可以通过很多方式来使用，比如使用官方SDK，第三方项目，但其实只需要一个HTTP请求就可以。以下是官方文档给出的例子：
curl https://api.openai.com/v1/chat/completions \
-H 'Content-Type: application/json' \
-H 'Authorization: Bearer YOUR_API_KEY' \
-d '{
"model": "gpt-3.5-turbo",
"messages": [{"role": "user", "content": "Hello!"}]
}'
从里边可以看到，需要的信息有：
① 请求地址： https://api.openai.com/v1/chat/completions 这个地址目前在国内大部分地区已经无法访问了，后边会讲解决办法
② 最常用的接口参数包括：
- model: 必填，建议使用
gpt-3.5-turbo，便宜。计费后边会讲。 - messages: AI 进行提问的问题或信息。
- max_tokens: 选填，指定生成回答的最大Token数。
- stream: 选填，是否按流的方式发送内容。
其中 messages的格式为：{"role","content"}。一般用 user 发送用户问题；system 发送给模型提示信息。
例如：
[
{"role": "system", "content": "You are a helpful assistant that translates English to French."},
{"role": "user", "content": "Translate the following English text to French: {text}"}
]
知道了这些基本就可以跑通GPT流程了，其他role可以稍后优化时来做。
Stream 参数
这里单独说一下 stream 参数，当它设置为 true 时，API 会以 SSE（ Server Side Event ）方式返回内容。
SSE 本质上还是 HTTP 协议，只不过它是一个长链接，先输出一个 header("Content-Type: text/event-stream") ， 然后持续不断地输出内容直到完成。如果不是做实时聊天，建议直接false掉。
需要注意的是，开启stream 后，将不会返回 usage 信息，这对精准计费有影响
{"id":"chatcmpl-6s3hNohxOliHi8zR7m5UTrLm4cWWc","object":"chat.completion.chunk","created":1678341949,"model":"gpt-3.5-turbo-0301","choices":[{"delta":{"content":"我"},"index":0,"finish_reason":null}]}
{"id":"chatcmpl-6s3hNohxOliHi8zR7m5UTrLm4cWWc","object":"chat.completion.chunk","created":1678341949,"model":"gpt-3.5-turbo-0301","choices":[{"delta":{"content":"没有"},"index":0,"finish_reason":null}]}
{"id":"chatcmpl-6s3hNohxOliHi8zR7m5UTrLm4cWWc","object":"chat.completion.chunk","created":1678341949,"model":"gpt-3.5-turbo-0301","choices":[{"delta":{"content":"当前"},"index":0,"finish_reason":null}]}
{"id":"chatcmpl-6s3hNohxOliHi8zR7m5UTrLm4cWWc","object":"chat.completion.chunk","created":1678341949,"model":"gpt-3.5-turbo-0301","choices":[{"delta":{"content":"日期"},"index":0,"finish_reason":null}]}
{"id":"chatcmpl-6s3hNohxOliHi8zR7m5UTrLm4cWWc","object":"chat.completion.chunk","created":1678341949,"model":"gpt-3.5-turbo-0301","choices":[{"delta":{"content":"的"},"index":0,"finish_reason":null}]}
{"id":"chatcmpl-6s3hNohxOliHi8zR7m5UTrLm4cWWc","object":"chat.completion.chunk","created":1678341949,"model":"gpt-3.5-turbo-0301","choices":[{"delta":{"content":"实"},"index":0,"finish_reason":null}]}
{"id":"chatcmpl-6s3hNohxOliHi8zR7m5UTrLm4cWWc","object":"chat.completion.chunk","created":1678341949,"model":"gpt-3.5-turbo-0301","choices":[{"delta":{"content":"时"},"index":0,"finish_reason":null}]}
{"id":"chatcmpl-6s3hNohxOliHi8zR7m5UTrLm4cWWc","object":"chat.completion.chunk","created":1678341949,"model":"gpt-3.5-turbo-0301","choices":[{"delta":{"content":"信息"},"index":0,"finish_reason":null}]}
请求和响应对应的struct
package openai import ( "context" "errors" "net/http" ) // Chat message role defined by the OpenAI API. const ( ChatMessageRoleSystem = "system" ChatMessageRoleUser = "user" ChatMessageRoleAssistant = "assistant" ) var ( ErrChatCompletionInvalidModel = errors.New("this model is not supported with this method, please use CreateCompletion client method instead") //nolint:lll ErrChatCompletionStreamNotSupported = errors.New("streaming is not supported with this method, please use CreateChatCompletionStream") //nolint:lll ) type ChatCompletionMessage struct { Role string `json:"role"` Content string `json:"content"` // This property isn't in the official documentation, but it's in // the documentation for the official library for python: // - https://github.com/openai/openai-python/blob/main/chatml.md // - https://github.com/openai/openai-cookbook/blob/main/examples/How_to_count_tokens_with_tiktoken.ipynb Name string `json:"name,omitempty"` } // ChatCompletionRequest represents a request structure for chat completion API. type ChatCompletionRequest struct { Model string `json:"model"` Messages []ChatCompletionMessage `json:"messages"` MaxTokens int `json:"max_tokens,omitempty"` Temperature float32 `json:"temperature,omitempty"` TopP float32 `json:"top_p,omitempty"` N int `json:"n,omitempty"` Stream bool `json:"stream,omitempty"` Stop []string `json:"stop,omitempty"` PresencePenalty float32 `json:"presence_penalty,omitempty"` FrequencyPenalty float32 `json:"frequency_penalty,omitempty"` LogitBias map[string]int `json:"logit_bias,omitempty"` User string `json:"user,omitempty"` } type ChatCompletionChoice struct { Index int `json:"index"` Message ChatCompletionMessage `json:"message"` FinishReason string `json:"finish_reason"` } // ChatCompletionResponse represents a response structure for chat completion API. type ChatCompletionResponse struct { ID string `json:"id"` Object string `json:"object"` Created int64 `json:"created"` Model string `json:"model"` Choices []ChatCompletionChoice `json:"choices"` Usage Usage `json:"usage"` } // CreateChatCompletion — API call to Create a completion for the chat message. func (c *Client) CreateChatCompletion( ctx context.Context, request ChatCompletionRequest, ) (response ChatCompletionResponse, err error) { if request.Stream { err = ErrChatCompletionStreamNotSupported return } urlSuffix := "/chat/completions" if !checkEndpointSupportsModel(urlSuffix, request.Model) { err = ErrChatCompletionInvalidModel return } req, err := c.requestBuilder.build(ctx, http.MethodPost, c.fullURL(urlSuffix), request) if err != nil { return } err = c.sendRequest(req, &response) return }
对应的测试脚本
curl -X POST "http://localhost:8081/sse" \ -H "Content-Type: application/json" \ -d '{ "model": "gpt-3.5-turbo", "stream":true, "messages": [ { "role": "system", "content": "Welcome to the chat!" }, { "role": "user", "content": "hi" } ], "max_tokens": 600 }'