a\uff0c\u53ef\u4ee5\u662f\u4e00\u4e2a\u95ee\u9898\uff0c\u53ef\u4ee5\u662f\u4f60\u9700\u8981\u6a21\u578b\u5e2e\u52a9\u4f60\u505a\u7684\u4e8b**\uff08\u6539\u53d8\u6587\u672c\u5199\u4f5c\u98ce\u683c\uff0c\u7ffb\u8bd1\uff0c\u56de\u590d\u6d88\u606f\u7b49\u7b49\uff09\u3002\n", " - `model` \u975e\u5fc5\u9700\u8f93\u5165\u53c2\u6570\u3002\u9ed8\u8ba4\u4f7f\u7528gpt-3.5-turbo\u3002\u4f60\u4e5f\u53ef\u4ee5\u9009\u62e9\u5176\u4ed6\u6a21\u578b\u3002\n", " \n", "\u8fd9\u91cc\u7684\u63d0\u793a\u5bf9\u5e94\u6211\u4eec\u7ed9chatgpt\u7684\u95ee\u9898\uff0c\u51fd\u6570\u7ed9\u51fa\u7684\u8f93\u51fa\u5219\u5bf9\u5e94chatpgt\u7ed9\u6211\u4eec\u7684\u7b54\u6848\u3002"]}, {"cell_type": "code", "execution_count": 3, "metadata": {"tags": []}, "outputs": [], "source": ["def get_completion(prompt, model=\"gpt-3.5-turbo\"):\n", " \n", " messages = [{\"role\": \"user\", \"content\": prompt}]\n", " \n", " response = openai.ChatCompletion.create(\n", " model=model,\n", " messages=messages,\n", " temperature=0, \n", " )\n", " return response.choices[0].message[\"content\"]"]}, {"cell_type": "markdown", "metadata": {"tags": []}, "source": ["### 2.1 \u8ba1\u7b971+1\n", "\n", "\u6211\u4eec\u6765\u4e00\u4e2a\u7b80\u5355\u7684\u4f8b\u5b50 - \u5206\u522b\u7528\u4e2d\u82f1\u6587\u95ee\u95ee\u6a21\u578b\n", "\n", "- \u4e2d\u6587\u63d0\u793a(Prompt in Chinese)\uff1a `1+1\u662f\u4ec0\u4e48\uff1f`\n", "- \u82f1\u6587\u63d0\u793a(Prompt in English)\uff1a `What is 1+1?`"]}, {"cell_type": "code", "execution_count": 4, "metadata": {}, "outputs": [{"data": {"text/plain": ["'1+1\u7b49\u4e8e2\u3002'"]}, "execution_count": 4, "metadata": {}, "output_type": "execute_result"}], "source": ["# \u4e2d\u6587\n", "get_completion(\"1+1\u662f\u4ec0\u4e48\uff1f\")"]}, {"cell_type": "code", "execution_count": 5, "metadata": {"tags": []}, "outputs": [{"data": {"text/plain": ["'1+1 equals 2.'"]}, "execution_count": 5, "metadata": {}, "output_type": "execute_result"}], "source": ["# \u82f1\u6587\n", "get_completion(\"What is 1+1?\")"]}, {"cell_type": "markdown", "metadata": {"tags": []}, "source": ["### 2.2 \u7528\u7f8e\u5f0f\u82f1\u8bed\u8868\u8fbe\u6d77\u76d7\u90ae\u4ef6\n", "\n", "\u4e0a\u9762\u7684\u7b80\u5355\u4f8b\u5b50\uff0c\u6a21\u578b`gpt-3.5-turbo`\u5bf9\u6211\u4eec\u7684\u5173\u4e8e1+1\u662f\u4ec0\u4e48\u7684\u63d0\u95ee\u7ed9\u51fa\u4e86\u56de\u7b54\u3002\n", "\n", "\u73b0\u5728\u6211\u4eec\u6765\u770b\u4e00\u4e2a\u590d\u6742\u4e00\u70b9\u7684\u4f8b\u5b50\uff1a \n", "\n", "\u5047\u8bbe\u6211\u4eec\u662f\u7535\u5546\u516c\u53f8\u5458\u5de5\uff0c\u6211\u4eec\u7684\u987e\u5ba2\u662f\u4e00\u540d\u6d77\u76d7A\uff0c\u4ed6\u5728\u6211\u4eec\u7684\u7f51\u7ad9\u4e0a\u4e70\u4e86\u4e00\u4e2a\u69a8\u6c41\u673a\u7528\u6765\u505a\u5976\u6614\uff0c\u5728\u5236\u4f5c\u5976\u6614\u7684\u8fc7\u7a0b\u4e2d\uff0c\u5976\u6614\u7684\u76d6\u5b50\u98de\u4e86\u51fa\u53bb\uff0c\u5f04\u5f97\u53a8\u623f\u5899\u4e0a\u5230\u5904\u90fd\u662f\u3002\u4e8e\u662f\u6d77\u76d7A\u7ed9\u6211\u4eec\u7684\u5ba2\u670d\u4e2d\u5fc3\u5199\u6765\u4ee5\u4e0b\u90ae\u4ef6\uff1a`customer_email`"]}, {"cell_type": "code", "execution_count": 6, "metadata": {"tags": []}, "outputs": [], "source": ["customer_email = \"\"\"\n", "Arrr, I be fuming that me blender lid \\\n", "flew off and splattered me kitchen walls \\\n", "with smoothie! And to make matters worse,\\\n", "the warranty don't cover the cost of \\\n", "cleaning up me kitchen. I need yer help \\\n", "right now, matey!\n", "\"\"\""]}, {"cell_type": "markdown", "metadata": {}, "source": ["\u6211\u4eec\u7684\u5ba2\u670d\u4eba\u5458\u5bf9\u4e8e\u6d77\u76d7\u7684\u63aa\u8f9e\u8868\u8fbe\u89c9\u5f97\u6709\u70b9\u96be\u4ee5\u7406\u89e3\u3002 \u73b0\u5728\u6211\u4eec\u60f3\u8981\u5b9e\u73b0\u4e24\u4e2a\u5c0f\u76ee\u6807\uff1a\n", "\n", "- \u8ba9\u6a21\u578b\u7528\u7f8e\u5f0f\u82f1\u8bed\u7684\u8868\u8fbe\u65b9\u5f0f\u5c06\u6d77\u76d7\u7684\u90ae\u4ef6\u8fdb\u884c\u7ffb\u8bd1\uff0c\u5ba2\u670d\u4eba\u5458\u53ef\u4ee5\u66f4\u597d\u7406\u89e3\u3002*\u8fd9\u91cc\u6d77\u76d7\u7684\u82f1\u6587\u8868\u8fbe\u53ef\u4ee5\u7406\u89e3\u4e3a\u82f1\u6587\u7684\u65b9\u8a00\uff0c\u5176\u4e0e\u7f8e\u5f0f\u82f1\u8bed\u7684\u5173\u7cfb\uff0c\u5c31\u5982\u56db\u5ddd\u8bdd\u4e0e\u666e\u901a\u8bdd\u7684\u5173\u7cfb\u3002\n", "- \u8ba9\u6a21\u578b\u5728\u7ffb\u8bd1\u662f\u7528\u5e73\u548c\u5c0a\u91cd\u7684\u8bed\u6c14\u8fdb\u884c\u8868\u8fbe\uff0c\u5ba2\u670d\u4eba\u5458\u7684\u5fc3\u60c5\u4e5f\u4f1a\u66f4\u597d\u3002\n", "\n", "\u6839\u636e\u8fd9\u4e24\u4e2a\u5c0f\u76ee\u6807\uff0c\u5b9a\u4e49\u4e00\u4e0b\u6587\u672c\u8868\u8fbe\u98ce\u683c\uff1a`style`"]}, {"cell_type": "code", "execution_count": 7, "metadata": {"tags": []}, "outputs": [], "source": ["# \u7f8e\u5f0f\u82f1\u8bed + \u5e73\u9759\u3001\u5c0a\u656c\u7684\u8bed\u8c03\n", "style = \"\"\"American English \\\n", "in a calm and respectful tone\n", "\"\"\""]}, {"cell_type": "markdown", "metadata": {}, "source": ["\u4e0b\u4e00\u6b65\u9700\u8981\u505a\u7684\u662f\u5c06`customer_email`\u548c`style`\u7ed3\u5408\u8d77\u6765\u6784\u9020\u6211\u4eec\u7684\u63d0\u793a:`prompt`"]}, {"cell_type": "code", "execution_count": 8, "metadata": {}, "outputs": [], "source": ["# \u975e\u6b63\u5f0f\u7528\u8bed\n", "customer_email = \"\"\" \n", "\u963f\uff0c\u6211\u5f88\u751f\u6c14\uff0c\\\n", "\u56e0\u4e3a\u6211\u7684\u6405\u62cc\u673a\u76d6\u6389\u4e86\uff0c\\\n", "\u628a\u5976\u6614\u6e85\u5230\u4e86\u53a8\u623f\u7684\u5899\u4e0a\uff01\\\n", "\u66f4\u7cdf\u7cd5\u7684\u662f\uff0c\u4fdd\u4fee\u4e0d\u5305\u62ec\u6253\u626b\u53a8\u623f\u7684\u8d39\u7528\u3002\\\n", "\u6211\u73b0\u5728\u9700\u8981\u4f60\u7684\u5e2e\u52a9\uff0c\u4f19\u8ba1\uff01\n", "\"\"\""]}, {"cell_type": "code", "execution_count": 9, "metadata": {"tags": []}, "outputs": [{"name": "stdout", "output_type": "stream", "text": ["Translate the text that is delimited by triple backticks \n", "into a style that is American English in a calm and respectful tone\n", ".\n", "text: ``` \n", "\u963f\uff0c\u6211\u5f88\u751f\u6c14\uff0c\u56e0\u4e3a\u6211\u7684\u6405\u62cc\u673a\u76d6\u6389\u4e86\uff0c\u628a\u5976\u6614\u6e85\u5230\u4e86\u53a8\u623f\u7684\u5899\u4e0a\uff01\u66f4\u7cdf\u7cd5\u7684\u662f\uff0c\u4fdd\u4fee\u4e0d\u5305\u62ec\u6253\u626b\u53a8\u623f\u7684\u8d39\u7528\u3002\u6211\u73b0\u5728\u9700\u8981\u4f60\u7684\u5e2e\u52a9\uff0c\u4f19\u8ba1\uff01\n", "```\n", "\n"]}], "source": ["# \u8981\u6c42\u6a21\u578b\u6839\u636e\u7ed9\u51fa\u7684\u8bed\u8c03\u8fdb\u884c\u8f6c\u5316\n", "prompt = f\"\"\"Translate the text \\\n", "that is delimited by triple backticks \n", "into a style that is {style}.\n", "text: ```{customer_email}```\n", "\"\"\"\n", "\n", "print(prompt)"]}, {"cell_type": "markdown", "metadata": {}, "source": ["`prompt` \u6784\u9020\u597d\u4e86\uff0c\u6211\u4eec\u53ef\u4ee5\u8c03\u7528`get_completion`\u5f97\u5230\u6211\u4eec\u60f3\u8981\u7684\u7ed3\u679c - \u7528\u5e73\u548c\u5c0a\u91cd\u7684\u8bed\u6c14\uff0c\u7f8e\u5f0f\u82f1\u8bed\u8868\u8fbe\u7684\u6d77\u76d7\u8bed\u8a00\u90ae\u4ef6"]}, {"cell_type": "code", "execution_count": 10, "metadata": {}, "outputs": [], "source": ["response = get_completion(prompt)"]}, {"cell_type": "code", "execution_count": 11, "metadata": {"tags": []}, "outputs": [{"data": {"text/plain": ["\"Oh, I'm really frustrated because the lid of my blender fell off and splattered the milkshake all over the kitchen wall! To make matters worse, the warranty doesn't cover the cost of cleaning the kitchen. I could really use your help right now, buddy!\""]}, "execution_count": 11, "metadata": {}, "output_type": "execute_result"}], "source": ["response"]}, {"cell_type": "markdown", "metadata": {}, "source": ["\u5bf9\u6bd4\u8bed\u8a00\u98ce\u683c\u8f6c\u6362\u524d\u540e\uff0c\u7528\u8bcd\u66f4\u4e3a\u6b63\u5f0f\uff0c\u66ff\u6362\u4e86\u6781\u7aef\u60c5\u7eea\u7684\u8868\u8fbe\uff0c\u5e76\u8868\u8fbe\u4e86\u611f\u8c22\u3002\n", "\n", "\u2728 \u4f60\u53ef\u4ee5\u5c1d\u8bd5\u4fee\u6539\u63d0\u793a\uff0c\u770b\u53ef\u4ee5\u5f97\u5230\u4ec0\u4e48\u4e0d\u4e00\u6837\u7684\u7ed3\u679c\ud83d\ude09"]}, {"cell_type": "markdown", "metadata": {"tags": []}, "source": ["### 2.3 \u4e2d\u6587\u7248"]}, {"cell_type": "code", "execution_count": 12, "metadata": {}, "outputs": [], "source": ["# \u666e\u901a\u8bdd + \u5e73\u9759\u3001\u5c0a\u656c\u7684\u8bed\u8c03\n", "style = \"\"\"\u6b63\u5f0f\u666e\u901a\u8bdd \\\n", "\u7528\u4e00\u4e2a\u5e73\u9759\u3001\u5c0a\u656c\u7684\u8bed\u8c03\n", "\"\"\""]}, {"cell_type": "code", "execution_count": 13, "metadata": {}, "outputs": [{"name": "stdout", "output_type": "stream", "text": ["\u628a\u7531\u4e09\u4e2a\u53cd\u5f15\u53f7\u5206\u9694\u7684\u6587\u672c\u7ffb\u8bd1\u6210\u4e00\u79cd\u6b63\u5f0f\u666e\u901a\u8bdd \u7528\u4e00\u4e2a\u5e73\u9759\u3001\u5c0a\u656c\u7684\u8bed\u8c03\n", "\u98ce\u683c\u3002\n", "\u6587\u672c: ``` \n", "\u963f\uff0c\u6211\u5f88\u751f\u6c14\uff0c\u56e0\u4e3a\u6211\u7684\u6405\u62cc\u673a\u76d6\u6389\u4e86\uff0c\u628a\u5976\u6614\u6e85\u5230\u4e86\u53a8\u623f\u7684\u5899\u4e0a\uff01\u66f4\u7cdf\u7cd5\u7684\u662f\uff0c\u4fdd\u4fee\u4e0d\u5305\u62ec\u6253\u626b\u53a8\u623f\u7684\u8d39\u7528\u3002\u6211\u73b0\u5728\u9700\u8981\u4f60\u7684\u5e2e\u52a9\uff0c\u4f19\u8ba1\uff01\n", "```\n", "\n"]}], "source": ["# \u8981\u6c42\u6a21\u578b\u6839\u636e\u7ed9\u51fa\u7684\u8bed\u8c03\u8fdb\u884c\u8f6c\u5316\n", "prompt = f\"\"\"\u628a\u7531\u4e09\u4e2a\u53cd\u5f15\u53f7\u5206\u9694\u7684\u6587\u672c\\\n", "\u7ffb\u8bd1\u6210\u4e00\u79cd{style}\u98ce\u683c\u3002\n", "\u6587\u672c: ```{customer_email}```\n", "\"\"\"\n", "\n", "print(prompt)\n", "\n"]}, {"cell_type": "code", "execution_count": 14, "metadata": {}, "outputs": [{"data": {"text/plain": ["'\u5c0a\u656c\u7684\u670b\u53cb\u4eec\uff0c\u6211\u611f\u5230\u975e\u5e38\u4e0d\u5b89\uff0c\u56e0\u4e3a\u6211\u7684\u6405\u62cc\u673a\u76d6\u5b50\u4e0d\u614e\u6389\u843d\uff0c\u5bfc\u81f4\u5976\u6614\u6e85\u5230\u4e86\u53a8\u623f\u7684\u5899\u58c1\u4e0a\uff01\u66f4\u52a0\u4ee4\u4eba\u7cdf\u5fc3\u7684\u662f\uff0c\u4fdd\u4fee\u670d\u52a1\u5e76\u4e0d\u5305\u542b\u53a8\u623f\u6e05\u6d01\u7684\u8d39\u7528\u3002\u6b64\u523b\uff0c\u6211\u771f\u8bda\u5730\u8bf7\u6c42\u5404\u4f4d\u7684\u5e2e\u52a9\uff0c\u670b\u53cb\u4eec\uff01'"]}, "execution_count": 14, "metadata": {}, "output_type": "execute_result"}], "source": ["response = get_completion(prompt)\n", "\n", "response"]}, {"cell_type": "markdown", "metadata": {"tags": []}, "source": ["## \u4e09\u3001\u901a\u8fc7LangChain\u4f7f\u7528OpenAI\n", "\n", "\u5728\u524d\u9762\u4e00\u90e8\u5206\uff0c\u6211\u4eec\u901a\u8fc7\u5c01\u88c5\u51fd\u6570`get_completion`\u76f4\u63a5\u8c03\u7528\u4e86OpenAI\u5b8c\u6210\u4e86\u5bf9\u65b9\u8a00\u90ae\u4ef6\u8fdb\u884c\u4e86\u7684\u7ffb\u8bd1\uff0c\u5f97\u5230\u7528\u5e73\u548c\u5c0a\u91cd\u7684\u8bed\u6c14\u3001\u6b63\u5f0f\u7684\u666e\u901a\u8bdd\u8868\u8fbe\u7684\u90ae\u4ef6\u3002\n", "\n", "\u8ba9\u6211\u4eec\u5c1d\u8bd5\u4f7f\u7528LangChain\u6765\u5b9e\u73b0\u76f8\u540c\u7684\u529f\u80fd\u3002"]}, {"cell_type": "code", "execution_count": 15, "metadata": {"tags": []}, "outputs": [], "source": ["# \u5982\u679c\u4f60\u9700\u8981\u67e5\u770b\u5b89\u88c5\u8fc7\u7a0b\u65e5\u5fd7\uff0c\u53ef\u5220\u9664 -q \n", "# --upgrade \u8ba9\u6211\u4eec\u53ef\u4ee5\u5b89\u88c5\u5230\u6700\u65b0\u7248\u672c\u7684 langchain\n", "!pip install -q --upgrade langchain"]}, {"cell_type": "markdown", "metadata": {"tags": []}, "source": ["### 3.1 \u6a21\u578b\n", "\n", "\u4ece`langchain.chat_models`\u5bfc\u5165`OpenAI`\u7684\u5bf9\u8bdd\u6a21\u578b`ChatOpenAI`\u3002 \u9664\u53bbOpenAI\u4ee5\u5916\uff0c`langchain.chat_models`\u8fd8\u96c6\u6210\u4e86\u5176\u4ed6\u5bf9\u8bdd\u6a21\u578b\uff0c\u66f4\u591a\u7ec6\u8282\u53ef\u4ee5\u67e5\u770b[Langchain\u5b98\u65b9\u6587\u6863](https://python.langchain.com/en/latest/modules/models/chat/integrations.html)\u3002"]}, {"cell_type": "code", "execution_count": 16, "metadata": {"tags": []}, "outputs": [], "source": ["from langchain.chat_models import ChatOpenAI"]}, {"cell_type": "code", "execution_count": 17, "metadata": {"tags": []}, "outputs": [{"data": {"text/plain": ["ChatOpenAI(cache=None, verbose=False, callbacks=None, callback_manager=None, tags=None, metadata=None, client=\n", " OPENAI_API_KEY=\"your_api_key\" \n", "
\n", " \n", " \u66ff\u6362\"your_api_key\"\u4e3a\u4f60\u81ea\u5df1\u7684 API Key"]}, {"cell_type": "code", "execution_count": 1, "metadata": {}, "outputs": [], "source": ["# \u4e0b\u8f7d\u9700\u8981\u7684\u5305python-dotenv\u548copenai\n", "# \u5982\u679c\u4f60\u9700\u8981\u67e5\u770b\u5b89\u88c5\u8fc7\u7a0b\u65e5\u5fd7\uff0c\u53ef\u5220\u9664 -q \n", "!pip install -q python-dotenv\n", "!pip install -q openai"]}, {"cell_type": "code", "execution_count": 2, "metadata": {"tags": []}, "outputs": [], "source": ["import os\n", "import openai\n", "from dotenv import load_dotenv, find_dotenv\n", "\n", "# \u8bfb\u53d6\u672c\u5730/\u9879\u76ee\u7684\u73af\u5883\u53d8\u91cf\u3002\n", "\n", "# find_dotenv()\u5bfb\u627e\u5e76\u5b9a\u4f4d.env\u6587\u4ef6\u7684\u8def\u5f84\n", "# load_dotenv()\u8bfb\u53d6\u8be5.env\u6587\u4ef6\uff0c\u5e76\u5c06\u5176\u4e2d\u7684\u73af\u5883\u53d8\u91cf\u52a0\u8f7d\u5230\u5f53\u524d\u7684\u8fd0\u884c\u73af\u5883\u4e2d \n", "# \u5982\u679c\u4f60\u8bbe\u7f6e\u7684\u662f\u5168\u5c40\u7684\u73af\u5883\u53d8\u91cf\uff0c\u8fd9\u884c\u4ee3\u7801\u5219\u6ca1\u6709\u4efb\u4f55\u4f5c\u7528\u3002\n", "_ = load_dotenv(find_dotenv())\n", "\n", "# \u83b7\u53d6\u73af\u5883\u53d8\u91cf OPENAI_API_KEY\n", "openai.api_key = os.environ['OPENAI_API_KEY'] "]}, {"cell_type": "markdown", "metadata": {"tags": []}, "source": ["## \u4e8c\u3001\u76f4\u63a5\u4f7f\u7528OpenAI\n", "\n", "\u6211\u4eec\u5148\u4ece\u76f4\u63a5\u8c03\u7528OpenAI\u7684API\u5f00\u59cb\u3002\n", "\n", "`get_completion`\u51fd\u6570\u662f\u57fa\u4e8e`openai`\u7684\u5c01\u88c5\u51fd\u6570\uff0c\u5bf9\u4e8e\u7ed9\u5b9a\u63d0\u793a\uff08prompt\uff09\u8f93\u51fa\u76f8\u5e94\u7684\u56de\u7b54\u3002\u5176\u5305\u542b\u4e24\u4e2a\u53c2\u6570\n", " \n", " - `prompt` \u5fc5\u9700\u8f93\u5165\u53c2\u6570\u3002 \u4f60\u7ed9\u6a21\u578b\u7684**\u63d0\u793a\uff0c\u53ef\u4ee5\u662f\u4e00\u4e2a\u95ee\u9898\uff0c\u53ef\u4ee5\u662f\u4f60\u9700\u8981\u6a21\u578b\u5e2e\u52a9\u4f60\u505a\u7684\u4e8b**\uff08\u6539\u53d8\u6587\u672c\u5199\u4f5c\u98ce\u683c\uff0c\u7ffb\u8bd1\uff0c\u56de\u590d\u6d88\u606f\u7b49\u7b49\uff09\u3002\n", " - `model` \u975e\u5fc5\u9700\u8f93\u5165\u53c2\u6570\u3002\u9ed8\u8ba4\u4f7f\u7528gpt-3.5-turbo\u3002\u4f60\u4e5f\u53ef\u4ee5\u9009\u62e9\u5176\u4ed6\u6a21\u578b\u3002\n", " \n", "\u8fd9\u91cc\u7684\u63d0\u793a\u5bf9\u5e94\u6211\u4eec\u7ed9chatgpt\u7684\u95ee\u9898\uff0c\u51fd\u6570\u7ed9\u51fa\u7684\u8f93\u51fa\u5219\u5bf9\u5e94chatpgt\u7ed9\u6211\u4eec\u7684\u7b54\u6848\u3002"]}, {"cell_type": "code", "execution_count": 3, "metadata": {"tags": []}, "outputs": [], "source": ["def get_completion(prompt, model=\"gpt-3.5-turbo\"):\n", " \n", " messages = [{\"role\": \"user\", \"content\": prompt}]\n", " \n", " response = openai.ChatCompletion.create(\n", " model=model,\n", " messages=messages,\n", " temperature=0, \n", " )\n", " return response.choices[0].message[\"content\"]"]}, {"cell_type": "markdown", "metadata": {"tags": []}, "source": ["### 2.1 \u8ba1\u7b971+1\n", "\n", "\u6211\u4eec\u6765\u4e00\u4e2a\u7b80\u5355\u7684\u4f8b\u5b50 - \u5206\u522b\u7528\u4e2d\u82f1\u6587\u95ee\u95ee\u6a21\u578b\n", "\n", "- \u4e2d\u6587\u63d0\u793a(Prompt in Chinese)\uff1a `1+1\u662f\u4ec0\u4e48\uff1f`\n", "- \u82f1\u6587\u63d0\u793a(Prompt in English)\uff1a `What is 1+1?