{"id":1245,"date":"2024-04-28T19:32:35","date_gmt":"2024-04-28T19:32:35","guid":{"rendered":"https:\/\/www.nicekj.com\/?p=1245"},"modified":"2024-04-28T19:32:48","modified_gmt":"2024-04-28T19:32:48","slug":"ruheliyongopenaidehanshudiaoyongtexing","status":"publish","type":"post","link":"https:\/\/www.nicekj.com\/ruheliyongopenaidehanshudiaoyongtexing.html","title":{"rendered":"\u5982\u4f55\u5229\u7528OpenAI\u7684\u51fd\u6570\u8c03\u7528\u7279\u6027"},"content":{"rendered":"

\u5982\u4f55\u5229\u7528OpenAI\u7684\u51fd\u6570\u8c03\u7528\u7279\u6027<\/h1>\n

\u51fd\u6570\u8c03\u7528\u80fd\u5b9e\u73b0\u54ea\u4e9b\u529f\u80fd\uff1f<\/h2>\n

\u7b80\u5355\u6765\u8bf4\uff0c\u51fd\u6570\u8c03\u7528\u529f\u80fd\u53ef\u4ee5\u52a9\u4f60\u5728\u8bf7\u6c42\u65b9\u6cd5\u65f6\u6784\u5efa\u7ed3\u6784\u5316\u7684\u6570\u636e<\/strong>\u3002\u56e0\u4e3a\u751f\u6210\u6a21\u578b\u7684\u7279\u6027\uff0c\u5b83\u4ea7\u751f\u7684\u6570\u636e\u5f80\u5f80\u662f\u65e0\u7ed3\u6784\u7684\uff0c\u5373\u4f7f\u5728\u63d0\u793a(prompt)\u4e2d\u6307\u5b9a\u4e86\u8f93\u51fa\u683c\u5f0f\uff0c\u4f46\u5b9e\u9645\u8f93\u51fa\u7684\u7ed3\u679c\u5f80\u5f80\u504f\u79bb\u9884\u671f\u3002<\/p>\n

\u5982\u679c\u4f60\u66fe\u7ecf\u4f7f\u7528\u8fc7Langchain\uff0c\u90a3\u4e48\u4f60\u4e00\u5b9a\u77e5\u9053\u5176\u4e2d\u6709\u4e00\u4e2a\u53eb\u505aOutputParser<\/code>\u7684\u6a21\u5757\uff0c\u8fd9\u4e2a\u6a21\u5757\u4e2d\u5b9a\u4e49\u4e86\u8bb8\u591a\u7528\u4e8e\u5feb\u901f\u7ed3\u6784\u5316\u6570\u636e\u7684\u5de5\u5177\u3002<\/p>\n

\n

\"\u5982\u4f55\u5229\u7528OpenAI\u7684\u51fd\u6570\u8c03\u7528\u7279\u6027\"<\/figure>\n<\/p>\n

\u7136\u800c\uff0c\u9762\u5bf9\u5927\u89c4\u6a21\u7684\u6a21\u578b\uff0c\u8fd9\u4e9b\u5de5\u5177\u5e76\u4e0d\u603b\u662f\u90a3\u4e48\u6709\u6548\u3002\u56e0\u4e3a\u6a21\u578b\u7684\u8f93\u51fa\u7ed3\u679c\u96be\u4ee5\u9884\u6d4b\uff0c\u8fd9\u4f7f\u5f97\u5728\u4e00\u4e9b\u590d\u6742\u573a\u666f\u4e0b\uff0c\u6a21\u578b\u8f93\u51fa\u7684\u7ed3\u679c\u603b\u662f\u65e0\u6cd5\u88ab\u6b63\u786e\u89e3\u6790\u3002\u5373\u4f7f\u662f\u4f7f\u7528\u4e86Langchain\u4e2d\u7684OutputFixingParser<\/code>\uff0c\u4e5f\u96be\u4ee5\u8ba9\u6574\u4e2a\u6d41\u7a0b\u53d8\u5f97\u6d41\u7545\u3002<\/p>\n

\u56e0\u6b64\uff0c\u6211\u4eec\u4ecd\u7136\u9700\u8981\u56de\u5f52\u5230OpenAI\u5b98\u65b9\u63a8\u51fa\u7684\u51fd\u6570\u8c03\u7528<\/code>\u7279\u6027\u3002\u5728\u5b98\u65b9\u63d0\u4f9b\u7684\u4e00\u4e2a\u793a\u4f8b\u4e2d\uff0c\u5b83\u4f7f\u7528function calling<\/code>\u8ba9GPT\u6a21\u578b\u63d0\u53d6\u81ea\u7136\u8bed\u8a00\u4e2d\u53ef\u7528\u4e8e\u51fd\u6570\u8c03\u7528\u7684\u4fe1\u606f<\/strong>\uff0c\u7136\u540e\u5229\u7528\u7b2c\u4e09\u65b9\u63a5\u53e3\u83b7\u53d6\u5b9e\u65f6\u4fe1\u606f\uff0c\u6700\u540e\u751f\u6210\u4e00\u4e2a\u5168\u65b0\u7684\u603b\u7ed3\u3002<\/p>\n

