Skip to content

Commit 144c54b

Browse files
bigwhitebigwhite
committed
add ai-app-primer
1 parent 4212bc6 commit 144c54b

File tree

2 files changed

+354
-0
lines changed

2 files changed

+354
-0
lines changed

‎ai-app-primer/ch03/hello_ai.go

Lines changed: 129 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,129 @@
1+
package main
2+
3+
import (
4+
"bytes"
5+
"encoding/json"
6+
"fmt"
7+
"io"
8+
"net/http"
9+
"os"
10+
"time"
11+
)
12+
13+
// 请求结构体 (简化版，只包含核心字段)
14+
type ChatCompletionRequest struct {
15+
Model string `json:"model"`
16+
Messages []ChatMessage `json:"messages"`
17+
}
18+
19+
type ChatMessage struct {
20+
Role string `json:"role"`
21+
Content string `json:"content"`
22+
}
23+
24+
// 响应结构体 (简化版，只关注核心信息)
25+
type ChatCompletionResponse struct {
26+
ID string `json:"id"`
27+
Choices []Choice `json:"choices"`
28+
Usage UsageInfo `json:"usage"`
29+
// 为简化，省略了 object, created, model, system_fingerprint 等
30+
}
31+
32+
type Choice struct {
33+
Message ResponseMessage `json:"message"`
34+
FinishReason string `json:"finish_reason"`
35+
// 为简化，省略了 index, logprobs 等
36+
}
37+
38+
type ResponseMessage struct {
39+
Role string `json:"role"`
40+
Content string `json:"content"` // 注意：在实际应用中，如果 content 可能为 null (如 tool_calls 时)，应使用 *string
41+
}
42+
43+
type UsageInfo struct {
44+
PromptTokens int `json:"prompt_tokens"`
45+
CompletionTokens int `json:"completion_tokens"`
46+
TotalTokens int `json:"total_tokens"`
47+
}
48+
49+
func main() {
50+
apiKey := os.Getenv("OPENAI_API_KEY")
51+
if apiKey == "" {
52+
fmt.Println("Error: OPENAI_API_KEY environment variable not set.")
53+
return
54+
}
55+
56+
// DeepSeek API 配置
57+
apiURL := "https://api.deepseek.com/chat/completions" // 或者 "https://api.deepseek.com/v1/chat/completions"
58+
modelID := "deepseek-chat" // 使用 DeepSeek 的一个聊天模型
59+
60+
// 构造请求体
61+
requestPayload := ChatCompletionRequest{
62+
Model: modelID,
63+
Messages: []ChatMessage{
64+
{Role: "system", Content: "You are a friendly and helpful assistant."}, // System (or Developer) message
65+
{Role: "user", Content: "Hello, AI! How are you today?"}, // User message
66+
},
67+
}
68+
69+
requestBodyBytes, err := json.Marshal(requestPayload)
70+
if err != nil {
71+
fmt.Printf("Error marshalling request: %v\n", err)
72+
return
73+
}
74+
75+
// 创建 HTTP 请求
76+
req, err := http.NewRequest("POST", apiURL, bytes.NewBuffer(requestBodyBytes))
77+
if err != nil {
78+
fmt.Printf("Error creating request: %v\n", err)
79+
return
80+
}
81+
req.Header.Set("Content-Type", "application/json")
82+
req.Header.Set("Authorization", "Bearer "+apiKey)
83+
84+
// 发送请求
85+
client := &http.Client{Timeout: 30 * time.Second}
86+
resp, err := client.Do(req)
87+
if err != nil {
88+
fmt.Printf("Error sending request: %v\n", err)
89+
return
90+
}
91+
defer resp.Body.Close()
92+
93+
// 读取响应体
94+
responseBodyBytes, err := io.ReadAll(resp.Body)
95+
if err != nil {
96+
fmt.Printf("Error reading response body: %v\n", err)
97+
return
98+
}
99+
100+
// 检查 HTTP 状态码
101+
if resp.StatusCode != http.StatusOK {
102+
fmt.Printf("Error: Non-OK status code: %d\nResponse: %s\n", resp.StatusCode, string(responseBodyBytes))
103+
return
104+
}
105+
106+
// 解析 JSON 响应
107+
var chatResponse ChatCompletionResponse
108+
err = json.Unmarshal(responseBodyBytes, &chatResponse)
109+
if err != nil {
110+
fmt.Printf("Error unmarshalling response: %v\n", err)
111+
fmt.Printf("Raw Response: %s\n", string(responseBodyBytes))
112+
return
113+
}
114+
115+
// 打印关键信息
116+
fmt.Println("--- AI Response ---")
117+
if len(chatResponse.Choices) > 0 {
118+
fmt.Printf("ID: %s\n", chatResponse.ID)
119+
fmt.Printf("Assistant: %s\n", chatResponse.Choices[0].Message.Content)
120+
fmt.Printf("Finish Reason: %s\n", chatResponse.Choices[0].FinishReason)
121+
fmt.Printf("Usage: PromptTokens=%d, CompletionTokens=%d, TotalTokens=%d\n",
122+
chatResponse.Usage.PromptTokens,
123+
chatResponse.Usage.CompletionTokens,
124+
chatResponse.Usage.TotalTokens,
125+
)
126+
} else {
127+
fmt.Println("No choices received.")
128+
}
129+
}