`"]}, {"cell_type": "code", "execution_count": 4, "metadata": {}, "outputs": [{"data": {"text/plain": ["'1+1\u7b49\u4e8e2\u3002'"]}, "execution_count": 4, "metadata": {}, "output_type": "execute_result"}], "source": ["# \u4e2d\u6587\n", "get_completion(\"1+1\u662f\u4ec0\u4e48\uff1f\")"]}, {"cell_type": "code", "execution_count": 5, "metadata": {"tags": []}, "outputs": [{"data": {"text/plain": ["'1+1 equals 2.'"]}, "execution_count": 5, "metadata": {}, "output_type": "execute_result"}], "source": ["# \u82f1\u6587\n", "get_completion(\"What is 1+1?\")"]}, {"cell_type": "markdown", "metadata": {"tags": []}, "source": ["### 2.2 \u7528\u7f8e\u5f0f\u82f1\u8bed\u8868\u8fbe\u6d77\u76d7\u90ae\u4ef6\n", "\n", "\u4e0a\u9762\u7684\u7b80\u5355\u4f8b\u5b50\uff0c\u6a21\u578b`gpt-3.5-turbo`\u5bf9\u6211\u4eec\u7684\u5173\u4e8e1+1\u662f\u4ec0\u4e48\u7684\u63d0\u95ee\u7ed9\u51fa\u4e86\u56de\u7b54\u3002\n", "\n", "\u73b0\u5728\u6211\u4eec\u6765\u770b\u4e00\u4e2a\u590d\u6742\u4e00\u70b9\u7684\u4f8b\u5b50\uff1a \n", "\n", "\u5047\u8bbe\u6211\u4eec\u662f\u7535\u5546\u516c\u53f8\u5458\u5de5\uff0c\u6211\u4eec\u7684\u987e\u5ba2\u662f\u4e00\u540d\u6d77\u76d7A\uff0c\u4ed6\u5728\u6211\u4eec\u7684\u7f51\u7ad9\u4e0a\u4e70\u4e86\u4e00\u4e2a\u69a8\u6c41\u673a\u7528\u6765\u505a\u5976\u6614\uff0c\u5728\u5236\u4f5c\u5976\u6614\u7684\u8fc7\u7a0b\u4e2d\uff0c\u5976\u6614\u7684\u76d6\u5b50\u98de\u4e86\u51fa\u53bb\uff0c\u5f04\u5f97\u53a8\u623f\u5899\u4e0a\u5230\u5904\u90fd\u662f\u3002\u4e8e\u662f\u6d77\u76d7A\u7ed9\u6211\u4eec\u7684\u5ba2\u670d\u4e2d\u5fc3\u5199\u6765\u4ee5\u4e0b\u90ae\u4ef6\uff1a`customer_email`"]}, {"cell_type": "code", "execution_count": 6, "metadata": {"tags": []}, "outputs": [], "source": ["customer_email = \"\"\"\n", "Arrr, I be fuming that me blender lid \\\n", "flew off and splattered me kitchen walls \\\n", "with smoothie! And to make matters worse,\\\n", "the warranty don't cover the cost of \\\n", "cleaning up me kitchen. I need yer help \\\n", "right now, matey!\n", "\"\"\""]}, {"cell_type": "markdown", "metadata": {}, "source": ["\u6211\u4eec\u7684\u5ba2\u670d\u4eba\u5458\u5bf9\u4e8e\u6d77\u76d7\u7684\u63aa\u8f9e\u8868\u8fbe\u89c9\u5f97\u6709\u70b9\u96be\u4ee5\u7406\u89e3\u3002 \u73b0\u5728\u6211\u4eec\u60f3\u8981\u5b9e\u73b0\u4e24\u4e2a\u5c0f\u76ee\u6807\uff1a\n", "\n", "- \u8ba9\u6a21\u578b\u7528\u7f8e\u5f0f\u82f1\u8bed\u7684\u8868\u8fbe\u65b9\u5f0f\u5c06\u6d77\u76d7\u7684\u90ae\u4ef6\u8fdb\u884c\u7ffb\u8bd1\uff0c\u5ba2\u670d\u4eba\u5458\u53ef\u4ee5\u66f4\u597d\u7406\u89e3\u3002*\u8fd9\u91cc\u6d77\u76d7\u7684\u82f1\u6587\u8868\u8fbe\u53ef\u4ee5\u7406\u89e3\u4e3a\u82f1\u6587\u7684\u65b9\u8a00\uff0c\u5176\u4e0e\u7f8e\u5f0f\u82f1\u8bed\u7684\u5173\u7cfb\uff0c\u5c31\u5982\u56db\u5ddd\u8bdd\u4e0e\u666e\u901a\u8bdd\u7684\u5173\u7cfb\u3002\n", "- \u8ba9\u6a21\u578b\u5728\u7ffb\u8bd1\u662f\u7528\u5e73\u548c\u5c0a\u91cd\u7684\u8bed\u6c14\u8fdb\u884c\u8868\u8fbe\uff0c\u5ba2\u670d\u4eba\u5458\u7684\u5fc3\u60c5\u4e5f\u4f1a\u66f4\u597d\u3002\n", "\n", "\u6839\u636e\u8fd9\u4e24\u4e2a\u5c0f\u76ee\u6807\uff0c\u5b9a\u4e49\u4e00\u4e0b\u6587\u672c\u8868\u8fbe\u98ce\u683c\uff1a`style`"]}, {"cell_type": "code", "execution_count": 7, "metadata": {"tags": []}, "outputs": [], "source": ["# \u7f8e\u5f0f\u82f1\u8bed + \u5e73\u9759\u3001\u5c0a\u656c\u7684\u8bed\u8c03\n", "style = \"\"\"American English \\\n", "in a calm and respectful tone\n", "\"\"\""]}, {"cell_type": "markdown", "metadata": {}, "source": ["\u4e0b\u4e00\u6b65\u9700\u8981\u505a\u7684\u662f\u5c06`customer_email`\u548c`style`\u7ed3\u5408\u8d77\u6765\u6784\u9020\u6211\u4eec\u7684\u63d0\u793a:`prompt`"]}, {"cell_type": "code", "execution_count": 8, "metadata": {}, "outputs": [], "source": ["# \u975e\u6b63\u5f0f\u7528\u8bed\n", "customer_email = \"\"\" \n", "\u963f\uff0c\u6211\u5f88\u751f\u6c14\uff0c\\\n", "\u56e0\u4e3a\u6211\u7684\u6405\u62cc\u673a\u76d6\u6389\u4e86\uff0c\\\n", "\u628a\u5976\u6614\u6e85\u5230\u4e86\u53a8\u623f\u7684\u5899\u4e0a\uff01\\\n", "\u66f4\u7cdf\u7cd5\u7684\u662f\uff0c\u4fdd\u4fee\u4e0d\u5305\u62ec\u6253\u626b\u53a8\u623f\u7684\u8d39\u7528\u3002\\\n", "\u6211\u73b0\u5728\u9700\u8981\u4f60\u7684\u5e2e\u52a9\uff0c\u4f19\u8ba1\uff01\n", "\"\"\""]}, {"cell_type": "code", "execution_count": 9, "metadata": {"tags": []}, "outputs": [{"name": "stdout", "output_type": "stream", "text": ["Translate the text that is delimited by triple backticks \n", "into a style that is American English in a calm and respectful tone\n", ".\n", "text: ``` \n", "\u963f\uff0c\u6211\u5f88\u751f\u6c14\uff0c\u56e0\u4e3a\u6211\u7684\u6405\u62cc\u673a\u76d6\u6389\u4e86\uff0c\u628a\u5976\u6614\u6e85\u5230\u4e86\u53a8\u623f\u7684\u5899\u4e0a\uff01\u66f4\u7cdf\u7cd5\u7684\u662f\uff0c\u4fdd\u4fee\u4e0d\u5305\u62ec\u6253\u626b\u53a8\u623f\u7684\u8d39\u7528\u3002\u6211\u73b0\u5728\u9700\u8981\u4f60\u7684\u5e2e\u52a9\uff0c\u4f19\u8ba1\uff01\n", "```\n", "\n"]}], "source": ["# \u8981\u6c42\u6a21\u578b\u6839\u636e\u7ed9\u51fa\u7684\u8bed\u8c03\u8fdb\u884c\u8f6c\u5316\n", "prompt = f\"\"\"Translate the text \\\n", "that is delimited by triple backticks \n", "into a style that is {style}.\n", "text: ```{customer_email}```\n", "\"\"\"\n", "\n", "print(prompt)"]}, {"cell_type": "markdown", "metadata": {}, "source": ["`prompt` \u6784\u9020\u597d\u4e86\uff0c\u6211\u4eec\u53ef\u4ee5\u8c03\u7528`get_completion`\u5f97\u5230\u6211\u4eec\u60f3\u8981\u7684\u7ed3\u679c - \u7528\u5e73\u548c\u5c0a\u91cd\u7684\u8bed\u6c14\uff0c\u7f8e\u5f0f\u82f1\u8bed\u8868\u8fbe\u7684\u6d77\u76d7\u8bed\u8a00\u90ae\u4ef6"]}, {"cell_type": "code", "execution_count": 10, "metadata": {}, "outputs": [], "source": ["response = get_completion(prompt)"]}, {"cell_type": "code", "execution_count": 11, "metadata": {"tags": []}, "outputs": [{"data": {"text/plain": ["\"Oh, I'm really frustrated because the lid of my blender fell off and splattered the milkshake all over the kitchen wall! To make matters worse, the warranty doesn't cover the cost of cleaning the kitchen. I could really use your help right now, buddy!\""]}, "execution_count": 11, "metadata": {}, "output_type": "execute_result"}], "source": ["response"]}, {"cell_type": "markdown", "metadata": {}, "source": ["\u5bf9\u6bd4\u8bed\u8a00\u98ce\u683c\u8f6c\u6362\u524d\u540e\uff0c\u7528\u8bcd\u66f4\u4e3a\u6b63\u5f0f\uff0c\u66ff\u6362\u4e86\u6781\u7aef\u60c5\u7eea\u7684\u8868\u8fbe\uff0c\u5e76\u8868\u8fbe\u4e86\u611f\u8c22\u3002\n", "\n", "\u2728 \u4f60\u53ef\u4ee5\u5c1d\u8bd5\u4fee\u6539\u63d0\u793a\uff0c\u770b\u53ef\u4ee5\u5f97\u5230\u4ec0\u4e48\u4e0d\u4e00\u6837\u7684\u7ed3\u679c\ud83d\ude09"]}, {"cell_type": "markdown", "metadata": {"tags": []}, "source": ["### 2.3 \u4e2d\u6587\u7248"]}, {"cell_type": "code", "execution_count": 12, "metadata": {}, "outputs": [], "source": ["# \u666e\u901a\u8bdd + \u5e73\u9759\u3001\u5c0a\u656c\u7684\u8bed\u8c03\n", "style = \"\"\"\u6b63\u5f0f\u666e\u901a\u8bdd \\\n", "\u7528\u4e00\u4e2a\u5e73\u9759\u3001\u5c0a\u656c\u7684\u8bed\u8c03\n", "\"\"\""]}, {"cell_type": "code", "execution_count": 13, "metadata": {}, "outputs": [{"name": "stdout", "output_type": "stream", "text": ["\u628a\u7531\u4e09\u4e2a\u53cd\u5f15\u53f7\u5206\u9694\u7684\u6587\u672c\u7ffb\u8bd1\u6210\u4e00\u79cd\u6b63\u5f0f\u666e\u901a\u8bdd \u7528\u4e00\u4e2a\u5e73\u9759\u3001\u5c0a\u656c\u7684\u8bed\u8c03\n", "\u98ce\u683c\u3002\n", "\u6587\u672c: ``` \n", "\u963f\uff0c\u6211\u5f88\u751f\u6c14\uff0c\u56e0\u4e3a\u6211\u7684\u6405\u62cc\u673a\u76d6\u6389\u4e86\uff0c\u628a\u5976\u6614\u6e85\u5230\u4e86\u53a8\u623f\u7684\u5899\u4e0a\uff01\u66f4\u7cdf\u7cd5\u7684\u662f\uff0c\u4fdd\u4fee\u4e0d\u5305\u62ec\u6253\u626b\u53a8\u623f\u7684\u8d39\u7528\u3002\u6211\u73b0\u5728\u9700\u8981\u4f60\u7684\u5e2e\u52a9\uff0c\u4f19\u8ba1\uff01\n", "```\n", "\n"]}], "source": ["# \u8981\u6c42\u6a21\u578b\u6839\u636e\u7ed9\u51fa\u7684\u8bed\u8c03\u8fdb\u884c\u8f6c\u5316\n", "prompt = f\"\"\"\u628a\u7531\u4e09\u4e2a\u53cd\u5f15\u53f7\u5206\u9694\u7684\u6587\u672c\\\n", "\u7ffb\u8bd1\u6210\u4e00\u79cd{style}\u98ce\u683c\u3002\n", "\u6587\u672c: ```{customer_email}```\n", "\"\"\"\n", "\n", "print(prompt)\n", "\n"]}, {"cell_type": "code", "execution_count": 14, "metadata": {}, "outputs": [{"data": {"text/plain": ["'\u5c0a\u656c\u7684\u670b\u53cb\u4eec\uff0c\u6211\u611f\u5230\u975e\u5e38\u4e0d\u5b89\uff0c\u56e0\u4e3a\u6211\u7684\u6405\u62cc\u673a\u76d6\u5b50\u4e0d\u614e\u6389\u843d\uff0c\u5bfc\u81f4\u5976\u6614\u6e85\u5230\u4e86\u53a8\u623f\u7684\u5899\u58c1\u4e0a\uff01\u66f4\u52a0\u4ee4\u4eba\u7cdf\u5fc3\u7684\u662f\uff0c\u4fdd\u4fee\u670d\u52a1\u5e76\u4e0d\u5305\u542b\u53a8\u623f\u6e05\u6d01\u7684\u8d39\u7528\u3002\u6b64\u523b\uff0c\u6211\u771f\u8bda\u5730\u8bf7\u6c42\u5404\u4f4d\u7684\u5e2e\u52a9\uff0c\u670b\u53cb\u4eec\uff01'"]}, "execution_count": 14, "metadata": {}, "output_type": "execute_result"}], "source": ["response = get_completion(prompt)\n", "\n", "response"]}, {"cell_type": "markdown", "metadata": {"tags": []}, "source": ["## \u4e09\u3001\u901a\u8fc7LangChain\u4f7f\u7528OpenAI\n", "\n", "\u5728\u524d\u9762\u4e00\u90e8\u5206\uff0c\u6211\u4eec\u901a\u8fc7\u5c01\u88c5\u51fd\u6570`get_completion`\u76f4\u63a5\u8c03\u7528\u4e86OpenAI\u5b8c\u6210\u4e86\u5bf9\u65b9\u8a00\u90ae\u4ef6\u8fdb\u884c\u4e86\u7684\u7ffb\u8bd1\uff0c\u5f97\u5230\u7528\u5e73\u548c\u5c0a\u91cd\u7684\u8bed\u6c14\u3001\u6b63\u5f0f\u7684\u666e\u901a\u8bdd\u8868\u8fbe\u7684\u90ae\u4ef6\u3002\n", "\n", "\u8ba9\u6211\u4eec\u5c1d\u8bd5\u4f7f\u7528LangChain\u6765\u5b9e\u73b0\u76f8\u540c\u7684\u529f\u80fd\u3002"]}, {"cell_type": "code", "execution_count": 15, "metadata": {"tags": []}, "outputs": [], "source": ["# \u5982\u679c\u4f60\u9700\u8981\u67e5\u770b\u5b89\u88c5\u8fc7\u7a0b\u65e5\u5fd7\uff0c\u53ef\u5220\u9664 -q \n", "# --upgrade \u8ba9\u6211\u4eec\u53ef\u4ee5\u5b89\u88c5\u5230\u6700\u65b0\u7248\u672c\u7684 langchain\n", "!pip install -q --upgrade langchain"]}, {"cell_type": "markdown", "metadata": {"tags": []}, "source": ["### 3.1 \u6a21\u578b\n", "\n", "\u4ece`langchain.chat_models`\u5bfc\u5165`OpenAI`\u7684\u5bf9\u8bdd\u6a21\u578b`ChatOpenAI`\u3002 \u9664\u53bbOpenAI\u4ee5\u5916\uff0c`langchain.chat_models`\u8fd8\u96c6\u6210\u4e86\u5176\u4ed6\u5bf9\u8bdd\u6a21\u578b\uff0c\u66f4\u591a\u7ec6\u8282\u53ef\u4ee5\u67e5\u770b[Langchain\u5b98\u65b9\u6587\u6863](https://python.langchain.com/en/latest/modules/models/chat/integrations.html)\u3002"]}, {"cell_type": "code", "execution_count": 16, "metadata": {"tags": []}, "outputs": [], "source": ["from langchain.chat_models import ChatOpenAI"]}, {"cell_type": "code", "execution_count": 17, "metadata": {"tags": []}, "outputs": [{"data": {"text/plain": ["ChatOpenAI(cache=None, verbose=False, callbacks=None, callback_manager=None, tags=None, metadata=None, client=\n", - " OPENAI_API_KEY=\"your_api_key\" \n", - "
\n", - " \n", - " 替换\"your_api_key\"为你自己的 API Key" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "id": "6932bd47-c6d5-4794-8102-a12b84412a93", - "metadata": {}, - "outputs": [], - "source": [ - "# 下载需要的包python-dotenv和openai\n", - "# 如果你需要查看安装过程日志，可删除 -q\n", - "!pip install -q python-dotenv\n", - "!pip install -q openai" - ] - }, - { - "cell_type": "code", - "execution_count": 1, - "id": "10446712-9fa6-4d71-94ce-2ea4cf197e54", - "metadata": {}, - "outputs": [], - "source": [ - "import os\n", - "\n", - "import openai\n", - "from dotenv import find_dotenv, load_dotenv\n", - "\n", - "# 读取本地/项目的环境变量。\n", - "\n", - "# find_dotenv()寻找并定位.env文件的路径\n", - "# load_dotenv()读取该.env文件，并将其中的环境变量加载到当前的运行环境中\n", - "# 如果你设置的是全局的环境变量，这行代码则没有任何作用。\n", - "_ = load_dotenv(find_dotenv())\n", - "\n", - "# 获取环境变量 OPENAI_API_KEY\n", - "openai.api_key = os.environ[\"OPENAI_API_KEY\"]" - ] - }, - { - "cell_type": "markdown", - "id": "1297dcd5", - "metadata": { - "tags": [] - }, - "source": [ - "## 二、对话缓存储存\n", - " " - ] - }, - { - "cell_type": "markdown", - "id": "ffa8cd26-45c5-4bae-90b2-b07d23bf3bb2", - "metadata": {}, - "source": [ - "### 2.1 英文版" - ] - }, - { - "cell_type": "markdown", - "id": "b7e77a3d-7aaa-48c1-b219-19bd6f4eb674", - "metadata": { - "tags": [] - }, - "source": [ - "#### 2.1.1 初始化对话模型" - ] - }, - { - "cell_type": "code", - "execution_count": 47, - "id": "20ad6fe2", - "metadata": { - "height": 98 - }, - "outputs": [], - "source": [ - "from langchain.chains import ConversationChain\n", - "from langchain.chat_models import ChatOpenAI\n", - "from langchain.memory import ConversationBufferMemory" - ] - }, - { - "cell_type": "code", - "execution_count": 48, - "id": "88bdf13d", - "metadata": { - "height": 133 - }, - "outputs": [], - "source": [ - "# 这里我们将参数temperature设置为0.0，从而减少生成答案的随机性。\n", - "# 如果你想要每次得到不一样的有新意的答案，可以尝试增大该参数。\n", - "llm = ChatOpenAI(temperature=0.0) \n", - "\n", - "memory = ConversationBufferMemory()\n", - "\n", - "# 新建一个 ConversationChain Class 实例\n", - "# verbose参数设置为True时，程序会输出更详细的信息，以提供更多的调试或运行时信息。\n", - "# 相反，当将verbose参数设置为False时，程序会以更简洁的方式运行，只输出关键的信息。\n", - "conversation = ConversationChain(llm=llm, memory = memory, verbose=True )" - ] - }, - { - "cell_type": "markdown", - "id": "dea83837", - "metadata": { - "tags": [] - }, - "source": [ - "#### 2.1.2 第一轮对话" - ] - }, - { - "cell_type": "markdown", - "id": "1a3b4c42", - "metadata": {}, - "source": [ - "当我们运行预测(predict)时，生成了一些提示，如下所见，他说“以下是人类和AI之间友好的对话，AI健谈“等等，这实际上是LangChain生成的提示，以使系统进行希望和友好的对话，并且必须保存对话，并提示了当前已完成的模型链。" - ] - }, - { - "cell_type": "code", - "execution_count": 49, - "id": "db24677d", - "metadata": { - "height": 47 - }, - "outputs": [ - { - "name": "stdout", - "output_type": "stream", - "text": [ - "\n", - "\n", - "\u001b[1m> Entering new chain...\u001b[0m\n", - "Prompt after formatting:\n", - "\u001b[32;1m\u001b[1;3mThe following is a friendly conversation between a human and an AI. The AI is talkative and provides lots of specific details from its context. If the AI does not know the answer to a question, it truthfully says it does not know.\n", - "\n", - "Current conversation:\n", - "\n", - "Human: Hi, my name is Andrew\n", - "AI:\u001b[0m\n", - "\n", - "\u001b[1m> Finished chain.\u001b[0m\n" - ] - }, - { - "data": { - "text/plain": [ - "\"Hello Andrew! It's nice to meet you. How can I assist you today?\"" - ] - }, - "execution_count": 49, - "metadata": {}, - "output_type": "execute_result" - } - ], - "source": [ - "conversation.predict(input=\"Hi, my name is Andrew\")" - ] - }, - { - "cell_type": "markdown", - "id": "e71564ad", - "metadata": {}, - "source": [ - "#### 2.1.3 第二轮对话" - ] - }, - { - "cell_type": "markdown", - "id": "54d006bd", - "metadata": {}, - "source": [ - "当我们进行第二轮对话时，它会保留上面的提示" - ] - }, - { - "cell_type": "code", - "execution_count": 50, - "id": "cc3ef937", - "metadata": { - "height": 31 - }, - "outputs": [ - { - "name": "stdout", - "output_type": "stream", - "text": [ - "\n", - "\n", - "\u001b[1m> Entering new chain...\u001b[0m\n", - "Prompt after formatting:\n", - "\u001b[32;1m\u001b[1;3mThe following is a friendly conversation between a human and an AI. The AI is talkative and provides lots of specific details from its context. If the AI does not know the answer to a question, it truthfully says it does not know.\n", - "\n", - "Current conversation:\n", - "Human: Hi, my name is Andrew\n", - "AI: Hello Andrew! It's nice to meet you. How can I assist you today?\n", - "Human: What is 1+1?