\n

\"\u5982\u4f55\u5229\u7528OpenAI\u7684\u51fd\u6570\u8c03\u7528\u7279\u6027\"<\/figure>\n<\/p>\n

\u5982\u4f55\u4f7f\u7528\u51fd\u6570\u8c03\u7528\u4ee5\u53ca\u5176\u5de5\u4f5c\u539f\u7406\uff1f<\/h2>\n

\u63a5\u4e0b\u6765\uff0c\u8ba9\u6211\u4eec\u501f\u52a9\u5b98\u65b9\u7684\u793a\u4f8b\u6765\u89e3\u6790\u4e00\u4e0b\u5982\u4f55\u4f7f\u7528function calling<\/code>\u8fd9\u4e2a\u7279\u6027\uff0c\u4ee5\u53ca\u5b83\u7a76\u7adf\u662f\u5982\u4f55\u8fd0\u4f5c\u7684\u3002<\/p>\n

\u7b2c\u4e00\u6b65\uff1a\u643a\u5e26\u51fd\u6570\u8c03\u7528\u53bb\u8bf7\u6c42GPT\u63a5\u53e3<\/h3>\n

\u9996\u5148\uff0c\u4f60\u9700\u8981\u628a\u4e00\u4e2a\u95ee\u9898\u548c\u51fd\u6570\u8c03\u7528\u4e00\u8d77\uff0c\u4f5c\u4e3a\u8bf7\u6c42\u63d0\u4ea4\u5230GPT\u7684\u63a5\u53e3\u3002<\/p>\n