‎ai-app-primer/ch03/multi_turn_chat.go

Lines changed: 225 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,225 @@
1+
package main
2+
3+
import (
4+
"bytes"
5+
"encoding/json"
6+
"fmt"
7+
"io"
8+
"net/http"
9+
"os"
10+
"strings"
11+
"time"
12+
)
13+
14+
// ChatMessage
15+
type ChatMessage struct {
16+
Role string `json:"role"`
17+
Content string `json:"content"`
18+
// 省略 ToolCalls 等仅用于工具调用的字段
19+
}
20+
21+
// ChatCompletionRequest
22+
type ChatCompletionRequest struct {
23+
Model string `json:"model"`
24+
Messages []ChatMessage `json:"messages"` // 包含完整的对话历史
25+
Temperature float64 `json:"temperature,omitempty"`
26+
// 使用明确的 max_completion_tokens (如果 API 支持) 或 max_tokens
27+
MaxTokens int `json:"max_tokens,omitempty"` // 兼容旧模型或 Ollama
28+
Stream bool `json:"stream,omitempty"`
29+
}
30+
31+
// ResponseMessage
32+
type ResponseMessage struct {
33+
Role string `json:"role"`
34+
Content *string `json:"content"` // 使用指针处理 null
35+
// 省略 ToolCalls
36+
}
37+
38+
// Choice
39+
type Choice struct {
40+
Index int `json:"index"`
41+
Message ResponseMessage `json:"message"`
42+
FinishReason string `json:"finish_reason"`
43+
// 省略 logprobs
44+
}
45+
46+
// UsageInfo
47+
type UsageInfo struct {
48+
PromptTokens int `json:"prompt_tokens"`
49+
CompletionTokens int `json:"completion_tokens"`
50+
TotalTokens int `json:"total_tokens"`
51+
}
52+
53+
// ChatCompletionResponse
54+
type ChatCompletionResponse struct {
55+
ID string `json:"id"`
56+
Object string `json:"object"`
57+
Created int64 `json:"created"`
58+
Model string `json:"model"`
59+
Choices []Choice `json:"choices"`
60+
Usage UsageInfo `json:"usage"`
61+
// 省略 system_fingerprint
62+
}
63+
64+
// --- 主函数 ---
65+
66+
const maxTurns = 5 // 控制对话轮数
67+
68+
func main() {
69+
apiKey := os.Getenv("OPENAI_API_KEY")
70+
// 对本地 Ollama 等不需要 key 的服务，apiKey 可以为空
71+
// if apiKey == "" {
72+
// fmt.Println("Warning: OPENAI_API_KEY environment variable not set. Assuming local service.")
73+
// }
74+
75+
// --- 配置 API 端点和模型 ---
76+
// apiURL := "https://api.openai.com/v1/chat/completions" // OpenAI
77+
// modelID := "gpt-3.5-turbo"
78+
apiURL := "https://api.deepseek.com/chat/completions"
79+
modelID := "deepseek-chat" // DeepSeek V3
80+
81+
// --- 初始化对话历史 ---
82+
// 开发者消息设定角色和目标
83+
conversationHistory := []ChatMessage{
84+
{
85+
Role: "system",
86+
Content: "You are a Go language assistant. Answer questions concisely about Go features.",
87+
},
88+
}
89+
90+
fmt.Printf("Starting a %d-turn conversation with model %s via %s...\n", maxTurns, modelID, apiURL)
91+
fmt.Println("--------------------------------------------------")
92+
93+
// --- 创建 HTTP 客户端 ---
94+
client := &http.Client{Timeout: 60 * time.Second}
95+
96+
// --- 自动进行多轮对话 ---
97+
for turn := 1; turn <= maxTurns; turn++ {
98+
fmt.Printf("--- Turn %d ---\n", turn)
99+
100+
// 1. 模拟生成用户本轮输入 (基于上一轮的简单逻辑)
101+
var userPrompt string
102+
if turn == 1 {
103+
userPrompt = "What are Go channels used for?"
104+
} else {
105+
// 简单追问，实际应用会更复杂
106+
// 获取上一轮助手的回答来构造问题
107+
lastAssistantMessage := ""
108+
if len(conversationHistory) > 0 && conversationHistory[len(conversationHistory)-1].Role == "assistant" {
109+
lastAssistantMessage = conversationHistory[len(conversationHistory)-1].Content