\n", - "AI:\u001b[0m\n", - "\n", - "\u001b[1m> Finished chain.\u001b[0m\n" - ] - }, - { - "data": { - "text/plain": [ - "'1+1 is equal to 2.'" - ] - }, - "execution_count": 50, - "metadata": {}, - "output_type": "execute_result" - } - ], - "source": [ - "conversation.predict(input=\"What is 1+1?\")" - ] - }, - { - "cell_type": "markdown", - "id": "33cb734b", - "metadata": {}, - "source": [ - "#### 2.1.4 第三轮对话" - ] - }, - { - "cell_type": "markdown", - "id": "0393df3d", - "metadata": {}, - "source": [ - "为了验证他是否记忆了前面的对话内容，我们让他回答前面已经说过的内容（我的名字），可以看到他确实输出了正确的名字，因此这个对话链随着往下进行会越来越长" - ] - }, - { - "cell_type": "code", - "execution_count": 51, - "id": "acf3339a", - "metadata": { - "height": 31 - }, - "outputs": [ - { - "name": "stdout", - "output_type": "stream", - "text": [ - "\n", - "\n", - "\u001b[1m> Entering new chain...\u001b[0m\n", - "Prompt after formatting:\n", - "\u001b[32;1m\u001b[1;3mThe following is a friendly conversation between a human and an AI. The AI is talkative and provides lots of specific details from its context. If the AI does not know the answer to a question, it truthfully says it does not know.\n", - "\n", - "Current conversation:\n", - "Human: Hi, my name is Andrew\n", - "AI: Hello Andrew! It's nice to meet you. How can I assist you today?\n", - "Human: What is 1+1?\n", - "AI: 1+1 is equal to 2.\n", - "Human: What is my name?\n", - "AI:\u001b[0m\n", - "\n", - "\u001b[1m> Finished chain.\u001b[0m\n" - ] - }, - { - "data": { - "text/plain": [ - "'Your name is Andrew.'" - ] - }, - "execution_count": 51, - "metadata": {}, - "output_type": "execute_result" - } - ], - "source": [ - "conversation.predict(input=\"What is my name?\")" - ] - }, - { - "cell_type": "markdown", - "id": "5a96a8d9", - "metadata": {}, - "source": [ - "#### 2.1.5 查看储存缓存\n", - "\n", - "储存缓存(memory.buffer)\n", - "储存了当前为止所有的对话信息" - ] - }, - { - "cell_type": "code", - "execution_count": 52, - "id": "2529400d", - "metadata": { - "height": 31 - }, - "outputs": [ - { - "name": "stdout", - "output_type": "stream", - "text": [ - "Human: Hi, my name is Andrew\n", - "AI: Hello Andrew! It's nice to meet you. How can I assist you today?\n", - "Human: What is 1+1?\n", - "AI: 1+1 is equal to 2.\n", - "Human: What is my name?\n", - "AI: Your name is Andrew.\n" - ] - } - ], - "source": [ - "print(memory.buffer) " - ] - }, - { - "cell_type": "markdown", - "id": "0b5de846", - "metadata": {}, - "source": [ - "也可以通过memory.load_memory_variables({})打印缓存中的历史消息。这里的`{}`是一个空字典，有一些更高级的功能，使用户可以使用更复杂的输入，但我们不会在这个短期课程中讨论它们，所以不要担心为什么这里有一个空的花括号。" - ] - }, - { - "cell_type": "code", - "execution_count": 53, - "id": "5018cb0a", - "metadata": { - "height": 31 - }, - "outputs": [ - { - "name": "stdout", - "output_type": "stream", - "text": [ - "{'history': \"Human: Hi, my name is Andrew\\nAI: Hello Andrew! It's nice to meet you. How can I assist you today?\\nHuman: What is 1+1?\\nAI: 1+1 is equal to 2.\\nHuman: What is my name?\\nAI: Your name is Andrew.\"}\n" - ] - } - ], - "source": [ - "print(memory.load_memory_variables({}))" - ] - }, - { - "cell_type": "markdown", - "id": "07d2e892", - "metadata": {}, - "source": [ - "#### 2.1.6 直接添加内容到储存缓存" - ] - }, - { - "cell_type": "code", - "execution_count": 54, - "id": "14219b70", - "metadata": { - "height": 31 - }, - "outputs": [], - "source": [ - "memory = ConversationBufferMemory() # 新建一个空的对话缓存记忆" - ] - }, - { - "cell_type": "code", - "execution_count": 55, - "id": "a36e9905", - "metadata": { - "height": 48 - }, - "outputs": [], - "source": [ - "memory.save_context({\"input\": \"Hi\"}, {\"output\": \"What's up\"}) # 向缓存区添加指定对话的输入输出" - ] - }, - { - "cell_type": "code", - "execution_count": 56, - "id": "61631b1f", - "metadata": { - "height": 31 - }, - "outputs": [ - { - "name": "stdout", - "output_type": "stream", - "text": [ - "Human: Hi\n", - "AI: What's up\n" - ] - } - ], - "source": [ - "print(memory.buffer) # 查看缓存区结果" - ] - }, - { - "cell_type": "code", - "execution_count": 57, - "id": "a2fdf9ec", - "metadata": { - "height": 31 - }, - "outputs": [ - { - "name": "stdout", - "output_type": "stream", - "text": [ - "{'history': \"Human: Hi\\nAI: What's up\"}\n" - ] - } - ], - "source": [ - "print(memory.load_memory_variables({}))# 再次加载记忆变量" - ] - }, - { - "cell_type": "markdown", - "id": "2ac544f2", - "metadata": {}, - "source": [ - "继续添加新的内容，对话历史都保存下来在了！" - ] - }, - { - "cell_type": "code", - "execution_count": 58, - "id": "7ca79256", - "metadata": { - "height": 64 - }, - "outputs": [], - "source": [ - "memory.save_context({\"input\": \"Not much, just hanging\"}, {\"output\": \"Cool\"})" - ] - }, - { - "cell_type": "code", - "execution_count": 59, - "id": "890a4497", - "metadata": { - "height": 31 - }, - "outputs": [ - { - "data": { - "text/plain": [ - "{'history': \"Human: Hi\\nAI: What's up\\nHuman: Not much, just hanging\\nAI: Cool\"}" - ] - }, - "execution_count": 59, - "metadata": {}, - "output_type": "execute_result" - } - ], - "source": [ - "memory.load_memory_variables({})" - ] - }, - { - "cell_type": "markdown", - "id": "6eddd91a-bf1c-4b82-b99c-c585420e4ecb", - "metadata": {}, - "source": [ - "### 2.2 中文版" - ] - }, - { - "cell_type": "markdown", - "id": "55b3e4e9-7a6a-4a09-9ac3-0096a67849c7", - "metadata": {}, - "source": [ - "#### 2.1.1 初始化对话模型" - ] - }, - { - "cell_type": "code", - "execution_count": 30, - "id": "3577aaff-7edb-40b0-866a-e407e63d55e0", - "metadata": { - "height": 98 - }, - "outputs": [], - "source": [ - "from langchain.chains import ConversationChain\n", - "from langchain.chat_models import ChatOpenAI\n", - "from langchain.memory import ConversationBufferMemory" - ] - }, - { - "cell_type": "code", - "execution_count": 33, - "id": "a77d37ab-1f75-4ae8-8d7c-5066773ead81", - "metadata": { - "height": 133 - }, - "outputs": [ - { - "name": "stdout", - "output_type": "stream", - "text": [ - "\n", - "\n", - "\u001b[1m> Entering new chain...\u001b[0m\n", - "Prompt after formatting:\n", - "\u001b[32;1m\u001b[1;3mThe following is a friendly conversation between a human and an AI. The AI is talkative and provides lots of specific details from its context. If the AI does not know the answer to a question, it truthfully says it does not know.\n", - "\n", - "Current conversation:\n", - "\n", - "Human: 你好, 我叫皮皮鲁\n", - "AI:\u001b[0m\n", - "\n", - "\u001b[1m> Finished chain.\u001b[0m\n" - ] - }, - { - "data": { - "text/plain": [ - "'你好，皮皮鲁！很高兴认识你。我是一个AI助手，可以回答你的问题和提供帮助。有什么我可以帮你的吗？'" - ] - }, - "execution_count": 33, - "metadata": {}, - "output_type": "execute_result" - } - ], - "source": [ - "llm = ChatOpenAI(temperature=0.0) \n", - "\n", - "memory = ConversationBufferMemory()\n", - "\n", - "conversation = ConversationChain(llm=llm, memory = memory, verbose=True )" - ] - }, - { - "cell_type": "markdown", - "id": "747cb539-abc4-4e47-8cb9-1ee608ab07fc", - "metadata": {}, - "source": [ - "#### 2.1.2 第一轮对话" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "id": "99e48462-7a92-4842-bdaa-2a478ba2252c", - "metadata": {}, - "outputs": [], - "source": [ - "conversation.predict(input=\"你好, 我叫皮皮鲁\")" - ] - }, - { - "cell_type": "markdown", - "id": "979d320b-6b20-4722-99db-c48a43711d6c", - "metadata": {}, - "source": [ - "#### 2.1.3 第二轮对话" - ] - }, - { - "cell_type": "code", - "execution_count": 34, - "id": "6ed6e97a-d7ea-4188-a6d7-f91d2a29d14a", - "metadata": {}, - "outputs": [ - { - "name": "stdout", - "output_type": "stream", - "text": [ - "\n", - "\n", - "\u001b[1m> Entering new chain...\u001b[0m\n", - "Prompt after formatting:\n", - "\u001b[32;1m\u001b[1;3mThe following is a friendly conversation between a human and an AI. The AI is talkative and provides lots of specific details from its context. If the AI does not know the answer to a question, it truthfully says it does not know.\n", - "\n", - "Current conversation:\n", - "Human: 你好, 我叫皮皮鲁\n", - "AI: 你好，皮皮鲁！很高兴认识你。我是一个AI助手，可以回答你的问题和提供帮助。有什么我可以帮你的吗？\n", - "Human: 1+1等于多少？\n", - "AI:\u001b[0m\n", - "\n", - "\u001b[1m> Finished chain.\u001b[0m\n" - ] - }, - { - "data": { - "text/plain": [ - "'1+1等于2。'" - ] - }, - "execution_count": 34, - "metadata": {}, - "output_type": "execute_result" - } - ], - "source": [ - "conversation.predict(input=\"1+1等于多少？\")" - ] - }, - { - "cell_type": "markdown", - "id": "8a1fd531-216e-42d8-b226-839747ad7dd3", - "metadata": {}, - "source": [ - "#### 2.1.4 第三轮对话" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "id": "b5dfe488-2758-42c7-9c20-e483b4c22ab8", - "metadata": {}, - "outputs": [ - { - "name": "stdout", - "output_type": "stream", - "text": [ - "\n", - "\n", - "\u001b[1m> Entering new chain...\u001b[0m\n", - "Prompt after formatting:\n", - "\u001b[32;1m\u001b[1;3mThe following is a friendly conversation between a human and an AI. The AI is talkative and provides lots of specific details from its context. If the AI does not know the answer to a question, it truthfully says it does not know.\n", - "\n", - "Current conversation:\n", - "Human: 你好, 我叫皮皮鲁\n", - "AI: 你好，皮皮鲁！很高兴认识你。我是一个AI助手，可以回答你的问题和提供帮助。有什么我可以帮你的吗？\n", - "Human: 1+1等于多少？\n", - "AI: 1+1等于2。\n", - "Human: What is my name?\n", - "AI: 你的名字是皮皮鲁。\n", - "Human: 我叫什么名字？\n", - "AI:\u001b[0m\n", - "\n", - "\u001b[1m> Finished chain.\u001b[0m\n" - ] - }, - { - "data": { - "text/plain": [ - "'你叫皮皮鲁。'" - ] - }, - "execution_count": 36, - "metadata": {}, - "output_type": "execute_result" - } - ], - "source": [ - "conversation.predict(input=\"我叫什么名字？\")" - ] - }, - { - "cell_type": "markdown", - "id": "05d9822e-943d-4905-a1f8-a0d28c215d60", - "metadata": {}, - "source": [ - "#### 2.1.5 查看储存缓存" - ] - }, - { - "cell_type": "code", - "execution_count": 37, - "id": "0795580f-b6b6-47e0-8882-26fe204560bd", - "metadata": { - "height": 31 - }, - "outputs": [ - { - "name": "stdout", - "output_type": "stream", - "text": [ - "Human: 你好, 我叫皮皮鲁\n", - "AI: 你好，皮皮鲁！很高兴认识你。我是一个AI助手，可以回答你的问题和提供帮助。有什么我可以帮你的吗？\n", - "Human: 1+1等于多少？\n", - "AI: 1+1等于2。\n", - "Human: What is my name?\n", - "AI: 你的名字是皮皮鲁。\n", - "Human: 我叫什么名字？\n", - "AI: 你叫皮皮鲁。\n" - ] - } - ], - "source": [ - "print(memory.buffer) " - ] - }, - { - "cell_type": "code", - "execution_count": 38, - "id": "dfe7824f-bd6e-4b95-92e2-1c85c62a92e9", - "metadata": {}, - "outputs": [ - { - "name": "stdout", - "output_type": "stream", - "text": [ - "Human: 你好, 我叫皮皮鲁\n", - "AI: 你好，皮皮鲁！很高兴认识你。我是一个AI助手，可以回答你的问题和提供帮助。有什么我可以帮你的吗？\n", - "Human: 1+1等于多少？\n", - "AI: 1+1等于2。\n", - "Human: What is my name?\n", - "AI: 你的名字是皮皮鲁。\n", - "Human: 我叫什么名字？\n", - "AI: 你叫皮皮鲁。\n" - ] - } - ], - "source": [ - "print(memory.buffer) " - ] - }, - { - "cell_type": "code", - "execution_count": 39, - "id": "48942759-8afb-4aed-80c5-a48952a2b0c0", - "metadata": { - "height": 31 - }, - "outputs": [ - { - "name": "stdout", - "output_type": "stream", - "text": [ - "{'history': 'Human: 你好, 我叫皮皮鲁\\nAI: 你好，皮皮鲁！很高兴认识你。我是一个AI助手，可以回答你的问题和提供帮助。有什么我可以帮你的吗？\\nHuman: 1+1等于多少？\\nAI: 1+1等于2。\\nHuman: What is my name?\\nAI: 你的名字是皮皮鲁。\\nHuman: 我叫什么名字？\\nAI: 你叫皮皮鲁。'}\n" - ] - } - ], - "source": [ - "print(memory.load_memory_variables({}))" - ] - }, - { - "cell_type": "code", - "execution_count": 40, - "id": "4d0c4625-e928-45dc-b8da-4ab865ac5f7e", - "metadata": {}, - "outputs": [ - { - "data": { - "text/plain": [ - "{'history': 'Human: 你好, 我叫皮皮鲁\\nAI: 你好，皮皮鲁！很高兴认识你。我是一个AI助手，可以回答你的问题和提供帮助。有什么我可以帮你的吗？\\nHuman: 1+1等于多少？\\nAI: 1+1等于2。\\nHuman: What is my name?\\nAI: 你的名字是皮皮鲁。\\nHuman: 我叫什么名字？\\nAI: 你叫皮皮鲁。'}" - ] - }, - "execution_count": 40, - "metadata": {}, - "output_type": "execute_result" - } - ], - "source": [ - "memory.load_memory_variables({})" - ] - }, - { - "cell_type": "markdown", - "id": "80f3778e-5fbf-43ed-9df1-d57d98ec6fb0", - "metadata": {}, - "source": [ - "#### 2.1.6 直接添加内容到储存缓存" - ] - }, - { - "cell_type": "code", - "execution_count": 42, - "id": "147b2c30-1662-4b49-aaf8-c228428e5cc6", - "metadata": {}, - "outputs": [ - { - "data": { - "text/plain": [ - "{'history': 'Human: 你好，我叫皮皮鲁\\nAI: 你好啊，我叫鲁西西'}" - ] - }, - "execution_count": 42, - "metadata": {}, - "output_type": "execute_result" - } - ], - "source": [ - "memory = ConversationBufferMemory()\n", - "memory.save_context({\"input\": \"你好，我叫皮皮鲁\"}, {\"output\": \"你好啊，我叫鲁西西\"})\n", - "memory.load_memory_variables({})" - ] - }, - { - "cell_type": "code", - "execution_count": 43, - "id": "6b5e27f0-dad7-41b6-9326-bebf6299638f", - "metadata": { - "height": 64 - }, - "outputs": [], - "source": [ - "memory.save_context({\"input\": \"Not much, just hanging\"}, {\"output\": \"Cool\"})" - ] - }, - { - "cell_type": "code", - "execution_count": 44, - "id": "cfa7c555-06ab-4906-b3dc-906f789e08f5", - "metadata": { - "height": 31 - }, - "outputs": [ - { - "data": { - "text/plain": [ - "{'history': 'Human: 你好，我叫皮皮鲁\\nAI: 你好啊，我叫鲁西西\\nHuman: Not much, just hanging\\nAI: Cool'}" - ] - }, - "execution_count": 44, - "metadata": {}, - "output_type": "execute_result" - } - ], - "source": [ - "memory.load_memory_variables({})" - ] - }, - { - "cell_type": "code", - "execution_count": 45, - "id": "deb33de8-37ea-4180-a73e-0fc456b14eb0", - "metadata": {}, - "outputs": [ - { - "data": { - "text/plain": [ - "{'history': 'Human: 你好，我叫皮皮鲁\\nAI: 你好啊，我叫鲁西西\\nHuman: Not much, just hanging\\nAI: Cool\\nHuman: 很高兴和你成为朋友！\\nAI: 是的，让我们一起去冒险吧！'}" - ] - }, - "execution_count": 45, - "metadata": {}, - "output_type": "execute_result" - } - ], - "source": [ - "memory.save_context({\"input\": \"很高兴和你成为朋友！\"}, {\"output\": \"是的，让我们一起去冒险吧！\"})\n", - "memory.load_memory_variables({})" - ] - }, - { - "cell_type": "markdown", - "id": "10146f1a-0114-4902-8122-d19ae6f7c461", - "metadata": {}, - "source": [ - "### 2.3 总结" - ] - }, - { - "cell_type": "markdown", - "id": "2759b6bc-edb2-4cfe-b0f8-1bf6c4d796f9", - "metadata": {}, - "source": [ - "当我们在使用大型语言模型进行聊天对话时，**大型语言模型本身实际上是无状态的。语言模型本身并不记得到目前为止的历史对话**。每次调用API结点都是独立的。储存(Memory)可以储存到目前为止的所有术语或对话，并将其输入或附加上下文到LLM中用于生成输出。如此看起来就好像它在进行下一轮对话的时候，记得之前说过什么。\n" - ] - }, - { - "cell_type": "markdown", - "id": "cf98e9ff", - "metadata": {}, - "source": [ - "## 三、对话缓存窗口储存\n", - " \n", - "随着对话变得越来越长，所需的内存量也变得非常长。将大量的tokens发送到LLM的成本，也会变得更加昂贵,这也就是为什么API的调用费用，通常是基于它需要处理的tokens数量而收费的。\n", - " \n", - "针对以上问题，LangChain也提供了几种方便的储存方式来保存历史对话。其中，对话缓存窗口储存只保留一个窗口大小的对话。它只使用最近的n次交互。这可以用于保持最近交互的滑动窗口，以便缓冲区不会过大" - ] - }, - { - "cell_type": "markdown", - "id": "b63c9061-9916-4524-b497-93a0aa2b7d06", - "metadata": {}, - "source": [ - "### 3.1 英文版" - ] - }, - { - "cell_type": "markdown", - "id": "641477a4", - "metadata": {}, - "source": [ - "#### 3.1.1 添加两轮对话到窗口储存" - ] - }, - { - "cell_type": "code", - "execution_count": 65, - "id": "3ea6233e", - "metadata": { - "height": 47 - }, - "outputs": [], - "source": [ - "from langchain.memory import ConversationBufferWindowMemory\n", - "\n", - "# k 为窗口参数，k=1表明只保留一个对话记忆\n", - "memory = ConversationBufferWindowMemory(k=1) " - ] - }, - { - "cell_type": "code", - "execution_count": 66, - "id": "dc4553fb", - "metadata": { - "height": 115 - }, - "outputs": [], - "source": [ - "# 向memory添加两轮对话\n", - "memory.save_context({\"input\": \"Hi\"}, {\"output\": \"What's up\"})\n", - "memory.save_context({\"input\": \"Not much, just hanging\"}, {\"output\": \"Cool\"})" - ] - }, - { - "cell_type": "code", - "execution_count": 67, - "id": "6a788403", - "metadata": { - "height": 31 - }, - "outputs": [ - { - "data": { - "text/plain": [ - "{'history': 'Human: Not much, just hanging\\nAI: Cool'}" - ] - }, - "execution_count": 67, - "metadata": {}, - "output_type": "execute_result" - } - ], - "source": [ - "# 并查看记忆变量当前的记录\n", - "memory.load_memory_variables({})" - ] - }, - { - "cell_type": "markdown", - "id": "63bda148", - "metadata": {}, - "source": [ - "#### 3.1.2 在对话链中应用窗口储存" - ] - }, - { - "cell_type": "code", - "execution_count": 68, - "id": "4087bc87", - "metadata": { - "height": 133 - }, - "outputs": [], - "source": [ - "llm = ChatOpenAI(temperature=0.0)\n", - "memory = ConversationBufferWindowMemory(k=1)\n", - "conversation = ConversationChain(llm=llm, memory=memory, verbose=False )" - ] - }, - { - "cell_type": "markdown", - "id": "b6d661e3", - "metadata": {}, - "source": [ - "注意此处！