<\/div>
makefile<\/span>
\u590d\u5236\u4ee3\u7801<\/div><\/div><\/div>messages = [{\"role\"<\/span>: \"user\"<\/span>, \"content\"<\/span>: \"What's the weather like in Boston?\"<\/span>}]<\/span>\nfunctions = [<\/span>\n        {<\/span>\n            \"name\"<\/span>: \"get_current_weather\"<\/span>,<\/span>\n            \"description\"<\/span>: \"Get the current weather in a given location\"<\/span>,<\/span>\n            \"parameters\"<\/span>: {<\/span>\n                \"type\"<\/span>: \"object\"<\/span>,<\/span>\n                \"properties\"<\/span>: {<\/span>\n                    \"location\"<\/span>: {<\/span>\n                        \"type\"<\/span>: \"string\"<\/span>,<\/span>\n                        \"description\"<\/span>: \"The city and state, e.g. San Francisco, CA\"<\/span>,<\/span>\n                    },<\/span>\n                    \"unit\"<\/span>: {\"type\"<\/span>: \"string\"<\/span>, \"enum\"<\/span>: [\"celsius\"<\/span>, \"fahrenheit\"<\/span>]},<\/span>\n                },<\/span>\n                \"required\"<\/span>: [\"location\"<\/span>],<\/span>\n            },<\/span>\n        }<\/span>\n    ]<\/span>\nresponse = openai.ChatCompletion.create(<\/span>\n        model=\"gpt-3.5-turbo-0613\"<\/span>,<\/span>\n        messages=messages,<\/span>\n        functions=functions,<\/span>\n        function_call=\"auto\"<\/span>,  # auto is default, but we'll be explicit<\/span><\/span>\n    )<\/span>\nresponse_message = response[\"choices\"<\/span>][0][\"message\"<\/span>]<\/span>\n\u200b<\/span>\n\"\"<\/span>\"<\/span><\/span>\n\u8f93\u51fa\uff1a<\/span>\n <OpenAIObject at 0x7a90d69f1620> JSON: {<\/span>\n  \"role\": \"<\/span>assistant\",<\/span><\/span>\n  \"content\": null,<\/span><\/span>\n  \"function_call\": {<\/span><\/span>\n    \"name\": \"<\/span>get_current_weather\",<\/span><\/span>\n    \"arguments\": \"<\/span>{n  \"location\"<\/span>: \"Boston, MA\"<\/span>n}\"<\/span><\/span>\n  }<\/span>\n}<\/span>\n \"\"\"<\/span><\/span>\n<\/code><\/pre>\n
\u5728\u8fd9\u4e2a\u4f8b\u5b50\u4e2d\uff0c\u5148\u662f\u5b9a\u4e49\u4e86\u4e00\u4e2a\u95ee\u9898\uff1a What's the weather like in Boston?<\/code>, \u8fd9\u4e2a\u662f\u6211\u4eec\u60f3\u8981\u56de\u7b54\u7684\u95ee\u9898\uff0c\u5176\u4e2dBoston<\/code>\u662f\u95ee\u9898\u4e2d\u7684\u5730\u540d\uff0cweather<\/code>\u662f\u6211\u4eec\u60f3\u8981\u77e5\u9053\u7684\u4fe1\u606f\u3002\u5728\u4e0b\u9762\u7684function\u4e2d\uff0c\u6211\u4eec\u5b9a\u4e49\u4e86\u4e00\u4e2afunction\u6765\u6307\u660e\u6211\u4eec\u5728\u83b7\u53d6\u4e09\u65b9\u7684api\u65f6\u9700\u8981\u54ea\u4e9b\u53c2\u6570\uff0c\u8fdb\u800c\u8ba9gpt\u77e5\u9053\u5b83\u5e94\u8be5\u4eceWhat's the weather like in Boston?<\/code> \u8fd9\u53e5\u8bdd\u4e2d\u63d0\u53d6\u4ec0\u4e48\u6837\u7684\u4fe1\u606f\uff0c\u4ee5\u53ca\u4ec0\u4e48\u683c\u5f0f\u7684\u4fe1\u606f\u3002<\/p>\n
\u7b2c\u4e8c\u6b65\uff1a\u62ff\u7740\u89e3\u6790\u51fa\u6765\u7684\u4fe1\u606f\u8bf7\u6c42\u63a5\u53e3<\/h3>\n
<\/div>
python<\/span>
\u590d\u5236\u4ee3\u7801<\/div><\/div><\/div>import<\/span> json<\/span>\n# \u8fd9\u662f\u4e00\u4e2a\u865a\u5047\u7684\u51fd\u6570\uff0c\u7528\u6765\u6a21\u62df\u4e09\u65b9\u63a5\u53e3\u7684\u8fd4\u56de\u503c<\/span><\/span>\ndef<\/span> get_current_weather<\/span>(location, unit=\"fahrenheit\"<\/span><\/span>):<\/span>\n    \"\"\"Get the current weather in a given location\"\"\"<\/span><\/span>\n    weather_info = {<\/span>\n        \"location\"<\/span>: location,<\/span>\n        \"temperature\"<\/span>: \"72\"<\/span>,<\/span>\n        \"unit\"<\/span>: unit,<\/span>\n        \"forecast\"<\/span>: [\"sunny\"<\/span>, \"windy\"<\/span>],<\/span>\n    }<\/span>\n    return<\/span> json.dumps(weather_info)<\/span>\n\u200b<\/span>\nif<\/span> response_message.get(\"function_call\"<\/span>):<\/span>\n        # Step 3: call the function<\/span><\/span>\n        # Note: the JSON response may not always be valid; be sure to handle errors<\/span><\/span>\n        available_functions = {<\/span>\n            \"get_current_weather\"<\/span>: get_current_weather,<\/span>\n        }  # only one function in this example, but you can have multiple<\/span><\/span>\n        function_name = response_message[\"function_call\"<\/span>][\"name\"<\/span>]<\/span>\n        fuction_to_call = available_functions[function_name]<\/span>\n        function_args = json.loads(response_message[\"function_call\"<\/span>][\"arguments\"<\/span>])<\/span>\n        function_response = fuction_to_call(<\/span>\n            location=function_args.get(\"location\"<\/span>),<\/span>\n            unit=function_args.get(\"unit\"<\/span>),<\/span>\n        )<\/span>\n\"\"\"<\/span><\/span>\n\u8f93\u51fa\u7684function_response:<\/span>\n{\"location\": \"Boston, MA\", \"temperature\": \"72\", \"unit\": null, \"forecast\": [\"sunny\", \"windy\"]}<\/span>\n\"\"\"<\/span>\n<\/code><\/pre>\n
\u5728\u8fd9\u4e00\u6b65\u4e2d\uff0c\u6211\u4eec\u901a\u8fc7function_name = response_message[\"function_call\"][\"name\"]<\/code> \u8fd9\u4e2a\u6765\u83b7\u53d6\u5230\u4e86gpt\u6a21\u578b\u89e3\u6790\u51fa\u7684\u51fd\u6570\u540d\uff0c\u56e0\u4e3a\u5728\u66f4\u590d\u6742\u7684\u573a\u666f\u4e2d\uff0c\u6211\u4eec\u53ef\u80fd\u4e00\u6b21\u6027\u8ba9gpt\u89e3\u6790\u66f4\u591a\u7684\u65b9\u6cd5\uff0c\u56e0\u6b64\u6211\u4eec\u9700\u8981\u83b7\u53d6\u5230\u6240\u6709\u7684\u51fd\u6570\u540d\u3002<\/p>\n
\u63a5\u7740\u6211\u4eec\u53c8\u901a\u8fc7response_message[\"function_call\"][\"arguments\"]<\/code>\u8fd9\u4e2a\u83b7\u53d6\u5230\u4e86\u89e3\u6790\u51fa\u6765\u7684\u53c2\u6570\u3002\u8fd9\u8fb9\u7684\u53c2\u6570\u5176\u5b9e\u5c31gpt\u901a\u8fc7\u6211\u4eec\u7684\u95ee\u9898\u63d0\u53d6\u5230\u7684\u80fd\u591f\u88ab\u7528\u5230get_current_weather<\/code>\u8fd9\u4e2a\u51fd\u6570\u4e2d\u4f5c\u4e3a\u53c2\u6570\u7684\u5185\u5bb9\u3002\u6bd4\u5982\u73b0\u5728\u6211\u4eec\u9700\u8981\u7684\u4e24\u4e2a\u53c2\u6570\u662flocation<\/code>\u548cunit<\/code>, gpt\u5c31\u4f1a\u6839\u636e\u73b0\u6709\u7684\u8fd9\u4e2a\u95ee\u9898\u6765\u63d0\u53d6\u662flocaltion<\/code>\u548cunit<\/code>\u610f\u601d\u7684\u4fe1\u606f\u3002<\/p>\n
\u7b2c\u4e09\u6b65\u751f\u6210\u4e00\u4e2asummary<\/h3>\n
<\/div>
ini<\/span>
\u590d\u5236\u4ee3\u7801<\/div><\/div><\/div> messages.append(response_message)  # extend conversation with assistant's reply<\/span><\/span>\n        messages.append(<\/span>\n            {<\/span>\n                \"role\": \"function\",<\/span>\n                \"name\": function_name,<\/span>\n                \"content\": function_response,<\/span>\n            }<\/span>\n        )  # extend conversation with function response<\/span><\/span>\n        second_response<\/span> = openai.ChatCompletion.create(<\/span>\n            model<\/span>=\"gpt-3.5-turbo-0613\"<\/span>,<\/span>\n            messages<\/span>=messages,<\/span>\n        )  # get a new response from GPT where it can see the function response<\/span><\/span>\n<\/code><\/pre>\n
\u5728\u6700\u540e\u4e00\u6b65\u53ea\u662f\u5355\u7eaf\u7684\u6839\u636e\u4e0a\u4e0b\u6587\u4fe1\u606f\u751f\u6210\u4e00\u4e2a\u65b0\u7684\u53e5\u5b50\uff0c\u5176\u5b9e\u6ca1\u6709\u4ec0\u4e48\u66f4\u591a\u7684\u4fe1\u606f\u3002<\/p>\n
\u63a5\u4e0b\u6765\u6211\u4eec\u901a\u8fc7\u53e6\u4e00\u4e2a\u7a0d\u5fae\u590d\u6742\u4e00\u70b9\u70b9\u7684\u6848\u4f8b\u518d\u6765\u89e3\u91ca\u4e0bfunction calling<\/code>\u7684\u7528\u6cd5<\/p>\n
\u4e3e\u4e2a\u6817\u5b50 \u8ba9GPT\u673a\u5668\u7ffb\u8bd1<\/h2>\n