由于这里用的是一个窗口的记忆，因此只能保存一轮的历史消息，因此AI并不能知道你第一轮对话中提到的名字，他最多只能记住上一轮（第二轮）的对话信息" - ] - }, - { - "cell_type": "code", - "execution_count": 69, - "id": "4faaa952", - "metadata": { - "height": 47 - }, - "outputs": [ - { - "data": { - "text/plain": [ - "\"Hello Andrew! It's nice to meet you. How can I assist you today?\"" - ] - }, - "execution_count": 69, - "metadata": {}, - "output_type": "execute_result" - } - ], - "source": [ - "conversation.predict(input=\"Hi, my name is Andrew\")" - ] - }, - { - "cell_type": "code", - "execution_count": 70, - "id": "bb20ddaa", - "metadata": { - "height": 31 - }, - "outputs": [ - { - "data": { - "text/plain": [ - "'1+1 is equal to 2.'" - ] - }, - "execution_count": 70, - "metadata": {}, - "output_type": "execute_result" - } - ], - "source": [ - "conversation.predict(input=\"What is 1+1?\")" - ] - }, - { - "cell_type": "code", - "execution_count": 71, - "id": "489b2194", - "metadata": { - "height": 31 - }, - "outputs": [ - { - "data": { - "text/plain": [ - "\"I'm sorry, but I don't have access to personal information.\"" - ] - }, - "execution_count": 71, - "metadata": {}, - "output_type": "execute_result" - } - ], - "source": [ - "conversation.predict(input=\"What is my name?\")" - ] - }, - { - "cell_type": "markdown", - "id": "88837e7c-cf4b-469e-b820-bbfc49ba876c", - "metadata": {}, - "source": [ - "### 3.2 中文版" - ] - }, - { - "cell_type": "markdown", - "id": "760ec3ad-6959-4a36-b1b5-4fcafe8088ad", - "metadata": {}, - "source": [ - "#### 3.1.1 添加两轮对话到窗口储存" - ] - }, - { - "cell_type": "code", - "execution_count": 72, - "id": "68a2907c", - "metadata": {}, - "outputs": [ - { - "data": { - "text/plain": [ - "{'history': 'Human: 很高兴和你成为朋友！\\nAI: 是的，让我们一起去冒险吧！'}" - ] - }, - "execution_count": 72, - "metadata": {}, - "output_type": "execute_result" - } - ], - "source": [ - "from langchain.memory import ConversationBufferWindowMemory\n", - "\n", - "# k=1表明只保留一个对话记忆\n", - "memory = ConversationBufferWindowMemory(k=1) \n", - "memory.save_context({\"input\": \"你好，我叫皮皮鲁\"}, {\"output\": \"你好啊，我叫鲁西西\"})\n", - "memory.save_context({\"input\": \"很高兴和你成为朋友！\"}, {\"output\": \"是的，让我们一起去冒险吧！\"})\n", - "memory.load_memory_variables({})" - ] - }, - { - "cell_type": "markdown", - "id": "dcabf017-5bfd-4904-9f06-388f994eddc9", - "metadata": {}, - "source": [ - "#### 3.1.2 在对话链中应用窗口储存" - ] - }, - { - "cell_type": "code", - "execution_count": 74, - "id": "1ee854d9", - "metadata": {}, - "outputs": [ - { - "name": "stdout", - "output_type": "stream", - "text": [ - "你好，皮皮鲁！很高兴认识你。我是一个AI助手，可以回答你的问题和提供帮助。有什么我可以帮你的吗？\n", - "1+1等于2。\n", - "很抱歉，我无法知道您的名字。\n" - ] - } - ], - "source": [ - "llm = ChatOpenAI(temperature=0.0)\n", - "memory = ConversationBufferWindowMemory(k=1)\n", - "conversation = ConversationChain(llm=llm, memory=memory, verbose=False )\n", - "print(conversation.predict(input=\"你好, 我叫皮皮鲁\"))\n", - "print(conversation.predict(input=\"1+1等于多少？\"))\n", - "print(conversation.predict(input=\"我叫什么名字？\"))" - ] - }, - { - "cell_type": "markdown", - "id": "d2931b92", - "metadata": {}, - "source": [ - "## 四、对话token缓存储存" - ] - }, - { - "cell_type": "markdown", - "id": "dff5b4c7", - "metadata": {}, - "source": [ - "使用对话token缓存记忆，内存将限制保存的token数量。如果token数量超出指定数目，它会切掉这个对话的早期部分\n", - "以保留与最近的交流相对应的token数量，但不超过token限制。" - ] - }, - { - "cell_type": "code", - "execution_count": 76, - "id": "9f6d063c", - "metadata": { - "height": 31 - }, - "outputs": [], - "source": [ - "!pip install -q tiktoken " - ] - }, - { - "cell_type": "code", - "execution_count": 77, - "id": "fb9020ed", - "metadata": { - "height": 81 - }, - "outputs": [], - "source": [ - "from langchain.llms import OpenAI\n", - "from langchain.memory import ConversationTokenBufferMemory" - ] - }, - { - "cell_type": "markdown", - "id": "f3a84112", - "metadata": {}, - "source": [ - "### 4.1 英文版\n", - "添加对话到Token缓存储存,限制token数量，进行测试" - ] - }, - { - "cell_type": "code", - "execution_count": 78, - "id": "43582ee6", - "metadata": { - "height": 149 - }, - "outputs": [], - "source": [ - "memory = ConversationTokenBufferMemory(llm=llm, max_token_limit=30)\n", - "memory.save_context({\"input\": \"AI is what?!\"}, {\"output\": \"Amazing!\"})\n", - "memory.save_context({\"input\": \"Backpropagation is what?\"}, {\"output\": \"Beautiful!\"})\n", - "memory.save_context({\"input\": \"Chatbots are what?\"}, {\"output\": \"Charming!\"})" - ] - }, - { - "cell_type": "code", - "execution_count": 79, - "id": "284288e1", - "metadata": { - "height": 31 - }, - "outputs": [ - { - "data": { - "text/plain": [ - "{'history': 'AI: Beautiful!\\nHuman: Chatbots are what?\\nAI: Charming!'}" - ] - }, - "execution_count": 79, - "metadata": {}, - "output_type": "execute_result" - } - ], - "source": [ - "memory.load_memory_variables({})" - ] - }, - { - "cell_type": "markdown", - "id": "7b62b2e1", - "metadata": {}, - "source": [ - "可以看到前面超出的的token已经被舍弃了！！！" - ] - }, - { - "cell_type": "markdown", - "id": "f7f6be43", - "metadata": {}, - "source": [ - "### 4.2 中文版" - ] - }, - { - "cell_type": "code", - "execution_count": 80, - "id": "e9191020", - "metadata": {}, - "outputs": [ - { - "data": { - "text/plain": [ - "{'history': 'AI: 轻舟已过万重山。'}" - ] - }, - "execution_count": 80, - "metadata": {}, - "output_type": "execute_result" - } - ], - "source": [ - "memory = ConversationTokenBufferMemory(llm=llm, max_token_limit=30)\n", - "memory.save_context({\"input\": \"朝辞白帝彩云间，\"}, {\"output\": \"千里江陵一日还。\"})\n", - "memory.save_context({\"input\": \"两岸猿声啼不住，\"}, {\"output\": \"轻舟已过万重山。\"})\n", - "memory.load_memory_variables({})" - ] - }, - { - "cell_type": "markdown", - "id": "fb08ef4a-876f-422a-81f9-4805288e5955", - "metadata": {}, - "source": [ - "### 4.3 补充" - ] - }, - { - "cell_type": "markdown", - "id": "5e4d918b", - "metadata": {}, - "source": [ - "ChatGPT使用一种基于字节对编码（Byte Pair Encoding，BPE）的方法来进行tokenization（将输入文本拆分为token）。BPE是一种常见的tokenization技术，它将输入文本分割成较小的子词单元。 \n", - "\n", - "OpenAI在其官方GitHub上公开了一个最新的开源Python库 [tiktoken](https://github.com/openai/tiktoken)，这个库主要是用来计算tokens数量的。相比较HuggingFace的tokenizer，其速度提升了好几倍。\n", - "\n", - "具体token计算方式,特别是汉字和英文单词的token区别，具体课参考[知乎文章](https://www.zhihu.com/question/594159910) 。" - ] - }, - { - "cell_type": "markdown", - "id": "5ff55d5d", - "metadata": {}, - "source": [ - "## 五、对话摘要缓存储存" - ] - }, - { - "cell_type": "markdown", - "id": "7d39b83a", - "metadata": {}, - "source": [ - "对话摘要缓存储存，**使用LLM编写到目前为止历史对话的摘要**，并将其保存" - ] - }, - { - "cell_type": "code", - "execution_count": 82, - "id": "72dcf8b1", - "metadata": { - "height": 64 - }, - "outputs": [], - "source": [ - "from langchain.chains import ConversationChain\n", - "from langchain.chat_models import ChatOpenAI\n", - "from langchain.memory import ConversationSummaryBufferMemory" - ] - }, - { - "cell_type": "markdown", - "id": "243b213e-ce17-46a0-8652-03658ca58dd8", - "metadata": {}, - "source": [ - "### 5.1 英文版" - ] - }, - { - "cell_type": "markdown", - "id": "6572ef39", - "metadata": {}, - "source": [ - "#### 5.1.1 使用对话摘要缓存储存\n", - "\n", - "创建一个长字符串，其中包含某人的日程安排" - ] - }, - { - "cell_type": "code", - "execution_count": 84, - "id": "4a5b238f", - "metadata": { - "height": 285 - }, - "outputs": [], - "source": [ - "# 创建一个长字符串\n", - "schedule = \"There is a meeting at 8am with your product team. \\\n", - "You will need your powerpoint presentation prepared. \\\n", - "9am-12pm have time to work on your LangChain \\\n", - "project which will go quickly because Langchain is such a powerful tool. \\\n", - "At Noon, lunch at the italian resturant with a customer who is driving \\\n", - "from over an hour away to meet you to understand the latest in AI. \\\n", - "Be sure to bring your laptop to show the latest LLM demo.\"\n", - "\n", - "# 使用对话摘要缓存记忆\n", - "llm = ChatOpenAI(temperature=0.0)\n", - "memory = ConversationSummaryBufferMemory(llm=llm, max_token_limit=100) \n", - "memory.save_context({\"input\": \"Hello\"}, {\"output\": \"What's up\"})\n", - "memory.save_context({\"input\": \"Not much, just hanging\"}, {\"output\": \"Cool\"})\n", - "memory.save_context(\n", - " {\"input\": \"What is on the schedule today?\"}, {\"output\": f\"{schedule}\"}\n", - ")" - ] - }, - { - "cell_type": "code", - "execution_count": 89, - "id": "15226a41-ab36-43a0-93f7-c03c6b374936", - "metadata": {}, - "outputs": [ - { - "name": "stdout", - "output_type": "stream", - "text": [ - "System: The human and AI exchange greetings. The human asks about the schedule for the day. The AI provides a detailed schedule, including a meeting with the product team, work on the LangChain project, and a lunch meeting with a customer interested in AI. The AI emphasizes the importance of bringing a laptop to showcase the latest LLM demo during the lunch meeting.\n" - ] - } - ], - "source": [ - "print(memory.load_memory_variables({})['history'])" - ] - }, - { - "cell_type": "markdown", - "id": "7ccb97b6", - "metadata": {}, - "source": [ - "#### 5.1.2 基于对话摘要缓存储存的对话链\n", - "基于上面的对话摘要缓存储存，新建一个对话链" - ] - }, - { - "cell_type": "code", - "execution_count": 90, - "id": "6728edba", - "metadata": { - "height": 99 - }, - "outputs": [], - "source": [ - "conversation = ConversationChain(llm=llm, memory=memory, verbose=True)" - ] - }, - { - "cell_type": "code", - "execution_count": 91, - "id": "9a221b1d", - "metadata": { - "height": 47 - }, - "outputs": [ - { - "name": "stdout", - "output_type": "stream", - "text": [ - "\n", - "\n", - "\u001b[1m> Entering new chain...\u001b[0m\n", - "Prompt after formatting:\n", - "\u001b[32;1m\u001b[1;3mThe following is a friendly conversation between a human and an AI. The AI is talkative and provides lots of specific details from its context. If the AI does not know the answer to a question, it truthfully says it does not know.\n", - "\n", - "Current conversation:\n", - "System: The human and AI exchange greetings. The human asks about the schedule for the day. The AI provides a detailed schedule, including a meeting with the product team, work on the LangChain project, and a lunch meeting with a customer interested in AI. The AI emphasizes the importance of bringing a laptop to showcase the latest LLM demo during the lunch meeting.\n", - "Human: What would be a good demo to show?\n", - "AI:\u001b[0m\n", - "\n", - "\u001b[1m> Finished chain.\u001b[0m\n" - ] - }, - { - "data": { - "text/plain": [ - "'A good demo to show during the lunch meeting with the customer interested in AI would be the latest LLM (Language Model) demo. The LLM is a cutting-edge AI model that can generate human-like text based on a given prompt. It has been trained on a vast amount of data and can generate coherent and contextually relevant responses. By showcasing the LLM demo, you can demonstrate the capabilities of AI in natural language processing and how it can be applied to various industries and use cases.'" - ] - }, - "execution_count": 91, - "metadata": {}, - "output_type": "execute_result" - } - ], - "source": [ - "conversation.predict(input=\"What would be a good demo to show?\")" - ] - }, - { - "cell_type": "code", - "execution_count": 92, - "id": "bb582617", - "metadata": { - "height": 31 - }, - "outputs": [ - { - "name": "stdout", - "output_type": "stream", - "text": [ - "System: The human and AI exchange greetings. The human asks about the schedule for the day. The AI provides a detailed schedule, including a meeting with the product team, work on the LangChain project, and a lunch meeting with a customer interested in AI. The AI emphasizes the importance of bringing a laptop to showcase the latest LLM demo during the lunch meeting. A good demo to show during the lunch meeting with the customer interested in AI would be the latest LLM (Language Model) demo. The LLM is a cutting-edge AI model that can generate human-like text based on a given prompt. It has been trained on a vast amount of data and can generate coherent and contextually relevant responses. By showcasing the LLM demo, you can demonstrate the capabilities of AI in natural language processing and how it can be applied to various industries and use cases.\n" - ] - } - ], - "source": [ - "print(memory.load_memory_variables({})['history'])" - ] - }, - { - "cell_type": "markdown", - "id": "4ba827aa", - "metadata": { - "height": 31 - }, - "source": [ - "### 5.2 中文版" - ] - }, - { - "cell_type": "markdown", - "id": "64898f33-c538-4e68-b008-7123870b692b", - "metadata": {}, - "source": [ - "#### 5.2.1 使用对话摘要缓存储存\n", - "\n", - "创建一个长字符串，其中包含某人的日程安排" - ] - }, - { - "cell_type": "code", - "execution_count": 97, - "id": "2c07922b", - "metadata": { - "height": 31 - }, - "outputs": [], - "source": [ - "# 创建一个长字符串\n", - "schedule = \"在八点你和你的产品团队有一个会议。 \\\n", - "你需要做一个PPT。 \\\n", - "上午9点到12点你需要忙于LangChain。\\\n", - "Langchain是一个有用的工具，因此你的项目进展的非常快。\\\n", - "中午，在意大利餐厅与一位开车来的顾客共进午餐 \\\n", - "走了一个多小时的路程与你见面，只为了解最新的 AI。 \\\n", - "确保你带了笔记本电脑可以展示最新的 LLM 样例.\"\n", - "\n", - "llm = ChatOpenAI(temperature=0.0)\n", - "memory = ConversationSummaryBufferMemory(llm=llm, max_token_limit=100)\n", - "memory.save_context({\"input\": \"你好，我叫皮皮鲁\"}, {\"output\": \"你好啊，我叫鲁西西\"})\n", - "memory.save_context({\"input\": \"很高兴和你成为朋友！\"}, {\"output\": \"是的，让我们一起去冒险吧！\"})\n", - "memory.save_context({\"input\": \"今天的日程安排是什么？\"}, {\"output\": f\"{schedule}\"})" - ] - }, - { - "cell_type": "code", - "execution_count": 98, - "id": "17424a12-430f-4529-9067-300978c6169e", - "metadata": {}, - "outputs": [ - { - "name": "stdout", - "output_type": "stream", - "text": [ - "System: The human introduces themselves as Pipilu and the AI introduces themselves as Luxixi. They express happiness at becoming friends and decide to go on an adventure together. The human asks about the schedule for the day. The AI informs them that they have a meeting with their product team at 8 o'clock and need to prepare a PowerPoint presentation. From 9 am to 12 pm, they will be busy with LangChain, a useful tool that helps their project progress quickly. At noon, they will have lunch with a customer who has driven for over an hour just to learn about the latest AI. The AI advises the human to bring their laptop to showcase the latest LLM samples.\n" - ] - } - ], - "source": [ - "print(memory.load_memory_variables({})['history'])" - ] - }, - { - "cell_type": "markdown", - "id": "9e29a956-607a-4247-9eb5-01285a370991", - "metadata": {}, - "source": [ - "#### 5.1.2 基于对话摘要缓存储存的对话链\n", - "基于上面的对话摘要缓存储存，新建一个对话链" - ] - }, - { - "cell_type": "code", - "execution_count": 99, - "id": "52696c8c", - "metadata": { - "height": 31 - }, - "outputs": [], - "source": [ - "conversation = ConversationChain(llm=llm, memory=memory, verbose=True)" - ] - }, - { - "cell_type": "code", - "execution_count": 100, - "id": "48690d13", - "metadata": { - "height": 31 - }, - "outputs": [ - { - "name": "stdout", - "output_type": "stream", - "text": [ - "\n", - "\n", - "\u001b[1m> Entering new chain...