\u73b0\u5728\u6211\u4eec\u60f3\u8ba9GPT\u5e2e\u6211\u505a\u4e00\u4e2a\u7ffb\u8bd1\u7684\u529f\u80fd\uff0c\u5e76\u4e14\u7ffb\u8bd1\u7684\u4fe1\u606f\u53ef\u4ee5\u7ed3\u6784\u5316\u7684\u8f93\u51fa\uff0c\u4e8e\u662f\u6211\u5199\u4e86\u8fd9\u4e48\u4e00\u4e2aprompt:<\/p>\n
<\/div>
ini<\/span>
\u590d\u5236\u4ee3\u7801<\/div><\/div><\/div>prompt_template<\/span> = \"\"\"<\/span><\/span>\n Now you have to translate the following text into {lang}:<\/span>\n    ### Text to translate:<\/span>\n    {text}<\/span>\n    ### Target language:<\/span>\n    {lang}<\/span>\n    ### Output format:<\/span>\n    your translation should in the following format:<\/span>\n    {{'res': 'your translation'}}<\/span>\n\"\"\"<\/span>\n<\/code><\/pre>\n
\u5176\u5b9e\uff0c\u5728\u5927\u90e8\u5206\u60c5\u51b5\u53ea\u4e0b\uff0c\u8fd9\u4e2aprompt\u5c31\u5df2\u7ecf\u53ef\u4ee5\u8f93\u51fa\u4e00\u4efd\u7ed3\u6784\u5316\u7684\u6570\u636e\u4e86\uff0c\u4f46\u662f\u5728\u590d\u6742\u573a\u666f\u6bd4\u5982\u4f60\u7684\u6587\u672c\u8f83\u957f\uff0c\u6216\u8005\u5c1d\u8bd5\u7ffb\u8bd1\u5927\u91cf\u6587\u672c\u7684\u65f6\u5019\uff0c\u6a21\u578b\u603b\u4f1a\u51fa\u73b0\u4e00\u4e9b\u989d\u5916\u7684\u4fe1\u606f\uff0c\u4e0d\u4ec5\u4ec5\u53ea\u6709\u4f60\u60f3\u8981\u7684\u8fd9\u4e2ajson\u683c\u5f0f\uff0c\u53ef\u80fd\u8fd8\u6709\u5176\u4ed6\u8fde\u5e26\u51fa\u6765\u7684\u8f93\u51fa\u3002<\/p>\n
\u73b0\u5728\u6211\u60f3\u8981\u505a\u7684\u662f\uff0c\u4f7f\u7528function calling<\/code> \u6765\u89e3\u6790\u6211\u4eec\u7684\u8f93\u51fa\uff0c\u6211\u53ea\u60f3\u8981\u6700\u540e\u7684\u7ffb\u8bd1\u7ed3\u679c\u3002\u6bd4\u8f83\u7701\u94b1\u4f46\u662f\u8d39\u529b\u7684\u65b9\u6cd5\u5c31\u662f\u6211\u76f4\u63a5\u63d0\u53d6\u6700\u7ec8\u8fd4\u56de\u6570\u636e\u4e2d\u7684 {res : xxx}<\/code> \u8fd9\u4e2a\u5b57\u7b26\u4e32\uff0c\u5e76\u83b7\u5f97\u5b83\u7684\u503c\u3002 \u4f46\u662f\u5728\u8fd9\u91cc\u6211\u4eec\u8fd8\u662f\u6253\u7b97\u4f7f\u7528function calling<\/code>\u6765\u5e2e\u6211\u4eec\u89e3\u6790\u8fd9\u4e2a\u6570\u636e\u3002<\/p>\n
\u73b0\u5728\u6211\u4eec\u7684\u7ffb\u8bd1\u7684\u6267\u884c\u662f\u8fd9\u6837\u7684\uff1a<\/p>\n
<\/div>
python<\/span>
\u590d\u5236\u4ee3\u7801<\/div><\/div><\/div>text = 'Collection of Open Source Projects Related to GPT\uff0cGPT\u76f8\u5173\u5f00\u6e90\u9879\u76ee\u5408\u96c6\ud83d\ude80\u3001\u7cbe\u9009\ud83d\udd25\ud83d\udd25'<\/span><\/span>\nlang = 'en'<\/span><\/span>\n\u200b<\/span>\nprompt = prompt_template.format<\/span>(text=text, lang=lang)<\/span>\n# completion\u662f\u6211\u4eec\u5c01\u88c5\u7684\u4e00\u4e2a\u8f6c\u53d1openai\u8bf7\u6c42\u7684\u51fd\u6570\uff0c\u6ca1\u6709\u4ec0\u4e48\u7279\u522b\u7684\u5b9e\u73b0<\/span><\/span>\nresponse = completion(model_name=\"gpt-3.5-turbo-0613\"<\/span>, prompt=prompt)<\/span>\n\u200b<\/span>\nresponse.content<\/span>\n\"\"\"<\/span><\/span>\n\u8f93\u51fa\uff1a<\/span>\n{'res': 'Collection of Open Source Projects Related to GPT, Compilation of open source projects related to GPT \ud83d\ude80, selected \ud83d\udd25\ud83d\udd25'}<\/span>\n\"\"\"<\/span>\n<\/code><\/pre>\n
\u73b0\u5728\u6211\u4eec\u5b9a\u4e49\u4e00\u4e2aResponseSchema<\/code>\u6765\u4ee3\u8868\u8fd9\u4e2a\u8f93\u51fa\u3002 \u56e0\u4e3a\u5728\u590d\u6742\u7684\u751f\u4ea7\u573a\u666f\u4e2d\uff0c\u6211\u4eec\u5f80\u5f80\u9700\u8981\u89e3\u6790\u7684\u5bf9\u8c61\u662f\u5f88\u590d\u6742\u7684\uff0c\u662f\u6709\u591a\u4e2a\u5b57\u6bb5\u4ee5\u53ca\u4e0d\u540c\u7c7b\u578b\u7684\uff0c\u6211\u4eec\u4e0d\u53ef\u80fd\u50cf\u5b98\u65b9\u7684\u4f8b\u5b50\u4e00\u6837\u624b\u52a8\u53bb\u53bb\u5199\u6bcf\u4e00\u4e2aparameters\u3002<\/p>\n
\u5728\u8fd9\u91cc\u6211\u4f7f\u7528\u4e86python\u7684pydantic<\/code>\u5305\u6765\u5b9e\u73b0\u8fd9\u4e48\u4e00\u4e2a\u529f\u80fd\u3002<\/p>\n
<\/div>
python<\/span>
\u590d\u5236\u4ee3\u7801<\/div><\/div><\/div>from<\/span> pydantic import<\/span> BaseModel, Field<\/span>\nimport<\/span> json<\/span>\n\u200b<\/span>\nclass<\/span> ResponseSchema<\/span>(BaseModel<\/span>):<\/span>\n    result: str<\/span> = Field(..., description=\"the translated value\"<\/span>)<\/span>\n\u200b<\/span>\n    @classmethod<\/span><\/span>\n    def<\/span> schema<\/span>(cls, by_alias=True<\/span><\/span>):<\/span>\n        schema = super<\/span>().schema(by_alias)<\/span>\n        schema.pop('title'<\/span>, None<\/span>)<\/span>\n        return<\/span> schema<\/span>\n    @classmethod<\/span><\/span>\n    def<\/span> schema_json<\/span>(cls, by_alias=True<\/span>, **dumps_kwargs<\/span>):<\/span>\n\u200b<\/span>\n        schema_str = json.dumps(cls.schema(by_alias), **dumps_kwargs)<\/span>\n        return<\/span> json.loads(schema_str)<\/span>\n<\/code><\/pre>\n
\u5728\u8fd9\u91cc\u6211\u624b\u52a8\u7684\u5b9e\u73b0\u4e86schema<\/code>\u548c schema_json<\/code> \u51fd\u6570\uff0c\u56e0\u4e3a pydantic<\/code>\u7684\u6a21\u578b\u7c7b\u603b\u4f1a\u81ea\u5e26\u4e00\u4e2atitle<\/code>\u7684\u5c5e\u6027\uff0c\u8fd9\u4e2a\u5c5e\u6027\u5f88\u5927\u7a0b\u5ea6\u4e0a\u4f1a\u5f71\u54cdGPT\u5bf9\u6211\u4eec\u6700\u7ec8\u60f3\u8981\u7684\u8f93\u51fa\u7684\u7406\u89e3\u3002<\/p>\n
\u6211\u4eec\u6267\u884cResponseSchema.schema_json()<\/code>\u5c31\u53ef\u4ee5\u83b7\u53d6\u5230\u548c\u4e0a\u9762\u5b98\u65b9\u4f8b\u5b50\u4e00\u6837\u7684\u4e00\u4e2a\u5173\u4e8e\u5b57\u6bb5\u63cf\u8ff0\u7684\u7ed3\u679c\uff1a<\/p>\n
<\/div>
arduino<\/span>
\u590d\u5236\u4ee3\u7801<\/div><\/div><\/div>{'type'<\/span>: 'object'<\/span>,<\/span>\n 'properties'<\/span>: {'result'<\/span>: {'title'<\/span>: 'Result'<\/span>,<\/span>\n   'description'<\/span>: 'the translated value'<\/span>,<\/span>\n   'type'<\/span>: 'string'<\/span>}},<\/span>\n 'required'<\/span>: ['result'<\/span>]}<\/span>\n<\/code><\/pre>\n
\u4e8e\u662f\u6211\u4eec\u7684function\u53ea\u9700\u8981\u5199\u6210\u8fd9\u6837\u5c31\u53ef\u4ee5\u4e86\uff1a<\/p>\n
<\/div>
ini<\/span>
\u590d\u5236\u4ee3\u7801<\/div><\/div><\/div>functions<\/span>=[<\/span>\n            {<\/span>\n                \"name\"<\/span>: \"print_translation\"<\/span>,<\/span>\n                \"description\"<\/span>: \"print the translated value from value of res\"<\/span>,<\/span>\n                \"parameters\"<\/span>: ResponseSchema.schema_json(),<\/span>\n            },<\/span>\n        ]<\/span>\n<\/code><\/pre>\n
\u5728\u8fd9\u4e2afunction\u4e2d\uff0c\u6211\u4eec\u7684print_translation<\/code> \u51fd\u6570\u5176\u5b9e\u662f\u865a\u5047\u7684\u51fd\u6570<\/strong>\uff0c\u6211\u4eec\u53ea\u662f\u7ed9\u4ed6\u8d77\u4e86\u4e00\u4e2a\u5408\u7406\u7684\u540d\u5b57\u4ee5\u53ca\u5408\u7406\u7684\u63cf\u8ff0<\/strong>\uff0c\u8fd9\u6837GPT\u5c31\u53ef\u4ee5\u77e5\u9053\u6211\u4eec\u7684\u53c2\u6570\u7684\u4e0a\u4e0b\u6587\u662f\u4ec0\u4e48\u6837\u7684\u4e86\u3002<\/p>\n
\u5b8c\u6574\u7684\u4e00\u4e2aflow\u5982\u4e0b\uff1a<\/p>\n
<\/div>
ini<\/span>
\u590d\u5236\u4ee3\u7801<\/div><\/div><\/div>import openai<\/span>\nfrom pydantic import BaseModel, Field<\/span>\nimport json<\/span>\n\u200b<\/span>\n\u200b<\/span>\ndef completion(model_name, prompt):<\/span>\n  completion<\/span> = openai.ChatCompletion.create(<\/span>\n            stream<\/span>=False<\/span>,<\/span>\n            model<\/span>=model_name,<\/span>\n            messages<\/span>=[<\/span>\n                {\"role\"<\/span>: \"user\"<\/span>, \"content\"<\/span>: prompt}<\/span>\n            ]<\/span>\n        )<\/span>\n\u200b<\/span>\n\u200b<\/span>\n  return completion.choices[0]<\/span>.message<\/span>\n\u200b<\/span>\nclass ResponseSchema(BaseModel):<\/span>\n    result: str<\/span> = Field(..., description=\"the translated value\"<\/span>)<\/span>\n\u200b<\/span>\n    @classmethod<\/span>\n    def