\u001b[0m\n", - "Prompt after formatting:\n", - "\u001b[32;1m\u001b[1;3mThe following is a friendly conversation between a human and an AI. The AI is talkative and provides lots of specific details from its context. If the AI does not know the answer to a question, it truthfully says it does not know.\n", - "\n", - "Current conversation:\n", - "System: The human introduces themselves as Pipilu and the AI introduces themselves as Luxixi. They express happiness at becoming friends and decide to go on an adventure together. The human asks about the schedule for the day. The AI informs them that they have a meeting with their product team at 8 o'clock and need to prepare a PowerPoint presentation. From 9 am to 12 pm, they will be busy with LangChain, a useful tool that helps their project progress quickly. At noon, they will have lunch with a customer who has driven for over an hour just to learn about the latest AI. The AI advises the human to bring their laptop to showcase the latest LLM samples.\n", - "Human: 展示什么样的样例最好呢？\n", - "AI:\u001b[0m\n", - "\n", - "\u001b[1m> Finished chain.\u001b[0m\n" - ] - }, - { - "data": { - "text/plain": [ - "'展示一些具有多样性和创新性的样例可能是最好的选择。你可以选择展示一些基于图像识别的样例，比如人脸识别、物体识别等。另外，你也可以展示一些自然语言处理方面的样例，比如文本生成、情感分析等。最重要的是选择那些能够展示出你们团队的技术实力和创造力的样例。'" - ] - }, - "execution_count": 100, - "metadata": {}, - "output_type": "execute_result" - } - ], - "source": [ - "conversation.predict(input=\"展示什么样的样例最好呢？\")" - ] - }, - { - "cell_type": "code", - "execution_count": 101, - "id": "85bba1f8", - "metadata": { - "height": 31 - }, - "outputs": [ - { - "data": { - "text/plain": [ - "{'history': \"System: The human introduces themselves as Pipilu and the AI introduces themselves as Luxixi. They express happiness at becoming friends and decide to go on an adventure together. The human asks about the schedule for the day. The AI informs them that they have a meeting with their product team at 8 o'clock and need to prepare a PowerPoint presentation. From 9 am to 12 pm, they will be busy with LangChain, a useful tool that helps their project progress quickly. At noon, they will have lunch with a customer who has driven for over an hour just to learn about the latest AI. The AI advises the human to bring their laptop to showcase the latest LLM samples. The human asks what kind of samples would be best to showcase. The AI suggests that showcasing diverse and innovative samples would be the best choice. They recommend selecting samples based on image recognition, such as face recognition and object recognition. Additionally, they suggest showcasing samples related to natural language processing, such as text generation and sentiment analysis. The AI emphasizes the importance of choosing samples that demonstrate the team's technical expertise and creativity.\"}" - ] - }, - "execution_count": 101, - "metadata": {}, - "output_type": "execute_result" - } - ], - "source": [ - "memory.load_memory_variables({}) # 摘要记录更新了" - ] - } - ], - "metadata": { - "kernelspec": { - "display_name": "Python 3 (ipykernel)", - "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": "3.9.12" - }, - "toc": { - "base_numbering": 1, - "nav_menu": {}, - "number_sections": false, - "sideBar": true, - "skip_h1_title": false, - "title_cell": "", - "title_sidebar": "Contents", - "toc_cell": false, - "toc_position": {}, - "toc_section_display": true, - "toc_window_display": true - } - }, - "nbformat": 4, - "nbformat_minor": 5 -} +{"cells": [{"cell_type": "markdown", "id": "a786c77c", "metadata": {"tags": []}, "source": ["# \u7b2c\u4e09\u7ae0 \u50a8\u5b58\n", "\n", " - [\u4e00\u3001\u8bbe\u7f6eOpenAI API Key](#\u4e00\u3001\u8bbe\u7f6eOpenAI-API-Key)\n", " - [\u4e8c\u3001\u5bf9\u8bdd\u7f13\u5b58\u50a8\u5b58](#\u4e8c\u3001\u5bf9\u8bdd\u7f13\u5b58\u50a8\u5b58)\n", " - [2.1 \u82f1\u6587\u7248](#2.1-\u82f1\u6587\u7248)\n", " - [2.1.1 \u521d\u59cb\u5316\u5bf9\u8bdd\u6a21\u578b](#2.1.1-\u521d\u59cb\u5316\u5bf9\u8bdd\u6a21\u578b)\n", " - [2.1.2 \u7b2c\u4e00\u8f6e\u5bf9\u8bdd](#2.1.2-\u7b2c\u4e00\u8f6e\u5bf9\u8bdd)\n", " - [2.1.3 \u7b2c\u4e8c\u8f6e\u5bf9\u8bdd](#2.1.3-\u7b2c\u4e8c\u8f6e\u5bf9\u8bdd)\n", " - [2.1.4 \u7b2c\u4e09\u8f6e\u5bf9\u8bdd](#2.1.4-\u7b2c\u4e09\u8f6e\u5bf9\u8bdd)\n", " - [2.1.5 \u67e5\u770b\u50a8\u5b58\u7f13\u5b58](#2.1.5-\u67e5\u770b\u50a8\u5b58\u7f13\u5b58)\n", " - [2.1.6 \u76f4\u63a5\u6dfb\u52a0\u5185\u5bb9\u5230\u50a8\u5b58\u7f13\u5b58](#2.1.6-\u76f4\u63a5\u6dfb\u52a0\u5185\u5bb9\u5230\u50a8\u5b58\u7f13\u5b58)\n", " - [2.2 \u4e2d\u6587\u7248](#2.2-\u4e2d\u6587\u7248)\n", " - [2.1.1 \u521d\u59cb\u5316\u5bf9\u8bdd\u6a21\u578b](#2.1.1-\u521d\u59cb\u5316\u5bf9\u8bdd\u6a21\u578b)\n", " - [2.1.2 \u7b2c\u4e00\u8f6e\u5bf9\u8bdd](#2.1.2-\u7b2c\u4e00\u8f6e\u5bf9\u8bdd)\n", " - [2.1.3 \u7b2c\u4e8c\u8f6e\u5bf9\u8bdd](#2.1.3-\u7b2c\u4e8c\u8f6e\u5bf9\u8bdd)\n", " - [2.1.4 \u7b2c\u4e09\u8f6e\u5bf9\u8bdd](#2.1.4-\u7b2c\u4e09\u8f6e\u5bf9\u8bdd)\n", " - [2.1.5 \u67e5\u770b\u50a8\u5b58\u7f13\u5b58](#2.1.5-\u67e5\u770b\u50a8\u5b58\u7f13\u5b58)\n", " - [2.1.6 \u76f4\u63a5\u6dfb\u52a0\u5185\u5bb9\u5230\u50a8\u5b58\u7f13\u5b58](#2.1.6-\u76f4\u63a5\u6dfb\u52a0\u5185\u5bb9\u5230\u50a8\u5b58\u7f13\u5b58)\n", " - [2.3 \u603b\u7ed3](#2.3-\u603b\u7ed3)\n", " - [\u4e09\u3001\u5bf9\u8bdd\u7f13\u5b58\u7a97\u53e3\u50a8\u5b58](#\u4e09\u3001\u5bf9\u8bdd\u7f13\u5b58\u7a97\u53e3\u50a8\u5b58)\n", " - [3.1 \u82f1\u6587\u7248](#3.1-\u82f1\u6587\u7248)\n", " - [3.1.1 \u6dfb\u52a0\u4e24\u8f6e\u5bf9\u8bdd\u5230\u7a97\u53e3\u50a8\u5b58](#3.1.1-\u6dfb\u52a0\u4e24\u8f6e\u5bf9\u8bdd\u5230\u7a97\u53e3\u50a8\u5b58)\n", " - [3.1.2 \u5728\u5bf9\u8bdd\u94fe\u4e2d\u5e94\u7528\u7a97\u53e3\u50a8\u5b58](#3.1.2-\u5728\u5bf9\u8bdd\u94fe\u4e2d\u5e94\u7528\u7a97\u53e3\u50a8\u5b58)\n", " - [3.2 \u4e2d\u6587\u7248](#3.2-\u4e2d\u6587\u7248)\n", " - [3.1.1 \u6dfb\u52a0\u4e24\u8f6e\u5bf9\u8bdd\u5230\u7a97\u53e3\u50a8\u5b58](#3.1.1-\u6dfb\u52a0\u4e24\u8f6e\u5bf9\u8bdd\u5230\u7a97\u53e3\u50a8\u5b58)\n", " - [3.1.2 \u5728\u5bf9\u8bdd\u94fe\u4e2d\u5e94\u7528\u7a97\u53e3\u50a8\u5b58](#3.1.2-\u5728\u5bf9\u8bdd\u94fe\u4e2d\u5e94\u7528\u7a97\u53e3\u50a8\u5b58)\n", " - [\u56db\u3001\u5bf9\u8bddtoken\u7f13\u5b58\u50a8\u5b58](#\u56db\u3001\u5bf9\u8bddtoken\u7f13\u5b58\u50a8\u5b58)\n", " - [4.1 \u82f1\u6587\u7248](#4.1-\u82f1\u6587\u7248)\n", " - [4.2 \u4e2d\u6587\u7248](#4.2-\u4e2d\u6587\u7248)\n", " - [4.3 \u8865\u5145](#4.3-\u8865\u5145)\n", " - [\u4e94\u3001\u5bf9\u8bdd\u6458\u8981\u7f13\u5b58\u50a8\u5b58](#\u4e94\u3001\u5bf9\u8bdd\u6458\u8981\u7f13\u5b58\u50a8\u5b58)\n", " - [5.1 \u82f1\u6587\u7248](#5.1-\u82f1\u6587\u7248)\n", " - [5.1.1 \u4f7f\u7528\u5bf9\u8bdd\u6458\u8981\u7f13\u5b58\u50a8\u5b58](#5.1.1-\u4f7f\u7528\u5bf9\u8bdd\u6458\u8981\u7f13\u5b58\u50a8\u5b58)\n", " - [5.1.2 \u57fa\u4e8e\u5bf9\u8bdd\u6458\u8981\u7f13\u5b58\u50a8\u5b58\u7684\u5bf9\u8bdd\u94fe](#5.1.2-\u57fa\u4e8e\u5bf9\u8bdd\u6458\u8981\u7f13\u5b58\u50a8\u5b58\u7684\u5bf9\u8bdd\u94fe)\n", " - [5.2 \u4e2d\u6587\u7248](#5.2-\u4e2d\u6587\u7248)\n", " - [5.2.1 \u4f7f\u7528\u5bf9\u8bdd\u6458\u8981\u7f13\u5b58\u50a8\u5b58](#5.2.1-\u4f7f\u7528\u5bf9\u8bdd\u6458\u8981\u7f13\u5b58\u50a8\u5b58)\n", " - [5.1.2 \u57fa\u4e8e\u5bf9\u8bdd\u6458\u8981\u7f13\u5b58\u50a8\u5b58\u7684\u5bf9\u8bdd\u94fe](#5.1.2-\u57fa\u4e8e\u5bf9\u8bdd\u6458\u8981\u7f13\u5b58\u50a8\u5b58\u7684\u5bf9\u8bdd\u94fe)\n"]}, {"cell_type": "markdown", "id": "7e10db6f", "metadata": {}, "source": ["\u5f53\u4f60\u4e0e\u90a3\u4e9b\u8bed\u8a00\u6a21\u578b\u8fdb\u884c\u4ea4\u4e92\u7684\u65f6\u5019\uff0c\u4ed6\u4eec\u4e0d\u4f1a\u8bb0\u5f97\u4f60\u4e4b\u524d\u548c\u4ed6\u8fdb\u884c\u7684\u4ea4\u6d41\u5185\u5bb9\uff0c\u8fd9\u5728\u6211\u4eec\u6784\u5efa\u4e00\u4e9b\u5e94\u7528\u7a0b\u5e8f\uff08\u5982\u804a\u5929\u673a\u5668\u4eba\uff09\u7684\u65f6\u5019\uff0c\u662f\u4e00\u4e2a\u5f88\u5927\u7684\u95ee\u9898 -- \u663e\u5f97\u4e0d\u591f\u667a\u80fd\uff01\u56e0\u6b64\uff0c\u5728\u672c\u8282\u4e2d\u6211\u4eec\u5c06\u4ecb\u7ecd LangChain \u4e2d\u7684\u50a8\u5b58\u6a21\u5757\uff0c\u5373\u5982\u4f55\u5c06\u5148\u524d\u7684\u5bf9\u8bdd\u5d4c\u5165\u5230\u8bed\u8a00\u6a21\u578b\u4e2d\u7684\uff0c\u4f7f\u5176\u5177\u6709\u8fde\u7eed\u5bf9\u8bdd\u7684\u80fd\u529b\u3002\n", "\n", "\u5f53\u4f7f\u7528 LangChain \u4e2d\u7684\u50a8\u5b58(Memory)\u6a21\u5757\u65f6\uff0c\u5b83\u53ef\u4ee5\u5e2e\u52a9\u4fdd\u5b58\u548c\u7ba1\u7406\u5386\u53f2\u804a\u5929\u6d88\u606f\uff0c\u4ee5\u53ca\u6784\u5efa\u5173\u4e8e\u7279\u5b9a\u5b9e\u4f53\u7684\u77e5\u8bc6\u3002\u8fd9\u4e9b\u7ec4\u4ef6\u53ef\u4ee5\u8de8\u591a\u8f6e\u5bf9\u8bdd\u50a8\u5b58\u4fe1\u606f\uff0c\u5e76\u5141\u8bb8\u5728\u5bf9\u8bdd\u671f\u95f4\u8ddf\u8e2a\u7279\u5b9a\u4fe1\u606f\u548c\u4e0a\u4e0b\u6587\u3002\n", "\n", "LangChain \u63d0\u4f9b\u4e86\u591a\u79cd\u50a8\u5b58\u7c7b\u578b\u3002\u5176\u4e2d\uff0c\u7f13\u51b2\u533a\u50a8\u5b58\u5141\u8bb8\u4fdd\u7559\u6700\u8fd1\u7684\u804a\u5929\u6d88\u606f\uff0c\u6458\u8981\u50a8\u5b58\u5219\u63d0\u4f9b\u4e86\u5bf9\u6574\u4e2a\u5bf9\u8bdd\u7684\u6458\u8981\u3002\u5b9e\u4f53\u50a8\u5b58 \u5219\u5141\u8bb8\u5728\u591a\u8f6e\u5bf9\u8bdd\u4e2d\u4fdd\u7559\u6709\u5173\u7279\u5b9a\u5b9e\u4f53\u7684\u4fe1\u606f\u3002\u8fd9\u4e9b\u8bb0\u5fc6\u7ec4\u4ef6\u90fd\u662f\u6a21\u5757\u5316\u7684\uff0c\u53ef\u4e0e\u5176\u4ed6\u7ec4\u4ef6\u7ec4\u5408\u4f7f\u7528\uff0c\u4ece\u800c\u589e\u5f3a\u673a\u5668\u4eba\u7684\u5bf9\u8bdd\u7ba1\u7406\u80fd\u529b\u3002\u50a8\u5b58\u6a21\u5757\u53ef\u4ee5\u901a\u8fc7\u7b80\u5355\u7684API\u8c03\u7528\u6765\u8bbf\u95ee\u548c\u66f4\u65b0\uff0c\u5141\u8bb8\u5f00\u53d1\u4eba\u5458\u66f4\u8f7b\u677e\u5730\u5b9e\u73b0\u5bf9\u8bdd\u5386\u53f2\u8bb0\u5f55\u7684\u7ba1\u7406\u548c\u7ef4\u62a4\u3002\n", "\n", "\u6b64\u6b21\u8bfe\u7a0b\u4e3b\u8981\u4ecb\u7ecd\u5176\u4e2d\u56db\u79cd\u50a8\u5b58\u6a21\u5757\uff0c\u5176\u4ed6\u6a21\u5757\u53ef\u67e5\u770b\u6587\u6863\u5b66\u4e60\u3002\n", "- \u5bf9\u8bdd\u7f13\u5b58\u50a8\u5b58 (ConversationBufferMemory\uff09\n", "- \u5bf9\u8bdd\u7f13\u5b58\u7a97\u53e3\u50a8\u5b58 (ConversationBufferWindowMemory\uff09\n", "- \u5bf9\u8bdd\u4ee4\u724c\u7f13\u5b58\u50a8\u5b58 (ConversationTokenBufferMemory\uff09\n", "- \u5bf9\u8bdd\u6458\u8981\u7f13\u5b58\u50a8\u5b58 (ConversationSummaryBufferMemory\uff09\n", "\n", "\u5728LangChain\u4e2d\uff0c\u50a8\u5b58 \u6307\u7684\u662f\u5927\u8bed\u8a00\u6a21\u578b\uff08LLM\uff09\u7684\u77ed\u671f\u8bb0\u5fc6\u3002\u4e3a\u4ec0\u4e48\u662f\u77ed\u671f\u8bb0\u5fc6\uff1f\u90a3\u662f\u56e0\u4e3aLLM\u8bad\u7ec3\u597d\u4e4b\u540e (\u83b7\u5f97\u4e86\u4e00\u4e9b\u957f\u671f\u8bb0\u5fc6)\uff0c\u5b83\u7684\u53c2\u6570\u4fbf\u4e0d\u4f1a\u56e0\u4e3a\u7528\u6237\u7684\u8f93\u5165\u800c\u53d1\u751f\u6539\u53d8\u3002\u5f53\u7528\u6237\u4e0e\u8bad\u7ec3\u597d\u7684LLM\u8fdb\u884c\u5bf9\u8bdd\u65f6\uff0cLLM\u4f1a\u6682\u65f6\u8bb0\u4f4f\u7528\u6237\u7684\u8f93\u5165\u548c\u5b83\u5df2\u7ecf\u751f\u6210\u7684\u8f93\u51fa\uff0c\u4ee5\u4fbf\u9884\u6d4b\u4e4b\u540e\u7684\u8f93\u51fa\uff0c\u800c\u6a21\u578b\u8f93\u51fa\u5b8c\u6bd5\u540e\uff0c\u5b83\u4fbf\u4f1a\u201c\u9057\u5fd8\u201d\u4e4b\u524d\u7528\u6237\u7684\u8f93\u5165\u548c\u5b83\u7684\u8f93\u51fa\u3002\u56e0\u6b64\uff0c\u4e4b\u524d\u7684\u8fd9\u4e9b\u4fe1\u606f\u53ea\u80fd\u79f0\u4f5c\u4e3aLLM\u7684\u77ed\u671f\u8bb0\u5fc6\u3002 \n", " \n", "\u4e3a\u4e86\u5ef6\u957fLLM\u77ed\u671f\u8bb0\u5fc6\u7684\u4fdd\u7559\u65f6\u95f4\uff0c\u5219\u9700\u8981\u501f\u52a9\u4e00\u4e9b\u5916\u90e8\u50a8\u5b58\u65b9\u5f0f\u6765\u8fdb\u884c\u8bb0\u5fc6\uff0c\u4ee5\u4fbf\u5728\u7528\u6237\u4e0eLLM\u5bf9\u8bdd\u4e2d\uff0cLLM\u80fd\u591f\u5c3d\u53ef\u80fd\u7684\u77e5\u9053\u7528\u6237\u4e0e\u5b83\u6240\u8fdb\u884c\u7684\u5386\u53f2\u5bf9\u8bdd\u4fe1\u606f\u3002 "]}, {"cell_type": "markdown", "id": "1ca56e6b-1e07-4405-a1ca-f4237f20fa75", "metadata": {"jp-MarkdownHeadingCollapsed": true, "tags": []}, "source": ["## \u4e00\u3001\u8bbe\u7f6eOpenAI API Key\n", "\n", "\u767b\u9646 [OpenAI \u8d26\u6237](https://platform.openai.com/account/api-keys) \u83b7\u53d6API Key\uff0c\u7136\u540e\u5c06\u5176\u8bbe\u7f6e\u4e3a\u73af\u5883\u53d8\u91cf\u3002\n", "\n", "- \u5982\u679c\u4f60\u60f3\u8981\u8bbe\u7f6e\u4e3a\u5168\u5c40\u73af\u5883\u53d8\u91cf\uff0c\u53ef\u4ee5\u53c2\u8003[\u77e5\u4e4e\u6587\u7ae0](https://zhuanlan.zhihu.com/p/627665725)\u3002\n", "- \u5982\u679c\u4f60\u60f3\u8981\u8bbe\u7f6e\u4e3a\u672c\u5730/\u9879\u76ee\u73af\u5883\u53d8\u91cf\uff0c\u5728\u672c\u6587\u4ef6\u76ee\u5f55\u4e0b\u521b\u5efa`.env`\u6587\u4ef6, \u6253\u5f00\u6587\u4ef6\u8f93\u5165\u4ee5\u4e0b\u5185\u5bb9\u3002\n", "\n", "\n", " OPENAI_API_KEY=\"your_api_key\" \n", "
\n", " \n", " \u66ff\u6362\"your_api_key\"\u4e3a\u4f60\u81ea\u5df1\u7684 API Key"]}, {"cell_type": "code", "execution_count": null, "id": "6932bd47-c6d5-4794-8102-a12b84412a93", "metadata": {}, "outputs": [], "source": ["# \u4e0b\u8f7d\u9700\u8981\u7684\u5305python-dotenv\u548copenai\n", "# \u5982\u679c\u4f60\u9700\u8981\u67e5\u770b\u5b89\u88c5\u8fc7\u7a0b\u65e5\u5fd7\uff0c\u53ef\u5220\u9664 -q\n", "!pip install -q python-dotenv\n", "!pip install -q openai"]}, {"cell_type": "code", "execution_count": 1, "id": "10446712-9fa6-4d71-94ce-2ea4cf197e54", "metadata": {}, "outputs": [], "source": ["import os\n", "\n", "import openai\n", "from dotenv import find_dotenv, load_dotenv\n", "\n", "# \u8bfb\u53d6\u672c\u5730/\u9879\u76ee\u7684\u73af\u5883\u53d8\u91cf\u3002\n", "\n", "# find_dotenv()\u5bfb\u627e\u5e76\u5b9a\u4f4d.env\u6587\u4ef6\u7684\u8def\u5f84\n", "# load_dotenv()\u8bfb\u53d6\u8be5.env\u6587\u4ef6\uff0c\u5e76\u5c06\u5176\u4e2d\u7684\u73af\u5883\u53d8\u91cf\u52a0\u8f7d\u5230\u5f53\u524d\u7684\u8fd0\u884c\u73af\u5883\u4e2d\n", "# \u5982\u679c\u4f60\u8bbe\u7f6e\u7684\u662f\u5168\u5c40\u7684\u73af\u5883\u53d8\u91cf\uff0c\u8fd9\u884c\u4ee3\u7801\u5219\u6ca1\u6709\u4efb\u4f55\u4f5c\u7528\u3002\n", "_ = load_dotenv(find_dotenv())\n", "\n", "# \u83b7\u53d6\u73af\u5883\u53d8\u91cf OPENAI_API_KEY\n", "openai.api_key = os.environ[\"OPENAI_API_KEY\"]"]}, {"cell_type": "markdown", "id": "1297dcd5", "metadata": {"tags": []}, "source": ["## \u4e8c\u3001\u5bf9\u8bdd\u7f13\u5b58\u50a8\u5b58\n", " "]}, {"cell_type": "markdown", "id": "ffa8cd26-45c5-4bae-90b2-b07d23bf3bb2", "metadata": {}, "source": ["### 2.1 \u82f1\u6587\u7248"]}, {"cell_type": "markdown", "id": "b7e77a3d-7aaa-48c1-b219-19bd6f4eb674", "metadata": {"tags": []}, "source": ["#### 2.1.1 \u521d\u59cb\u5316\u5bf9\u8bdd\u6a21\u578b"]}, {"cell_type": "code", "execution_count": 47, "id": "20ad6fe2", "metadata": {"height": 98}, "outputs": [], "source": ["from langchain.chains import ConversationChain\n", "from langchain.chat_models import ChatOpenAI\n", "from langchain.memory import ConversationBufferMemory"]}, {"cell_type": "code", "execution_count": 48, "id": "88bdf13d", "metadata": {"height": 133}, "outputs": [], "source": ["# \u8fd9\u91cc\u6211\u4eec\u5c06\u53c2\u6570temperature\u8bbe\u7f6e\u4e3a0.0\uff0c\u4ece\u800c\u51cf\u5c11\u751f\u6210\u7b54\u6848\u7684\u968f\u673a\u6027\u3002\n", "# \u5982\u679c\u4f60\u60f3\u8981\u6bcf\u6b21\u5f97\u5230\u4e0d\u4e00\u6837\u7684\u6709\u65b0\u610f\u7684\u7b54\u6848\uff0c\u53ef\u4ee5\u5c1d\u8bd5\u589e\u5927\u8be5\u53c2\u6570\u3002\n", "llm = ChatOpenAI(temperature=0.0) \n", "\n", "memory = ConversationBufferMemory()\n", "\n", "# \u65b0\u5efa\u4e00\u4e2a ConversationChain Class \u5b9e\u4f8b\n", "# verbose\u53c2\u6570\u8bbe\u7f6e\u4e3aTrue\u65f6\uff0c\u7a0b\u5e8f\u4f1a\u8f93\u51fa\u66f4\u8be6\u7ec6\u7684\u4fe1\u606f\uff0c\u4ee5\u63d0\u4f9b\u66f4\u591a\u7684\u8c03\u8bd5\u6216\u8fd0\u884c\u65f6\u4fe1\u606f\u3002\n", "# \u76f8\u53cd\uff0c\u5f53\u5c06verbose\u53c2\u6570\u8bbe\u7f6e\u4e3aFalse\u65f6\uff0c\u7a0b\u5e8f\u4f1a\u4ee5\u66f4\u7b80\u6d01\u7684\u65b9\u5f0f\u8fd0\u884c\uff0c\u53ea\u8f93\u51fa\u5173\u952e\u7684\u4fe1\u606f\u3002\n", "conversation = ConversationChain(llm=llm, memory = memory, verbose=True )"]}, {"cell_type": "markdown", "id": "dea83837", "metadata": {"tags": []}, "source": ["#### 2.1.2 \u7b2c\u4e00\u8f6e\u5bf9\u8bdd"]}, {"cell_type": "markdown", "id": "1a3b4c42", "metadata": {}, "source": ["\u5f53\u6211\u4eec\u8fd0\u884c\u9884\u6d4b(predict)\u65f6\uff0c\u751f\u6210\u4e86\u4e00\u4e9b\u63d0\u793a\uff0c\u5982\u4e0b\u6240\u89c1\uff0c\u4ed6\u8bf4\u201c\u4ee5\u4e0b\u662f\u4eba\u7c7b\u548cAI\u4e4b\u95f4\u53cb\u597d\u7684\u5bf9\u8bdd\uff0cAI\u5065\u8c08\u201c\u7b49\u7b49\uff0c\u8fd9\u5b9e\u9645\u4e0a\u662fLangChain\u751f\u6210\u7684\u63d0\u793a\uff0c\u4ee5\u4f7f\u7cfb\u7edf\u8fdb\u884c\u5e0c\u671b\u548c\u53cb\u597d\u7684\u5bf9\u8bdd\uff0c\u5e76\u4e14\u5fc5\u987b\u4fdd\u5b58\u5bf9\u8bdd\uff0c\u5e76\u63d0\u793a\u4e86\u5f53\u524d\u5df2\u5b8c\u6210\u7684\u6a21\u578b\u94fe\u3002"]}, {"cell_type": "code", "execution_count": 49, "id": "db24677d", "metadata": {"height": 47}, "outputs": [{"name": "stdout", "output_type": "stream", "text": ["\n", "\n", "\u001b[1m> Entering new chain...\u001b[0m\n", "Prompt after formatting:\n", "\u001b[32;1m\u001b[1;3mThe following is a friendly conversation between a human and an AI. The AI is talkative and provides lots of specific details from its context. If the AI does not know the answer to a question, it truthfully says it does not know.\n", "\n", "Current conversation:\n", "\n", "Human: Hi, my name is Andrew\n", "AI:\u001b[0m\n", "\n", "\u001b[1m> Finished chain.\u001b[0m\n"]}, {"data": {"text/plain": ["\"Hello Andrew! It's nice to meet you. How can I assist you today?\""]}, "execution_count": 49, "metadata": {}, "output_type": "execute_result"}], "source": ["conversation.predict(input=\"Hi, my name is Andrew\")"]}, {"cell_type": "markdown", "id": "e71564ad", "metadata": {}, "source": ["#### 2.1.3 \u7b2c\u4e8c\u8f6e\u5bf9\u8bdd"]}, {"cell_type": "markdown", "id": "54d006bd", "metadata": {}, "source": ["\u5f53\u6211\u4eec\u8fdb\u884c\u7b2c\u4e8c\u8f6e\u5bf9\u8bdd\u65f6\uff0c\u5b83\u4f1a\u4fdd\u7559\u4e0a\u9762\u7684\u63d0\u793a"]}, {"cell_type": "code", "execution_count": 50, "id": "cc3ef937", "metadata": {"height": 31}, "outputs": [{"name": "stdout", "output_type": "stream", "text": ["\n", "\n", "\u001b[1m> Entering new chain...\u001b[0m\n", "Prompt after formatting:\n", "\u001b[32;1m\u001b[1;3mThe following is a friendly conversation between a human and an AI. The AI is talkative and provides lots of specific details from its context. If the AI does not know the answer to a question, it truthfully says it does not know.\n", "\n", "Current conversation:\n", "Human: Hi, my name is Andrew\n", "AI: Hello Andrew! It's nice to meet you. How can I assist you today?\n", "Human: What is 1+1?\n", "AI:\u001b[0m\n", "\n", "\u001b[1m> Finished chain.\u001b[0m\n"]}, {"data": {"text/plain": ["'1+1 is equal to 2.'"]}, "execution_count": 50, "metadata": {}, "output_type": "execute_result"}], "source": ["conversation.predict(input=\"What is 1+1?\")"]}, {"cell_type": "markdown", "id": "33cb734b", "metadata": {}, "source": ["#### 2.1.4 \u7b2c\u4e09\u8f6e\u5bf9\u8bdd"]}, {"cell_type": "markdown", "id": "0393df3d", "metadata": {}, "source": ["\u4e3a\u4e86\u9a8c\u8bc1\u4ed6\u662f\u5426\u8bb0\u5fc6\u4e86\u524d\u9762\u7684\u5bf9\u8bdd\u5185\u5bb9\uff0c\u6211\u4eec\u8ba9\u4ed6\u56de\u7b54\u524d\u9762\u5df2\u7ecf\u8bf4\u8fc7\u7684\u5185\u5bb9\uff08\u6211\u7684\u540d\u5b57\uff09\uff0c\u53ef\u4ee5\u770b\u5230\u4ed6\u786e\u5b9e\u8f93\u51fa\u4e86\u6b63\u786e\u7684\u540d\u5b57\uff0c\u56e0\u6b64\u8fd9\u4e2a\u5bf9\u8bdd\u94fe\u968f\u7740\u5f80\u4e0b\u8fdb\u884c\u4f1a\u8d8a\u6765\u8d8a\u957f"]}, {"cell_type": "code", "execution_count": 51, "id": "acf3339a", "metadata": {"height": 31}, "outputs": [{"name": "stdout", "output_type": "stream", "text": ["\n", "\n", "\u001b[1m> Entering new chain...\u001b[0m\n", "Prompt after formatting:\n", "\u001b[32;1m\u001b[1;3mThe following is a friendly conversation between a human and an AI. The AI is talkative and provides lots of specific details from its context. If the AI does not know the answer to a question, it truthfully says it does not know.\n", "\n", "Current conversation:\n", "Human: Hi, my name is Andrew\n", "AI: Hello Andrew! It's nice to meet you. How can I assist you today?\n", "Human: What is 1+1?\n", "AI: 1+1 is equal to 2.\n", "Human: What is my name?\n", "AI:\u001b[0m\n", "\n", "\u001b[1m> Finished chain.\u001b[0m\n"]}, {"data": {"text/plain": ["'Your name is Andrew.'"]}, "execution_count": 51, "metadata": {}, "output_type": "execute_result"}], "source": ["conversation.predict(input=\"What is my name?\")"]}, {"cell_type": "markdown", "id": "5a96a8d9", "metadata": {}, "source": ["#### 2.1.5 \u67e5\u770b\u50a8\u5b58\u7f13\u5b58\n", "\n", "\u50a8\u5b58\u7f13\u5b58(memory.buffer)\n", "\u50a8\u5b58\u4e86\u5f53\u524d\u4e3a\u6b62\u6240\u6709\u7684\u5bf9\u8bdd\u4fe1\u606f"]}, {"cell_type": "code", "execution_count": 52, "id": "2529400d", "metadata": {"height": 31}, "outputs": [{"name": "stdout", "output_type": "stream", "text": ["Human: Hi, my name is Andrew\n", "AI: Hello Andrew! It's nice to meet you. How can I assist you today?\n", "Human: What is 1+1?\n", "AI: 1+1 is equal to 2.\n", "Human: What is my name?\n", "AI: Your name is Andrew.\n"]}], "source": ["print(memory.buffer) "]}, {"cell_type": "markdown", "id": "0b5de846", "metadata": {}, "source": ["\u4e5f\u53ef\u4ee5\u901a\u8fc7memory.load_memory_variables({})\u6253\u5370\u7f13\u5b58\u4e2d\u7684\u5386\u53f2\u6d88\u606f\u3002\u8fd9\u91cc\u7684`{}`\u662f\u4e00\u4e2a\u7a7a\u5b57\u5178\uff0c\u6709\u4e00\u4e9b\u66f4\u9ad8\u7ea7\u7684\u529f\u80fd\uff0c\u4f7f\u7528\u6237\u53ef\u4ee5\u4f7f\u7528\u66f4\u590d\u6742\u7684\u8f93\u5165\uff0c\u4f46\u6211\u4eec\u4e0d\u4f1a\u5728\u8fd9\u4e2a\u77ed\u671f\u8bfe\u7a0b\u4e2d\u8ba8\u8bba\u5b83\u4eec\uff0c\u6240\u4ee5\u4e0d\u8981\u62c5\u5fc3\u4e3a\u4ec0\u4e48\u8fd9\u91cc\u6709\u4e00\u4e2a\u7a7a\u7684\u82b1\u62ec\u53f7\u3002"]}, {"cell_type": "code", "execution_count": 53, "id": "5018cb0a", "metadata": {"height": 31}, "outputs": [{"name": "stdout", "output_type": "stream", "text": ["{'history': \"Human: Hi, my name is Andrew\\nAI: Hello Andrew! It's nice to meet you. How can I assist you today?\\nHuman: What is 1+1?\\nAI: 1+1 is equal to 2.\\nHuman: What is my name?\\nAI: Your name is Andrew.\"}\n"]}], "source": ["print(memory.load_memory_variables({}))"]}, {"cell_type": "markdown", "id": "07d2e892", "metadata": {}, "source": ["#### 2.1.6 \u76f4\u63a5\u6dfb\u52a0\u5185\u5bb9\u5230\u50a8\u5b58\u7f13\u5b58"]}, {"cell_type": "code", "execution_count": 54, "id": "14219b70", "metadata": {"height": 31}, "outputs": [], "source": ["memory = ConversationBufferMemory() # \u65b0\u5efa\u4e00\u4e2a\u7a7a\u7684\u5bf9\u8bdd\u7f13\u5b58\u8bb0\u5fc6"]}, {"cell_type": "code", "execution_count": 55, "id": "a36e9905", "metadata": {"height": 48}, "outputs": [], "source": ["memory.save_context({\"input\": \"Hi\"}, {\"output\": \"What's up\"}) # \u5411\u7f13\u5b58\u533a\u6dfb\u52a0\u6307\u5b9a\u5bf9\u8bdd\u7684\u8f93\u5165\u8f93\u51fa"]}, {"cell_type": "code", "execution_count": 56, "id": "61631b1f", "metadata": {"height": 31}, "outputs": [{"name": "stdout", "output_type": "stream", "text": ["Human: Hi\n", "AI: What's up\n"]}], "source": ["print(memory.buffer) # \u67e5\u770b\u7f13\u5b58\u533a\u7ed3\u679c"]}, {"cell_type": "code", "execution_count": 57, "id": "a2fdf9ec", "metadata": {"height": 31}, "outputs": [{"name": "stdout", "output_type": "stream", "text": ["{'history': \"Human: Hi\\nAI: What's up\"}\n"]}], "source": ["print(memory.load_memory_variables({}))# \u518d\u6b21\u52a0\u8f7d\u8bb0\u5fc6\u53d8\u91cf"]}, {"cell_type": "markdown", "id": "2ac544f2", "metadata": {}, "source": ["\u7ee7\u7eed\u6dfb\u52a0\u65b0\u7684\u5185\u5bb9\uff0c\u5bf9\u8bdd\u5386\u53f2\u90fd\u4fdd\u5b58\u4e0b\u6765\u5728\u4e86\uff01"]}, {"cell_type": "code", "execution_count": 58, "id": "7ca79256", "metadata": {"height": 64}, "outputs": [], "source": ["memory.save_context({\"input\": \"Not much, just hanging\"}, {\"output\": \"Cool\"})"]}, {"cell_type": "code", "execution_count": 59, "id": "890a4497", "metadata": {"height": 31}, "outputs": [{"data": {"text/plain": ["{'history': \"Human: Hi\\nAI: What's up\\nHuman: Not much, just hanging\\nAI: Cool\"}"]}, "execution_count": 59, "metadata": {}, "output_type": "execute_result"}], "source": ["memory.load_memory_variables({})"]}, {"cell_type": "markdown", "id": "6eddd91a-bf1c-4b82-b99c-c585420e4ecb", "metadata": {}, "source": ["### 2.2 \u4e2d\u6587\u7248"]}, {"cell_type": "markdown", "id": "55b3e4e9-7a6a-4a09-9ac3-0096a67849c7", "metadata": {}, "source": ["#### 2.1.1 \u521d\u59cb\u5316\u5bf9\u8bdd\u6a21\u578b"]}, {"cell_type": "code", "execution_count": 30, "id": "3577aaff-7edb-40b0-866a-e407e63d55e0", "metadata": {"height": 98}, "outputs": [], "source": ["from langchain.chains import ConversationChain\n", "from langchain.chat_models import ChatOpenAI\n", "from langchain.memory import ConversationBufferMemory"]}, {"cell_type": "code", "execution_count": 33, "id": "a77d37ab-1f75-4ae8-8d7c-5066773ead81", "metadata": {"height": 133}, "outputs": [{"name": "stdout", "output_type": "stream", "text": ["\n", "\n", "\u001b[1m> Entering new chain...\u001b[0m\n", "Prompt after formatting:\n", "\u001b[32;1m\u001b[1;3mThe following is a friendly conversation between a human and an AI. The AI is talkative and provides lots of specific details from its context. If the AI does not know the answer to a question, it truthfully says it does not know.\n", "\n", "Current conversation:\n", "\n", "Human: \u4f60\u597d, \u6211\u53eb\u76ae\u76ae\u9c81\n", "AI:\u001b[0m\n", "\n", "\u001b[1m> Finished chain.\u001b[0m\n"]}, {"data": {"text/plain": ["'\u4f60\u597d\uff0c\u76ae\u76ae\u9c81\uff01\u5f88\u9ad8\u5174\u8ba4\u8bc6\u4f60\u3002\u6211\u662f\u4e00\u4e2aAI\u52a9\u624b\uff0c\u53ef\u4ee5\u56de\u7b54\u4f60\u7684\u95ee\u9898\u548c\u63d0\u4f9b\u5e2e\u52a9\u3002\u6709\u4ec0\u4e48\u6211\u53ef\u4ee5\u5e2e\u4f60\u7684\u5417\uff1f'"]}, "execution_count": 33, "metadata": {}, "output_type": "execute_result"}], "source": ["llm = ChatOpenAI(temperature=0.0) \n", "\n", "memory = ConversationBufferMemory()\n", "\n", "conversation = ConversationChain(llm=llm, memory = memory, verbose=True )"]}, {"cell_type": "markdown", "id": "747cb539-abc4-4e47-8cb9-1ee608ab07fc", "metadata": {}, "source": ["#### 2.1.2 \u7b2c\u4e00\u8f6e\u5bf9\u8bdd"]}, {"cell_type": "code", "execution_count": null, "id": "99e48462-7a92-4842-bdaa-2a478ba2252c", "metadata": {}, "outputs": [], "source": ["conversation.predict(input=\"\u4f60\u597d, \u6211\u53eb\u76ae\u76ae\u9c81\")"]}, {"cell_type": "markdown", "id": "979d320b-6b20-4722-99db-c48a43711d6c", "metadata": {}, "source": ["#### 2.1.3 \u7b2c\u4e8c\u8f6e\u5bf9\u8bdd"]}, {"cell_type": "code", "execution_count": 34, "id": "6ed6e97a-d7ea-4188-a6d7-f91d2a29d14a", "metadata": {}, "outputs": [{"name": "stdout", "output_type": "stream", "text": ["\n", "\n", "\u001b[1m> Entering new chain...\u001b[0m\n", "Prompt after formatting:\n", "\u001b[32;1m\u001b[1;3mThe following is a friendly conversation between a human and an AI. The AI is talkative and provides lots of specific details from its context. If the AI does not know the answer to a question, it truthfully says it does not know.\n", "\n", "Current conversation:\n", "Human: \u4f60\u597d, \u6211\u53eb\u76ae\u76ae\u9c81\n", "AI: \u4f60\u597d\uff0c\u76ae\u76ae\u9c81\uff01\u5f88\u9ad8\u5174\u8ba4\u8bc6\u4f60\u3002\u6211\u662f\u4e00\u4e2aAI\u52a9\u624b\uff0c\u53ef\u4ee5\u56de\u7b54\u4f60\u7684\u95ee\u9898\u548c\u63d0\u4f9b\u5e2e\u52a9\u3002\u6709\u4ec0\u4e48\u6211\u53ef\u4ee5\u5e2e\u4f60\u7684\u5417\uff1f\n", "Human: 1+1\u7b49\u4e8e\u591a\u5c11\uff1f\n", "AI:\u001b[0m\n", "\n", "\u001b[1m> Finished chain.\u001b[0m\n"]}, {"data": {"text/plain": ["'1+1\u7b49\u4e8e2\u3002'"]}, "execution_count": 34, "metadata": {}, "output_type": "execute_result"}], "source": ["conversation.predict(input=\"1+1\u7b49\u4e8e\u591a\u5c11\uff1f\")"]}, {"cell_type": "markdown", "id": "8a1fd531-216e-42d8-b226-839747ad7dd3", "metadata": {}, "source": ["#### 2.1.4 \u7b2c\u4e09\u8f6e\u5bf9\u8bdd"]}, {"cell_type": "code", "execution_count": null, "id": "b5dfe488-2758-42c7-9c20-e483b4c22ab8", "metadata": {}, "outputs": [{"name": "stdout", "output_type": "stream", "text": ["\n", "\n", "\u001b[1m> Entering new chain...\u001b[0m\n", "Prompt after formatting:\n", "\u001b[32;1m\u001b[1;3mThe following is a friendly conversation between a human and an AI. The AI is talkative and provides lots of specific details from its context. If the AI does not know the answer to a question, it truthfully says it does not know.\n", "\n", "Current conversation:\n", "Human: \u4f60\u597d, \u6211\u53eb\u76ae\u76ae\u9c81\n", "AI: \u4f60\u597d\uff0c\u76ae\u76ae\u9c81\uff01\u5f88\u9ad8\u5174\u8ba4\u8bc6\u4f60\u3002\u6211\u662f\u4e00\u4e2aAI\u52a9\u624b\uff0c\u53ef\u4ee5\u56de\u7b54\u4f60\u7684\u95ee\u9898\u548c\u63d0\u4f9b\u5e2e\u52a9\u3002\u6709\u4ec0\u4e48\u6211\u53ef\u4ee5\u5e2e\u4f60\u7684\u5417\uff1f\n", "Human: 1+1\u7b49\u4e8e\u591a\u5c11\uff1f\n", "AI: 1+1\u7b49\u4e8e2\u3002\n", "Human: What is my name?\n", "AI: \u4f60\u7684\u540d\u5b57\u662f\u76ae\u76ae\u9c81\u3002\n", "Human: \u6211\u53eb\u4ec0\u4e48\u540d\u5b57\uff1f\n", "AI:\u001b[0m\n", "\n", "\u001b[1m> Finished chain.\u001b[0m\n"]}, {"data": {"text/plain": ["'\u4f60\u53eb\u76ae\u76ae\u9c81\u3002'"]}, "execution_count": 36, "metadata": {}, "output_type": "execute_result"}], "source": ["conversation.predict(input=\"\u6211\u53eb\u4ec0\u4e48\u540d\u5b57\uff1f\")"]}, {"cell_type": "markdown", "id": "05d9822e-943d-4905-a1f8-a0d28c215d60", "metadata": {}, "source": ["#### 2.1.5 \u67e5\u770b\u50a8\u5b58\u7f13\u5b58"]}, {"cell_type": "code", "execution_count": 37, "id": "0795580f-b6b6-47e0-8882-26fe204560bd", "metadata": {"height": 31}, "outputs": [{"name": "stdout", "output_type": "stream", "text": ["Human: \u4f60\u597d, \u6211\u53eb\u76ae\u76ae\u9c81\n", "AI: \u4f60\u597d\uff0c\u76ae\u76ae\u9c81\uff01\u5f88\u9ad8\u5174\u8ba4\u8bc6\u4f60\u3002\u6211\u662f\u4e00\u4e2aAI\u52a9\u624b\uff0c\u53ef\u4ee5\u56de\u7b54\u4f60\u7684\u95ee\u9898\u548c\u63d0\u4f9b\u5e2e\u52a9\u3002\u6709\u4ec0\u4e48\u6211\u53ef\u4ee5\u5e2e\u4f60\u7684\u5417\uff1f\n", "Human: 1+1\u7b49\u4e8e\u591a\u5c11\uff1f\n", "AI: 1+1\u7b49\u4e8e2\u3002\n", "Human: What is my name?\n", "AI: \u4f60\u7684\u540d\u5b57\u662f\u76ae\u76ae\u9c81\u3002\n", "Human: \u6211\u53eb\u4ec0\u4e48\u540d\u5b57\uff1f\n", "AI: \u4f60\u53eb\u76ae\u76ae\u9c81\u3002\n"]}], "source": ["print(memory.buffer) "]}, {"cell_type": "code", "execution_count": 38, "id": "dfe7824f-bd6e-4b95-92e2-1c85c62a92e9", "metadata": {}, "outputs": [{"name": "stdout", "output_type": "stream", "text": ["Human: \u4f60\u597d, \u6211\u53eb\u76ae\u76ae\u9c81\n", "AI: \u4f60\u597d\uff0c\u76ae\u76ae\u9c81\uff01\u5f88\u9ad8\u5174\u8ba4\u8bc6\u4f60\u3002\u6211\u662f\u4e00\u4e2aAI\u52a9\u624b\uff0c\u53ef\u4ee5\u56de\u7b54\u4f60\u7684\u95ee\u9898\u548c\u63d0\u4f9b\u5e2e\u52a9\u3002\u6709\u4ec0\u4e48\u6211\u53ef\u4ee5\u5e2e\u4f60\u7684\u5417\uff1f\n", "Human: 1+1\u7b49\u4e8e\u591a\u5c11\uff1f\n", "AI: 1+1\u7b49\u4e8e2\u3002\n", "Human: What is my name?