schema(cls, by_alias<\/span>=True<\/span>):<\/span>\n        schema<\/span> = super().schema(by_alias)<\/span>\n        schema.pop('title', None)<\/span>\n        return schema<\/span>\n    @classmethod<\/span>\n    def schema_json(cls, by_alias<\/span>=True<\/span>, **dumps_kwargs):<\/span>\n\u200b<\/span>\n        schema_str<\/span> = json.dumps(cls.schema(by_alias), **dumps_kwargs)<\/span>\n        return json.loads(schema_str)<\/span>\n\u200b<\/span>\n###  \u8ba9GPT\u5e2e\u6211\u751f\u6210\u5bf9\u5e94\u8bed\u53e5\u7684\u7ffb\u8bd1\uff0c\u5e76\u8fd4\u56de\u6210json<\/span><\/span>\nprompt_template<\/span> = \"\"\"<\/span><\/span>\n Now you have to translate the following text into {lang}:<\/span>\n    ### Text to translate:<\/span>\n    {text}<\/span>\n    ### Target language:<\/span>\n    {lang}<\/span>\n    ### Output format:<\/span>\n    your translation should in the following format:<\/span>\n    {{'res': 'your translation'}}<\/span>\n\"\"\"<\/span>\n\u200b<\/span>\n\u200b<\/span>\ntext<\/span> = 'Collection of Open Source Projects Related to GPT\uff0cGPT\u76f8\u5173\u5f00\u6e90\u9879\u76ee\u5408\u96c6\ud83d\ude80\u3001\u7cbe\u9009\ud83d\udd25\ud83d\udd25'<\/span><\/span>\nlang<\/span> = 'en'<\/span><\/span>\n\u200b<\/span>\nfunctions<\/span>=[<\/span>\n            {<\/span>\n                \"name\"<\/span>: \"print_translation\"<\/span>,<\/span>\n                \"description\"<\/span>: \"print the translated value from value of res\"<\/span>,<\/span>\n                \"parameters\"<\/span>: ResponseSchema.schema_json(),<\/span>\n            },<\/span>\n       ]<\/span>\n\u200b<\/span>\nprompt<\/span> = prompt_template.format(text=text, lang=lang)<\/span>\nresponse<\/span> = completion(model_name=\"gpt-3.5-turbo-0613\"<\/span>, prompt=prompt)<\/span>\nfinal_result<\/span> = openai.ChatCompletion.create(<\/span>\n    model<\/span>=\"gpt-3.5-turbo-0613\"<\/span>,<\/span>\n    messages<\/span>=[{\"role\"<\/span>: \"user\"<\/span>, \"content\"<\/span>: prompt}],<\/span>\n    functions<\/span>=functions,<\/span>\n        function_call<\/span>=\"auto\"<\/span>,<\/span>\n    )<\/span>\nmessage<\/span> = final_result[\"choices\"<\/span>][0<\/span>][\"message\"<\/span>]<\/span>\neval(message['function_call']<\/span>['arguments']<\/span>)['result']<\/span><\/span>\n\u200b<\/span>\n\"\"\"<\/span>\n\u6700\u7ec8\u8f93\u51fa\uff1a<\/span>\nCollection of Open Source Projects Related to GPT, GPT Related Open Source Project Collection \ud83d\ude80, Selected \ud83d\udd25\ud83d\udd25<\/span>\n\"\"\"<\/span>\n\u200b<\/span>\n<\/code><\/pre>\n
\u603b\u7ed3<\/h2>\n
\u5728\u672c\u6587\u4e2d\uff0c\u6211\u4eec\u6df1\u5165\u63a2\u7d22\u4e86OpenAI\u7684\u51fd\u6570\u8c03\u7528\u7279\u6027\uff0c\u5e76\u89e3\u6790\u4e86\u5176\u5728\u5b9e\u73b0\u7ed3\u6784\u5316\u6570\u636e\u751f\u6210\u4e2d\u7684\u5de8\u5927\u4ef7\u503c\u3002\u867d\u7136\u6211\u4eec\u901a\u5e38\u8ba4\u4e3a\u751f\u6210\u6a21\u578b\u4ea7\u751f\u7684\u6570\u636e\u662f\u65e0\u7ed3\u6784\u7684\uff0c\u4f46\u5728\u6587\u7ae0\u4e2d\u8bc1\u660e\uff0c\u501f\u52a9\u51fd\u6570\u8c03\u7528\u7279\u6027\uff0c\u6211\u4eec\u5b8c\u5168\u6709\u53ef\u80fd\u521b\u5efa\u51fa\u7ed3\u6784\u5316\u7684\u6570\u636e\u3002<\/p>\n
\u901a\u8fc7\u7f16\u7a0b\u793a\u4f8b\uff0c\u6211\u4eec\u8be6\u7ec6\u89e3\u91ca\u4e86\u51fd\u6570\u8c03\u7528\u7684\u5de5\u4f5c\u539f\u7406\uff0c\u5305\u62ec\u5982\u4f55\u5411GPT\u63a5\u53e3\u63d0\u4ea4\u5e26\u6709\u95ee\u9898\u548c\u51fd\u6570\u8c03\u7528\u7684\u8bf7\u6c42\uff0c\u5982\u4f55\u5229\u7528\u6a21\u578b\u63d0\u53d6\u51fa\u7684\u4fe1\u606f\u53bb\u8bf7\u6c42\u7b2c\u4e09\u65b9\u63a5\u53e3\u83b7\u53d6\u5b9e\u65f6\u6570\u636e\uff0c\u4ee5\u53ca\u5982\u4f55\u751f\u6210\u4e00\u4e2a\u5168\u65b0\u7684\u603b\u7ed3\u3002<\/p>\n