\n", "AI: \u4f60\u7684\u540d\u5b57\u662f\u76ae\u76ae\u9c81\u3002\n", "Human: \u6211\u53eb\u4ec0\u4e48\u540d\u5b57\uff1f\n", "AI: \u4f60\u53eb\u76ae\u76ae\u9c81\u3002\n"]}], "source": ["print(memory.buffer) "]}, {"cell_type": "code", "execution_count": 39, "id": "48942759-8afb-4aed-80c5-a48952a2b0c0", "metadata": {"height": 31}, "outputs": [{"name": "stdout", "output_type": "stream", "text": ["{'history': 'Human: \u4f60\u597d, \u6211\u53eb\u76ae\u76ae\u9c81\\nAI: \u4f60\u597d\uff0c\u76ae\u76ae\u9c81\uff01\u5f88\u9ad8\u5174\u8ba4\u8bc6\u4f60\u3002\u6211\u662f\u4e00\u4e2aAI\u52a9\u624b\uff0c\u53ef\u4ee5\u56de\u7b54\u4f60\u7684\u95ee\u9898\u548c\u63d0\u4f9b\u5e2e\u52a9\u3002\u6709\u4ec0\u4e48\u6211\u53ef\u4ee5\u5e2e\u4f60\u7684\u5417\uff1f\\nHuman: 1+1\u7b49\u4e8e\u591a\u5c11\uff1f\\nAI: 1+1\u7b49\u4e8e2\u3002\\nHuman: What is my name?\\nAI: \u4f60\u7684\u540d\u5b57\u662f\u76ae\u76ae\u9c81\u3002\\nHuman: \u6211\u53eb\u4ec0\u4e48\u540d\u5b57\uff1f\\nAI: \u4f60\u53eb\u76ae\u76ae\u9c81\u3002'}\n"]}], "source": ["print(memory.load_memory_variables({}))"]}, {"cell_type": "code", "execution_count": 40, "id": "4d0c4625-e928-45dc-b8da-4ab865ac5f7e", "metadata": {}, "outputs": [{"data": {"text/plain": ["{'history': 'Human: \u4f60\u597d, \u6211\u53eb\u76ae\u76ae\u9c81\\nAI: \u4f60\u597d\uff0c\u76ae\u76ae\u9c81\uff01\u5f88\u9ad8\u5174\u8ba4\u8bc6\u4f60\u3002\u6211\u662f\u4e00\u4e2aAI\u52a9\u624b\uff0c\u53ef\u4ee5\u56de\u7b54\u4f60\u7684\u95ee\u9898\u548c\u63d0\u4f9b\u5e2e\u52a9\u3002\u6709\u4ec0\u4e48\u6211\u53ef\u4ee5\u5e2e\u4f60\u7684\u5417\uff1f\\nHuman: 1+1\u7b49\u4e8e\u591a\u5c11\uff1f\\nAI: 1+1\u7b49\u4e8e2\u3002\\nHuman: What is my name?\\nAI: \u4f60\u7684\u540d\u5b57\u662f\u76ae\u76ae\u9c81\u3002\\nHuman: \u6211\u53eb\u4ec0\u4e48\u540d\u5b57\uff1f\\nAI: \u4f60\u53eb\u76ae\u76ae\u9c81\u3002'}"]}, "execution_count": 40, "metadata": {}, "output_type": "execute_result"}], "source": ["memory.load_memory_variables({})"]}, {"cell_type": "markdown", "id": "80f3778e-5fbf-43ed-9df1-d57d98ec6fb0", "metadata": {}, "source": ["#### 2.1.6 \u76f4\u63a5\u6dfb\u52a0\u5185\u5bb9\u5230\u50a8\u5b58\u7f13\u5b58"]}, {"cell_type": "code", "execution_count": 42, "id": "147b2c30-1662-4b49-aaf8-c228428e5cc6", "metadata": {}, "outputs": [{"data": {"text/plain": ["{'history': 'Human: \u4f60\u597d\uff0c\u6211\u53eb\u76ae\u76ae\u9c81\\nAI: \u4f60\u597d\u554a\uff0c\u6211\u53eb\u9c81\u897f\u897f'}"]}, "execution_count": 42, "metadata": {}, "output_type": "execute_result"}], "source": ["memory = ConversationBufferMemory()\n", "memory.save_context({\"input\": \"\u4f60\u597d\uff0c\u6211\u53eb\u76ae\u76ae\u9c81\"}, {\"output\": \"\u4f60\u597d\u554a\uff0c\u6211\u53eb\u9c81\u897f\u897f\"})\n", "memory.load_memory_variables({})"]}, {"cell_type": "code", "execution_count": 43, "id": "6b5e27f0-dad7-41b6-9326-bebf6299638f", "metadata": {"height": 64}, "outputs": [], "source": ["memory.save_context({\"input\": \"Not much, just hanging\"}, {\"output\": \"Cool\"})"]}, {"cell_type": "code", "execution_count": 44, "id": "cfa7c555-06ab-4906-b3dc-906f789e08f5", "metadata": {"height": 31}, "outputs": [{"data": {"text/plain": ["{'history': 'Human: \u4f60\u597d\uff0c\u6211\u53eb\u76ae\u76ae\u9c81\\nAI: \u4f60\u597d\u554a\uff0c\u6211\u53eb\u9c81\u897f\u897f\\nHuman: Not much, just hanging\\nAI: Cool'}"]}, "execution_count": 44, "metadata": {}, "output_type": "execute_result"}], "source": ["memory.load_memory_variables({})"]}, {"cell_type": "code", "execution_count": 45, "id": "deb33de8-37ea-4180-a73e-0fc456b14eb0", "metadata": {}, "outputs": [{"data": {"text/plain": ["{'history': 'Human: \u4f60\u597d\uff0c\u6211\u53eb\u76ae\u76ae\u9c81\\nAI: \u4f60\u597d\u554a\uff0c\u6211\u53eb\u9c81\u897f\u897f\\nHuman: Not much, just hanging\\nAI: Cool\\nHuman: \u5f88\u9ad8\u5174\u548c\u4f60\u6210\u4e3a\u670b\u53cb\uff01\\nAI: \u662f\u7684\uff0c\u8ba9\u6211\u4eec\u4e00\u8d77\u53bb\u5192\u9669\u5427\uff01'}"]}, "execution_count": 45, "metadata": {}, "output_type": "execute_result"}], "source": ["memory.save_context({\"input\": \"\u5f88\u9ad8\u5174\u548c\u4f60\u6210\u4e3a\u670b\u53cb\uff01\"}, {\"output\": \"\u662f\u7684\uff0c\u8ba9\u6211\u4eec\u4e00\u8d77\u53bb\u5192\u9669\u5427\uff01\"})\n", "memory.load_memory_variables({})"]}, {"cell_type": "markdown", "id": "10146f1a-0114-4902-8122-d19ae6f7c461", "metadata": {}, "source": ["### 2.3 \u603b\u7ed3"]}, {"cell_type": "markdown", "id": "2759b6bc-edb2-4cfe-b0f8-1bf6c4d796f9", "metadata": {}, "source": ["\u5f53\u6211\u4eec\u5728\u4f7f\u7528\u5927\u578b\u8bed\u8a00\u6a21\u578b\u8fdb\u884c\u804a\u5929\u5bf9\u8bdd\u65f6\uff0c**\u5927\u578b\u8bed\u8a00\u6a21\u578b\u672c\u8eab\u5b9e\u9645\u4e0a\u662f\u65e0\u72b6\u6001\u7684\u3002\u8bed\u8a00\u6a21\u578b\u672c\u8eab\u5e76\u4e0d\u8bb0\u5f97\u5230\u76ee\u524d\u4e3a\u6b62\u7684\u5386\u53f2\u5bf9\u8bdd**\u3002\u6bcf\u6b21\u8c03\u7528API\u7ed3\u70b9\u90fd\u662f\u72ec\u7acb\u7684\u3002\u50a8\u5b58(Memory)\u53ef\u4ee5\u50a8\u5b58\u5230\u76ee\u524d\u4e3a\u6b62\u7684\u6240\u6709\u672f\u8bed\u6216\u5bf9\u8bdd\uff0c\u5e76\u5c06\u5176\u8f93\u5165\u6216\u9644\u52a0\u4e0a\u4e0b\u6587\u5230LLM\u4e2d\u7528\u4e8e\u751f\u6210\u8f93\u51fa\u3002\u5982\u6b64\u770b\u8d77\u6765\u5c31\u597d\u50cf\u5b83\u5728\u8fdb\u884c\u4e0b\u4e00\u8f6e\u5bf9\u8bdd\u7684\u65f6\u5019\uff0c\u8bb0\u5f97\u4e4b\u524d\u8bf4\u8fc7\u4ec0\u4e48\u3002\n"]}, {"cell_type": "markdown", "id": "cf98e9ff", "metadata": {}, "source": ["## \u4e09\u3001\u5bf9\u8bdd\u7f13\u5b58\u7a97\u53e3\u50a8\u5b58\n", " \n", "\u968f\u7740\u5bf9\u8bdd\u53d8\u5f97\u8d8a\u6765\u8d8a\u957f\uff0c\u6240\u9700\u7684\u5185\u5b58\u91cf\u4e5f\u53d8\u5f97\u975e\u5e38\u957f\u3002\u5c06\u5927\u91cf\u7684tokens\u53d1\u9001\u5230LLM\u7684\u6210\u672c\uff0c\u4e5f\u4f1a\u53d8\u5f97\u66f4\u52a0\u6602\u8d35,\u8fd9\u4e5f\u5c31\u662f\u4e3a\u4ec0\u4e48API\u7684\u8c03\u7528\u8d39\u7528\uff0c\u901a\u5e38\u662f\u57fa\u4e8e\u5b83\u9700\u8981\u5904\u7406\u7684tokens\u6570\u91cf\u800c\u6536\u8d39\u7684\u3002\n", " \n", "\u9488\u5bf9\u4ee5\u4e0a\u95ee\u9898\uff0cLangChain\u4e5f\u63d0\u4f9b\u4e86\u51e0\u79cd\u65b9\u4fbf\u7684\u50a8\u5b58\u65b9\u5f0f\u6765\u4fdd\u5b58\u5386\u53f2\u5bf9\u8bdd\u3002\u5176\u4e2d\uff0c\u5bf9\u8bdd\u7f13\u5b58\u7a97\u53e3\u50a8\u5b58\u53ea\u4fdd\u7559\u4e00\u4e2a\u7a97\u53e3\u5927\u5c0f\u7684\u5bf9\u8bdd\u3002\u5b83\u53ea\u4f7f\u7528\u6700\u8fd1\u7684n\u6b21\u4ea4\u4e92\u3002\u8fd9\u53ef\u4ee5\u7528\u4e8e\u4fdd\u6301\u6700\u8fd1\u4ea4\u4e92\u7684\u6ed1\u52a8\u7a97\u53e3\uff0c\u4ee5\u4fbf\u7f13\u51b2\u533a\u4e0d\u4f1a\u8fc7\u5927"]}, {"cell_type": "markdown", "id": "b63c9061-9916-4524-b497-93a0aa2b7d06", "metadata": {}, "source": ["### 3.1 \u82f1\u6587\u7248"]}, {"cell_type": "markdown", "id": "641477a4", "metadata": {}, "source": ["#### 3.1.1 \u6dfb\u52a0\u4e24\u8f6e\u5bf9\u8bdd\u5230\u7a97\u53e3\u50a8\u5b58"]}, {"cell_type": "code", "execution_count": 65, "id": "3ea6233e", "metadata": {"height": 47}, "outputs": [], "source": ["from langchain.memory import ConversationBufferWindowMemory\n", "\n", "# k \u4e3a\u7a97\u53e3\u53c2\u6570\uff0ck=1\u8868\u660e\u53ea\u4fdd\u7559\u4e00\u4e2a\u5bf9\u8bdd\u8bb0\u5fc6\n", "memory = ConversationBufferWindowMemory(k=1) "]}, {"cell_type": "code", "execution_count": 66, "id": "dc4553fb", "metadata": {"height": 115}, "outputs": [], "source": ["# \u5411memory\u6dfb\u52a0\u4e24\u8f6e\u5bf9\u8bdd\n", "memory.save_context({\"input\": \"Hi\"}, {\"output\": \"What's up\"})\n", "memory.save_context({\"input\": \"Not much, just hanging\"}, {\"output\": \"Cool\"})"]}, {"cell_type": "code", "execution_count": 67, "id": "6a788403", "metadata": {"height": 31}, "outputs": [{"data": {"text/plain": ["{'history': 'Human: Not much, just hanging\\nAI: Cool'}"]}, "execution_count": 67, "metadata": {}, "output_type": "execute_result"}], "source": ["# \u5e76\u67e5\u770b\u8bb0\u5fc6\u53d8\u91cf\u5f53\u524d\u7684\u8bb0\u5f55\n", "memory.load_memory_variables({})"]}, {"cell_type": "markdown", "id": "63bda148", "metadata": {}, "source": ["#### 3.1.2 \u5728\u5bf9\u8bdd\u94fe\u4e2d\u5e94\u7528\u7a97\u53e3\u50a8\u5b58"]}, {"cell_type": "code", "execution_count": 68, "id": "4087bc87", "metadata": {"height": 133}, "outputs": [], "source": ["llm = ChatOpenAI(temperature=0.0)\n", "memory = ConversationBufferWindowMemory(k=1)\n", "conversation = ConversationChain(llm=llm, memory=memory, verbose=False )"]}, {"cell_type": "markdown", "id": "b6d661e3", "metadata": {}, "source": ["\u6ce8\u610f\u6b64\u5904\uff01\u7531\u4e8e\u8fd9\u91cc\u7528\u7684\u662f\u4e00\u4e2a\u7a97\u53e3\u7684\u8bb0\u5fc6\uff0c\u56e0\u6b64\u53ea\u80fd\u4fdd\u5b58\u4e00\u8f6e\u7684\u5386\u53f2\u6d88\u606f\uff0c\u56e0\u6b64AI\u5e76\u4e0d\u80fd\u77e5\u9053\u4f60\u7b2c\u4e00\u8f6e\u5bf9\u8bdd\u4e2d\u63d0\u5230\u7684\u540d\u5b57\uff0c\u4ed6\u6700\u591a\u53ea\u80fd\u8bb0\u4f4f\u4e0a\u4e00\u8f6e\uff08\u7b2c\u4e8c\u8f6e\uff09\u7684\u5bf9\u8bdd\u4fe1\u606f"]}, {"cell_type": "code", "execution_count": 69, "id": "4faaa952", "metadata": {"height": 47}, "outputs": [{"data": {"text/plain": ["\"Hello Andrew! It's nice to meet you. How can I assist you today?\""]}, "execution_count": 69, "metadata": {}, "output_type": "execute_result"}], "source": ["conversation.predict(input=\"Hi, my name is Andrew\")"]}, {"cell_type": "code", "execution_count": 70, "id": "bb20ddaa", "metadata": {"height": 31}, "outputs": [{"data": {"text/plain": ["'1+1 is equal to 2.'"]}, "execution_count": 70, "metadata": {}, "output_type": "execute_result"}], "source": ["conversation.predict(input=\"What is 1+1?\")"]}, {"cell_type": "code", "execution_count": 71, "id": "489b2194", "metadata": {"height": 31}, "outputs": [{"data": {"text/plain": ["\"I'm sorry, but I don't have access to personal information.\""]}, "execution_count": 71, "metadata": {}, "output_type": "execute_result"}], "source": ["conversation.predict(input=\"What is my name?\")"]}, {"cell_type": "markdown", "id": "88837e7c-cf4b-469e-b820-bbfc49ba876c", "metadata": {}, "source": ["### 3.2 \u4e2d\u6587\u7248"]}, {"cell_type": "markdown", "id": "760ec3ad-6959-4a36-b1b5-4fcafe8088ad", "metadata": {}, "source": ["#### 3.1.1 \u6dfb\u52a0\u4e24\u8f6e\u5bf9\u8bdd\u5230\u7a97\u53e3\u50a8\u5b58"]}, {"cell_type": "code", "execution_count": 72, "id": "68a2907c", "metadata": {}, "outputs": [{"data": {"text/plain": ["{'history': 'Human: \u5f88\u9ad8\u5174\u548c\u4f60\u6210\u4e3a\u670b\u53cb\uff01\\nAI: \u662f\u7684\uff0c\u8ba9\u6211\u4eec\u4e00\u8d77\u53bb\u5192\u9669\u5427\uff01'}"]}, "execution_count": 72, "metadata": {}, "output_type": "execute_result"}], "source": ["from langchain.memory import ConversationBufferWindowMemory\n", "\n", "# k=1\u8868\u660e\u53ea\u4fdd\u7559\u4e00\u4e2a\u5bf9\u8bdd\u8bb0\u5fc6\n", "memory = ConversationBufferWindowMemory(k=1) \n", "memory.save_context({\"input\": \"\u4f60\u597d\uff0c\u6211\u53eb\u76ae\u76ae\u9c81\"}, {\"output\": \"\u4f60\u597d\u554a\uff0c\u6211\u53eb\u9c81\u897f\u897f\"})\n", "memory.save_context({\"input\": \"\u5f88\u9ad8\u5174\u548c\u4f60\u6210\u4e3a\u670b\u53cb\uff01\"}, {\"output\": \"\u662f\u7684\uff0c\u8ba9\u6211\u4eec\u4e00\u8d77\u53bb\u5192\u9669\u5427\uff01\"})\n", "memory.load_memory_variables({})"]}, {"cell_type": "markdown", "id": "dcabf017-5bfd-4904-9f06-388f994eddc9", "metadata": {}, "source": ["#### 3.1.2 \u5728\u5bf9\u8bdd\u94fe\u4e2d\u5e94\u7528\u7a97\u53e3\u50a8\u5b58"]}, {"cell_type": "code", "execution_count": 74, "id": "1ee854d9", "metadata": {}, "outputs": [{"name": "stdout", "output_type": "stream", "text": ["\u4f60\u597d\uff0c\u76ae\u76ae\u9c81\uff01\u5f88\u9ad8\u5174\u8ba4\u8bc6\u4f60\u3002\u6211\u662f\u4e00\u4e2aAI\u52a9\u624b\uff0c\u53ef\u4ee5\u56de\u7b54\u4f60\u7684\u95ee\u9898\u548c\u63d0\u4f9b\u5e2e\u52a9\u3002\u6709\u4ec0\u4e48\u6211\u53ef\u4ee5\u5e2e\u4f60\u7684\u5417\uff1f\n", "1+1\u7b49\u4e8e2\u3002\n", "\u5f88\u62b1\u6b49\uff0c\u6211\u65e0\u6cd5\u77e5\u9053\u60a8\u7684\u540d\u5b57\u3002\n"]}], "source": ["llm = ChatOpenAI(temperature=0.0)\n", "memory = ConversationBufferWindowMemory(k=1)\n", "conversation = ConversationChain(llm=llm, memory=memory, verbose=False )\n", "print(conversation.predict(input=\"\u4f60\u597d, \u6211\u53eb\u76ae\u76ae\u9c81\"))\n", "print(conversation.predict(input=\"1+1\u7b49\u4e8e\u591a\u5c11\uff1f\"))\n", "print(conversation.predict(input=\"\u6211\u53eb\u4ec0\u4e48\u540d\u5b57\uff1f\"))"]}, {"cell_type": "markdown", "id": "d2931b92", "metadata": {}, "source": ["## \u56db\u3001\u5bf9\u8bddtoken\u7f13\u5b58\u50a8\u5b58"]}, {"cell_type": "markdown", "id": "dff5b4c7", "metadata": {}, "source": ["\u4f7f\u7528\u5bf9\u8bddtoken\u7f13\u5b58\u8bb0\u5fc6\uff0c\u5185\u5b58\u5c06\u9650\u5236\u4fdd\u5b58\u7684token\u6570\u91cf\u3002\u5982\u679ctoken\u6570\u91cf\u8d85\u51fa\u6307\u5b9a\u6570\u76ee\uff0c\u5b83\u4f1a\u5207\u6389\u8fd9\u4e2a\u5bf9\u8bdd\u7684\u65e9\u671f\u90e8\u5206\n", "\u4ee5\u4fdd\u7559\u4e0e\u6700\u8fd1\u7684\u4ea4\u6d41\u76f8\u5bf9\u5e94\u7684token\u6570\u91cf\uff0c\u4f46\u4e0d\u8d85\u8fc7token\u9650\u5236\u3002"]}, {"cell_type": "code", "execution_count": 76, "id": "9f6d063c", "metadata": {"height": 31}, "outputs": [], "source": ["!pip install -q tiktoken "]}, {"cell_type": "code", "execution_count": 77, "id": "fb9020ed", "metadata": {"height": 81}, "outputs": [], "source": ["from langchain.llms import OpenAI\n", "from langchain.memory import ConversationTokenBufferMemory"]}, {"cell_type": "markdown", "id": "f3a84112", "metadata": {}, "source": ["### 4.1 \u82f1\u6587\u7248\n", "\u6dfb\u52a0\u5bf9\u8bdd\u5230Token\u7f13\u5b58\u50a8\u5b58,\u9650\u5236token\u6570\u91cf\uff0c\u8fdb\u884c\u6d4b\u8bd5"]}, {"cell_type": "code", "execution_count": 78, "id": "43582ee6", "metadata": {"height": 149}, "outputs": [], "source": ["memory = ConversationTokenBufferMemory(llm=llm, max_token_limit=30)\n", "memory.save_context({\"input\": \"AI is what?!\"}, {\"output\": \"Amazing!\"})\n", "memory.save_context({\"input\": \"Backpropagation is what?\"}, {\"output\": \"Beautiful!\"})\n", "memory.save_context({\"input\": \"Chatbots are what?\"}, {\"output\": \"Charming!\"})"]}, {"cell_type": "code", "execution_count": 79, "id": "284288e1", "metadata": {"height": 31}, "outputs": [{"data": {"text/plain": ["{'history': 'AI: Beautiful!\\nHuman: Chatbots are what?\\nAI: Charming!'}"]}, "execution_count": 79, "metadata": {}, "output_type": "execute_result"}], "source": ["memory.load_memory_variables({})"]}, {"cell_type": "markdown", "id": "7b62b2e1", "metadata": {}, "source": ["\u53ef\u4ee5\u770b\u5230\u524d\u9762\u8d85\u51fa\u7684\u7684token\u5df2\u7ecf\u88ab\u820d\u5f03\u4e86\uff01\uff01\uff01"]}, {"cell_type": "markdown", "id": "f7f6be43", "metadata": {}, "source": ["### 4.2 \u4e2d\u6587\u7248"]}, {"cell_type": "code", "execution_count": 80, "id": "e9191020", "metadata": {}, "outputs": [{"data": {"text/plain": ["{'history': 'AI: \u8f7b\u821f\u5df2\u8fc7\u4e07\u91cd\u5c71\u3002'}"]}, "execution_count": 80, "metadata": {}, "output_type": "execute_result"}], "source": ["memory = ConversationTokenBufferMemory(llm=llm, max_token_limit=30)\n", "memory.save_context({\"input\": \"\u671d\u8f9e\u767d\u5e1d\u5f69\u4e91\u95f4\uff0c\"}, {\"output\": \"\u5343\u91cc\u6c5f\u9675\u4e00\u65e5\u8fd8\u3002\"})\n", "memory.save_context({\"input\": \"\u4e24\u5cb8\u733f\u58f0\u557c\u4e0d\u4f4f\uff0c\"}, {\"output\": \"\u8f7b\u821f\u5df2\u8fc7\u4e07\u91cd\u5c71\u3002\"})\n", "memory.load_memory_variables({})"]}, {"cell_type": "markdown", "id": "fb08ef4a-876f-422a-81f9-4805288e5955", "metadata": {}, "source": ["### 4.3 \u8865\u5145"]}, {"cell_type": "markdown", "id": "5e4d918b", "metadata": {}, "source": ["ChatGPT\u4f7f\u7528\u4e00\u79cd\u57fa\u4e8e\u5b57\u8282\u5bf9\u7f16\u7801\uff08Byte Pair Encoding\uff0cBPE\uff09\u7684\u65b9\u6cd5\u6765\u8fdb\u884ctokenization\uff08\u5c06\u8f93\u5165\u6587\u672c\u62c6\u5206\u4e3atoken\uff09\u3002BPE\u662f\u4e00\u79cd\u5e38\u89c1\u7684tokenization\u6280\u672f\uff0c\u5b83\u5c06\u8f93\u5165\u6587\u672c\u5206\u5272\u6210\u8f83\u5c0f\u7684\u5b50\u8bcd\u5355\u5143\u3002 \n", "\n", "OpenAI\u5728\u5176\u5b98\u65b9GitHub\u4e0a\u516c\u5f00\u4e86\u4e00\u4e2a\u6700\u65b0\u7684\u5f00\u6e90Python\u5e93 [tiktoken](https://github.com/openai/tiktoken)\uff0c\u8fd9\u4e2a\u5e93\u4e3b\u8981\u662f\u7528\u6765\u8ba1\u7b97tokens\u6570\u91cf\u7684\u3002\u76f8\u6bd4\u8f83HuggingFace\u7684tokenizer\uff0c\u5176\u901f\u5ea6\u63d0\u5347\u4e86\u597d\u51e0\u500d\u3002\n", "\n", "\u5177\u4f53token\u8ba1\u7b97\u65b9\u5f0f,\u7279\u522b\u662f\u6c49\u5b57\u548c\u82f1\u6587\u5355\u8bcd\u7684token\u533a\u522b\uff0c\u5177\u4f53\u8bfe\u53c2\u8003[\u77e5\u4e4e\u6587\u7ae0](https://www.zhihu.com/question/594159910) \u3002"]}, {"cell_type": "markdown", "id": "5ff55d5d", "metadata": {}, "source": ["## \u4e94\u3001\u5bf9\u8bdd\u6458\u8981\u7f13\u5b58\u50a8\u5b58"]}, {"cell_type": "markdown", "id": "7d39b83a", "metadata": {}, "source": ["\u5bf9\u8bdd\u6458\u8981\u7f13\u5b58\u50a8\u5b58\uff0c**\u4f7f\u7528LLM\u7f16\u5199\u5230\u76ee\u524d\u4e3a\u6b62\u5386\u53f2\u5bf9\u8bdd\u7684\u6458\u8981**\uff0c\u5e76\u5c06\u5176\u4fdd\u5b58"]}, {"cell_type": "code", "execution_count": 82, "id": "72dcf8b1", "metadata": {"height": 64}, "outputs": [], "source": ["from langchain.chains import ConversationChain\n", "from langchain.chat_models import ChatOpenAI\n", "from langchain.memory import ConversationSummaryBufferMemory"]}, {"cell_type": "markdown", "id": "243b213e-ce17-46a0-8652-03658ca58dd8", "metadata": {}, "source": ["### 5.1 \u82f1\u6587\u7248"]}, {"cell_type": "markdown", "id": "6572ef39", "metadata": {}, "source": ["#### 5.1.1 \u4f7f\u7528\u5bf9\u8bdd\u6458\u8981\u7f13\u5b58\u50a8\u5b58\n", "\n", "\u521b\u5efa\u4e00\u4e2a\u957f\u5b57\u7b26\u4e32\uff0c\u5176\u4e2d\u5305\u542b\u67d0\u4eba\u7684\u65e5\u7a0b\u5b89\u6392"]}, {"cell_type": "code", "execution_count": 84, "id": "4a5b238f", "metadata": {"height": 285}, "outputs": [], "source": ["# \u521b\u5efa\u4e00\u4e2a\u957f\u5b57\u7b26\u4e32\n", "schedule = \"There is a meeting at 8am with your product team. \\\n", "You will need your powerpoint presentation prepared. \\\n", "9am-12pm have time to work on your LangChain \\\n", "project which will go quickly because Langchain is such a powerful tool. \\\n", "At Noon, lunch at the italian resturant with a customer who is driving \\\n", "from over an hour away to meet you to understand the latest in AI. \\\n", "Be sure to bring your laptop to show the latest LLM demo.\"\n", "\n", "# \u4f7f\u7528\u5bf9\u8bdd\u6458\u8981\u7f13\u5b58\u8bb0\u5fc6\n", "llm = ChatOpenAI(temperature=0.0)\n", "memory = ConversationSummaryBufferMemory(llm=llm, max_token_limit=100) \n", "memory.save_context({\"input\": \"Hello\"}, {\"output\": \"What's up\"})\n", "memory.save_context({\"input\": \"Not much, just hanging\"}, {\"output\": \"Cool\"})\n", "memory.save_context(\n", " {\"input\": \"What is on the schedule today?\"}, {\"output\": f\"{schedule}\"}\n", ")"]}, {"cell_type": "code", "execution_count": 89, "id": "15226a41-ab36-43a0-93f7-c03c6b374936", "metadata": {}, "outputs": [{"name": "stdout", "output_type": "stream", "text": ["System: The human and AI exchange greetings. The human asks about the schedule for the day. The AI provides a detailed schedule, including a meeting with the product team, work on the LangChain project, and a lunch meeting with a customer interested in AI. The AI emphasizes the importance of bringing a laptop to showcase the latest LLM demo during the lunch meeting.\n"]}], "source": ["print(memory.load_memory_variables({})['history'])"]}, {"cell_type": "markdown", "id": "7ccb97b6", "metadata": {}, "source": ["#### 5.1.2 \u57fa\u4e8e\u5bf9\u8bdd\u6458\u8981\u7f13\u5b58\u50a8\u5b58\u7684\u5bf9\u8bdd\u94fe\n", "\u57fa\u4e8e\u4e0a\u9762\u7684\u5bf9\u8bdd\u6458\u8981\u7f13\u5b58\u50a8\u5b58\uff0c\u65b0\u5efa\u4e00\u4e2a\u5bf9\u8bdd\u94fe"]}, {"cell_type": "code", "execution_count": 90, "id": "6728edba", "metadata": {"height": 99}, "outputs": [], "source": ["conversation = ConversationChain(llm=llm, memory=memory, verbose=True)"]}, {"cell_type": "code", "execution_count": 91, "id": "9a221b1d", "metadata": {"height": 47}, "outputs": [{"name": "stdout", "output_type": "stream", "text": ["\n", "\n", "\u001b[1m> Entering new chain...\u001b[0m\n", "Prompt after formatting:\n", "\u001b[32;1m\u001b[1;3mThe following is a friendly conversation between a human and an AI. The AI is talkative and provides lots of specific details from its context. If the AI does not know the answer to a question, it truthfully says it does not know.\n", "\n", "Current conversation:\n", "System: The human and AI exchange greetings. The human asks about the schedule for the day. The AI provides a detailed schedule, including a meeting with the product team, work on the LangChain project, and a lunch meeting with a customer interested in AI. The AI emphasizes the importance of bringing a laptop to showcase the latest LLM demo during the lunch meeting.\n", "Human: What would be a good demo to show?\n", "AI:\u001b[0m\n", "\n", "\u001b[1m> Finished chain.\u001b[0m\n"]}, {"data": {"text/plain": ["'A good demo to show during the lunch meeting with the customer interested in AI would be the latest LLM (Language Model) demo. The LLM is a cutting-edge AI model that can generate human-like text based on a given prompt. It has been trained on a vast amount of data and can generate coherent and contextually relevant responses. By showcasing the LLM demo, you can demonstrate the capabilities of AI in natural language processing and how it can be applied to various industries and use cases.'"]}, "execution_count": 91, "metadata": {}, "output_type": "execute_result"}], "source": ["conversation.predict(input=\"What would be a good demo to show?\")"]}, {"cell_type": "code", "execution_count": 92, "id": "bb582617", "metadata": {"height": 31}, "outputs": [{"name": "stdout", "output_type": "stream", "text": ["System: The human and AI exchange greetings. The human asks about the schedule for the day. The AI provides a detailed schedule, including a meeting with the product team, work on the LangChain project, and a lunch meeting with a customer interested in AI. The AI emphasizes the importance of bringing a laptop to showcase the latest LLM demo during the lunch meeting. A good demo to show during the lunch meeting with the customer interested in AI would be the latest LLM (Language Model) demo. The LLM is a cutting-edge AI model that can generate human-like text based on a given prompt. It has been trained on a vast amount of data and can generate coherent and contextually relevant responses. By showcasing the LLM demo, you can demonstrate the capabilities of AI in natural language processing and how it can be applied to various industries and use cases.\n"]}], "source": ["print(memory.load_memory_variables({})['history'])"]}, {"cell_type": "markdown", "id": "4ba827aa", "metadata": {"height": 31}, "source": ["### 5.2 \u4e2d\u6587\u7248"]}, {"cell_type": "markdown", "id": "64898f33-c538-4e68-b008-7123870b692b", "metadata": {}, "source": ["#### 5.2.1 \u4f7f\u7528\u5bf9\u8bdd\u6458\u8981\u7f13\u5b58\u50a8\u5b58\n", "\n", "\u521b\u5efa\u4e00\u4e2a\u957f\u5b57\u7b26\u4e32\uff0c\u5176\u4e2d\u5305\u542b\u67d0\u4eba\u7684\u65e5\u7a0b\u5b89\u6392"]}, {"cell_type": "code", "execution_count": 97, "id": "2c07922b", "metadata": {"height": 31}, "outputs": [], "source": ["# \u521b\u5efa\u4e00\u4e2a\u957f\u5b57\u7b26\u4e32\n", "schedule = \"\u5728\u516b\u70b9\u4f60\u548c\u4f60\u7684\u4ea7\u54c1\u56e2\u961f\u6709\u4e00\u4e2a\u4f1a\u8bae\u3002 \\\n", "\u4f60\u9700\u8981\u505a\u4e00\u4e2aPPT\u3002 \\\n", "\u4e0a\u53489\u70b9\u523012\u70b9\u4f60\u9700\u8981\u5fd9\u4e8eLangChain\u3002\\\n", "Langchain\u662f\u4e00\u4e2a\u6709\u7528\u7684\u5de5\u5177\uff0c\u56e0\u6b64\u4f60\u7684\u9879\u76ee\u8fdb\u5c55\u7684\u975e\u5e38\u5feb\u3002\\\n", "\u4e2d\u5348\uff0c\u5728\u610f\u5927\u5229\u9910\u5385\u4e0e\u4e00\u4f4d\u5f00\u8f66\u6765\u7684\u987e\u5ba2\u5171\u8fdb\u5348\u9910 \\\n", "\u8d70\u4e86\u4e00\u4e2a\u591a\u5c0f\u65f6\u7684\u8def\u7a0b\u4e0e\u4f60\u89c1\u9762\uff0c\u53ea\u4e3a\u4e86\u89e3\u6700\u65b0\u7684 AI\u3002 \\\n", "\u786e\u4fdd\u4f60\u5e26\u4e86\u7b14\u8bb0\u672c\u7535\u8111\u53ef\u4ee5\u5c55\u793a\u6700\u65b0\u7684 LLM \u6837\u4f8b.\"\n", "\n", "llm = ChatOpenAI(temperature=0.0)\n", "memory = ConversationSummaryBufferMemory(llm=llm, max_token_limit=100)\n", "memory.save_context({\"input\": \"\u4f60\u597d\uff0c\u6211\u53eb\u76ae\u76ae\u9c81\"}, {\"output\": \"\u4f60\u597d\u554a\uff0c\u6211\u53eb\u9c81\u897f\u897f\"})\n", "memory.save_context({\"input\": \"\u5f88\u9ad8\u5174\u548c\u4f60\u6210\u4e3a\u670b\u53cb\uff01\"}, {\"output\": \"\u662f\u7684\uff0c\u8ba9\u6211\u4eec\u4e00\u8d77\u53bb\u5192\u9669\u5427\uff01\"})\n", "memory.save_context({\"input\": \"\u4eca\u5929\u7684\u65e5\u7a0b\u5b89\u6392\u662f\u4ec0\u4e48\uff1f\"}, {\"output\": f\"{schedule}\"})"]}, {"cell_type": "code", "execution_count": 98, "id": "17424a12-430f-4529-9067-300978c6169e", "metadata": {}, "outputs": [{"name": "stdout", "output_type": "stream", "text": ["System: The human introduces themselves as Pipilu and the AI introduces themselves as Luxixi. They express happiness at becoming friends and decide to go on an adventure together. The human asks about the schedule for the day. The AI informs them that they have a meeting with their product team at 8 o'clock and need to prepare a PowerPoint presentation. From 9 am to 12 pm, they will be busy with LangChain, a useful tool that helps their project progress quickly. At noon, they will have lunch with a customer who has driven for over an hour just to learn about the latest AI. The AI advises the human to bring their laptop to showcase the latest LLM samples.\n"]}], "source": ["print(memory.load_memory_variables({})['history'])"]}, {"cell_type": "markdown", "id": "9e29a956-607a-4247-9eb5-01285a370991", "metadata": {}, "source": ["#### 5.1.2 \u57fa\u4e8e\u5bf9\u8bdd\u6458\u8981\u7f13\u5b58\u50a8\u5b58\u7684\u5bf9\u8bdd\u94fe\n", "\u57fa\u4e8e\u4e0a\u9762\u7684\u5bf9\u8bdd\u6458\u8981\u7f13\u5b58\u50a8\u5b58\uff0c\u65b0\u5efa\u4e00\u4e2a\u5bf9\u8bdd\u94fe"]}, {"cell_type": "code", "execution_count": 99, "id": "52696c8c", "metadata": {"height": 31}, "outputs": [], "source": ["conversation = ConversationChain(llm=llm, memory=memory, verbose=True)"]}, {"cell_type": "code", "execution_count": 100, "id": "48690d13", "metadata": {"height": 31}, "outputs": [{"name": "stdout", "output_type": "stream", "text": ["\n", "\n", "\u001b[1m> Entering new chain...\u001b[0m\n", "Prompt after formatting:\n", "\u001b[32;1m\u001b[1;3mThe following is a friendly conversation between a human and an AI. The AI is talkative and provides lots of specific details from its context. If the AI does not know the answer to a question, it truthfully says it does not know.\n", "\n", "Current conversation:\n", "System: The human introduces themselves as Pipilu and the AI introduces themselves as Luxixi. They express happiness at becoming friends and decide to go on an adventure together. The human asks about the schedule for the day. The AI informs them that they have a meeting with their product team at 8 o'clock and need to prepare a PowerPoint presentation. From 9 am to 12 pm, they will be busy with LangChain, a useful tool that helps their project progress quickly. At noon, they will have lunch with a customer who has driven for over an hour just to learn about the latest AI. The AI advises the human to bring their laptop to showcase the latest LLM samples.\n", "Human: \u5c55\u793a\u4ec0\u4e48\u6837\u7684\u6837\u4f8b\u6700\u597d\u5462\uff1f\n", "AI:\u001b[0m\n", "\n", "\u001b[1m> Finished chain.\u001b[0m\n"]}, {"data": {"text/plain": ["'\u5c55\u793a\u4e00\u4e9b\u5177\u6709\u591a\u6837\u6027\u548c\u521b\u65b0\u6027\u7684\u6837\u4f8b\u53ef\u80fd\u662f\u6700\u597d\u7684\u9009\u62e9\u3002\u4f60\u53ef\u4ee5\u9009\u62e9\u5c55\u793a\u4e00\u4e9b\u57fa\u4e8e\u56fe\u50cf\u8bc6\u522b\u7684\u6837\u4f8b\uff0c\u6bd4\u5982\u4eba\u8138\u8bc6\u522b\u3001\u7269\u4f53\u8bc6\u522b\u7b49\u3002\u53e6\u5916\uff0c\u4f60\u4e5f\u53ef\u4ee5\u5c55\u793a\u4e00\u4e9b\u81ea\u7136\u8bed\u8a00\u5904\u7406\u65b9\u9762\u7684\u6837\u4f8b\uff0c\u6bd4\u5982\u6587\u672c\u751f\u6210\u3001\u60c5\u611f\u5206\u6790\u7b49\u3002\u6700\u91cd\u8981\u7684\u662f\u9009\u62e9\u90a3\u4e9b\u80fd\u591f\u5c55\u793a\u51fa\u4f60\u4eec\u56e2\u961f\u7684\u6280\u672f\u5b9e\u529b\u548c\u521b\u9020\u529b\u7684\u6837\u4f8b\u3002'"]}, "execution_count": 100, "metadata": {}, "output_type": "execute_result"}], "source": ["conversation.predict(input=\"\u5c55\u793a\u4ec0\u4e48\u6837\u7684\u6837\u4f8b\u6700\u597d\u5462\uff1f\")"]}, {"cell_type": "code", "execution_count": 101, "id": "85bba1f8", "metadata": {"height": 31}, "outputs": [{"data": {"text/plain": ["{'history': \"System: The human introduces themselves as Pipilu and the AI introduces themselves as Luxixi. They express happiness at becoming friends and decide to go on an adventure together. The human asks about the schedule for the day. The AI informs them that they have a meeting with their product team at 8 o'clock and need to prepare a PowerPoint presentation. From 9 am to 12 pm, they will be busy with LangChain, a useful tool that helps their project progress quickly. At noon, they will have lunch with a customer who has driven for over an hour just to learn about the latest AI. The AI advises the human to bring their laptop to showcase the latest LLM samples. The human asks what kind of samples would be best to showcase. The AI suggests that showcasing diverse and innovative samples would be the best choice. They recommend selecting samples based on image recognition, such as face recognition and object recognition. Additionally, they suggest showcasing samples related to natural language processing, such as text generation and sentiment analysis. The AI emphasizes the importance of choosing samples that demonstrate the team's technical expertise and creativity.\"}"]}, "execution_count": 101, "metadata": {}, "output_type": "execute_result"}], "source": ["memory.load_memory_variables({}) # \u6458\u8981\u8bb0\u5f55\u66f4\u65b0\u4e86"]}], "metadata": {"kernelspec": {"display_name": "Python 3 (ipykernel)", "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": "3.9.12"}, "toc": {"base_numbering": 1, "nav_menu": {}, "number_sections": false, "sideBar": true, "skip_h1_title": false, "title_cell": "", "title_sidebar": "Contents", "toc_cell": false, "toc_position": {}, "toc_section_display": true, "toc_window_display": true}}, "nbformat": 4, "nbformat_minor": 5} \ No newline at end of file