\u63a5\u7740\uff0c\u6211\u4eec\u5c55\u793a\u4e86\u5982\u4f55\u501f\u52a9\u51fd\u6570\u8c03\u7528\u7279\u6027\u5b9e\u73b0\u673a\u5668\u7ffb\u8bd1\u529f\u80fd\uff0c\u5e76\u7ed3\u6784\u5316\u8f93\u51fa\u7ffb\u8bd1\u7684\u4fe1\u606f\u3002\u672c\u6587\u5411\u5927\u5bb6\u5c55\u793a\u4e86\uff0c\u901a\u8fc7\u4e3a\u865a\u62df\u51fd\u6570\u5b9a\u4e49\u5408\u7406\u7684\u540d\u5b57\u548c\u63cf\u8ff0\uff0cGPT\u6a21\u578b\u80fd\u591f\u7406\u89e3\u53c2\u6570\u7684\u4e0a\u4e0b\u6587\uff0c\u5e76\u6309\u7167\u9884\u8bbe\u7684\u683c\u5f0f\u751f\u6210\u7ed3\u6784\u5316\u7684\u7ffb\u8bd1\u7ed3\u679c\u3002<\/p>\n
\u603b\u7684\u6765\u8bf4\uff0c\u672c\u6587\u4e3a\u5927\u5bb6\u63d0\u4f9b\u4e86\u4e00\u4e2a\u5b9e\u7528\u7684\u3001\u7ed3\u6784\u5316\u7684\u6570\u636e\u751f\u6210\u65b9\u6cd5\uff0c\u5e0c\u671b\u8fd9\u5bf9\u4e8e\u60a8\u5728\u5927\u89c4\u6a21\u6a21\u578b\u7684\u6570\u636e\u89e3\u6790\u4ee5\u53ca\u591a\u6837\u5316\u7684\u4fe1\u606f\u63d0\u53d6\u4e0a\u80fd\u591f\u5e26\u6765\u5b9e\u8d28\u6027\u7684\u5e2e\u52a9\u3002<\/p>","protected":false},"excerpt":{"rendered":"
\u5982\u4f55\u5229\u7528OpenAI\u7684\u51fd\u6570\u8c03\u7528\u7279\u6027 \u51fd\u6570\u8c03\u7528\u80fd\u5b9e\u73b0\u54ea\u4e9b\u529f\u80fd\uff1f \u7b80\u5355\u6765\u8bf4\uff0c\u51fd\u6570\u8c03\u7528\u529f\u80fd\u53ef\u4ee5\u52a9\u4f60\u5728\u8bf7\u6c42\u65b9\u6cd5\u65f6\u6784\u5efa\u7ed3\u6784\u5316\u7684\u6570\u636e\u3002\u56e0\u4e3a\u751f\u6210\u6a21\u578b\u7684\u7279\u6027\uff0c\u5b83\u4ea7\u751f\u7684\u6570\u636e\u5f80\u5f80\u662f\u65e0\u7ed3\u6784\u7684\uff0c\u5373\u4f7f\u5728\u63d0\u793a(prompt)\u4e2d\u6307<\/p>\n","protected":false},"author":1,"featured_media":2865,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"rank_math_title":"OpenAI\u51fd\u6570\u8c03\u7528\uff1a\u7ed3\u6784\u5316\u6570\u636e\u751f\u6210\u5229\u5668 - \u5948\u65af\u79d1\u6280\u793e\u533a","rank_math_description":"\u63a2\u7d22OpenAI\u51fd\u6570\u8c03\u7528\u7279\u6027\uff0c\u4e86\u89e3\u5982\u4f55\u5229\u7528\u5b83\u5b9e\u73b0\u7ed3\u6784\u5316\u6570\u636e\u751f\u6210\uff0c\u5e76\u901a\u8fc7\u5b9e\u9645\u6848\u4f8b\u89e3\u6790\u5176\u5728\u673a\u5668\u7ffb\u8bd1\u4e2d\u7684\u5e94\u7528\u3002\u638c\u63e1\u8fd9\u9879\u6280\u672f\uff0c\u63d0\u5347\u5927\u89c4\u6a21\u6a21\u578b\u6570\u636e\u89e3\u6790\u548c\u4fe1\u606f\u63d0\u53d6\u6548\u7387\u3002","rank_math_focus_keyword":"OpenAI, \u51fd\u6570\u8c03\u7528, \u7ed3\u6784\u5316\u6570\u636e, GPT\u6a21\u578b, \u673a\u5668\u7ffb\u8bd1","views":"43","footnotes":""},"categories":[3],"tags":[126,127,128,129,136],"collection":[],"class_list":["post-1245","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-fenlei2","tag-gpt","tag-ai","tag-128","tag-129","tag-136"],"acf":[],"_links":{"self":[{"href":"https:\/\/www.nicekj.com\/nicekj2024\/wp\/v2\/posts\/1245","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.nicekj.com\/nicekj2024\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.nicekj.com\/nicekj2024\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.nicekj.com\/nicekj2024\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.nicekj.com\/nicekj2024\/wp\/v2\/comments?post=1245"}],"version-history":[{"count":0,"href":"https:\/\/www.nicekj.com\/nicekj2024\/wp\/v2\/posts\/1245\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.nicekj.com\/nicekj2024\/wp\/v2\/media\/2865"}],"wp:attachment":[{"href":"https:\/\/www.nicekj.com\/nicekj2024\/wp\/v2\/media?parent=1245"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.nicekj.com\/nicekj2024\/wp\/v2\/categories?post=1245"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.nicekj.com\/nicekj2024\/wp\/v2\/tags?post=1245"},{"taxonomy":"collection","embeddable":true,"href":"https:\/\/www.nicekj.com\/nicekj2024\/wp\/v2\/collection?post=1245"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}