110+
}
111+
if strings.Contains(lastAssistantMessage, "communication") {
112+
userPrompt = "Can you give a simple code example of channel communication?"
113+
} else if strings.Contains(lastAssistantMessage, "synchronization") {
114+
userPrompt = "How does channel synchronization compare to using sync.Mutex?"
115+
} else {
116+
// 如果上轮没捕捉到关键词，就问个相关问题
117+
userPrompt = "What about goroutines? How do they relate to channels?"
118+
}
119+
}
120+
fmt.Printf("User: %s\n", userPrompt)
121+
122+
// 2. 将用户消息添加到历史记录 (发送前)
123+
userMessage := ChatMessage{Role: "user", Content: userPrompt}
124+
conversationHistory = append(conversationHistory, userMessage)
125+
126+
// 3. 准备 API 请求体 (包含完整历史)
127+
requestPayload := ChatCompletionRequest{
128+
Model: modelID,
129+
Messages: conversationHistory, // **关键: 传递了更新后的完整历史**
130+
Temperature: 0.6,
131+
MaxTokens: 100, // **限制每轮回复长度**
132+
Stream: false,
133+
}
134+
requestBodyBytes, err := json.Marshal(requestPayload)
135+
if err != nil {
136+
fmt.Printf("[Error] Marshalling request for turn %d: %v\n", turn, err)
137+
// 出错时移除刚添加的用户消息，避免影响下一轮（如果继续）
138+
conversationHistory = conversationHistory[:len(conversationHistory)-1]
139+
continue // 或 break
140+
}
141+
142+
// 4. 创建并发送 HTTP 请求
143+
req, err := http.NewRequest("POST", apiURL, bytes.NewBuffer(requestBodyBytes))
144+
if err != nil {
145+
fmt.Printf("[Error] Creating request for turn %d: %v\n", turn, err)
146+
conversationHistory = conversationHistory[:len(conversationHistory)-1]
147+
continue
148+
}
149+
req.Header.Set("Content-Type", "application/json")
150+
// 仅在 apiKey 存在时添加 Authorization 头
151+
if apiKey != "" {
152+
req.Header.Set("Authorization", "Bearer "+apiKey)
153+
}
154+
155+
resp, err := client.Do(req)
156+
if err != nil {
157+
fmt.Printf("[Error] Sending request for turn %d: %v\n", turn, err)
158+
conversationHistory = conversationHistory[:len(conversationHistory)-1]
159+
continue
160+
}
161+
162+
// 5. 处理响应
163+
func() { // 使用匿名函数方便 defer resp.Body.Close()
164+
defer resp.Body.Close()
165+
responseBodyBytes, readErr := io.ReadAll(resp.Body)
166+
if readErr != nil {
167+
fmt.Printf("[Error] Reading response body for turn %d: %v\n", turn, readErr)
168+
conversationHistory = conversationHistory[:len(conversationHistory)-1]
169+
return // 从匿名函数返回
170+
}
171+
172+
if resp.StatusCode != http.StatusOK {
173+
fmt.Printf("[Error] Non-OK status code for turn %d: %d\nResponse: %s\n", turn, resp.StatusCode, string(responseBodyBytes))
174+
conversationHistory = conversationHistory[:len(conversationHistory)-1]
175+
return
176+
}
177+
178+
var chatResponse ChatCompletionResponse
179+
unmarshalErr := json.Unmarshal(responseBodyBytes, &chatResponse)
180+
if unmarshalErr != nil {
181+
fmt.Printf("[Error] Unmarshalling response for turn %d: %v\n", turn, unmarshalErr)
182+
fmt.Printf("Raw Response: %s\n", string(responseBodyBytes))
183+
conversationHistory = conversationHistory[:len(conversationHistory)-1]
184+
return
185+
}
186+
187+
// 6. 提取、打印并存储助手响应
188+
if len(chatResponse.Choices) > 0 {
189+
choice := chatResponse.Choices[0]
190+
assistantContent := ""
191+
if choice.Message.Content != nil { // 检查 content 是否为 null
192+
assistantContent = *choice.Message.Content
193+
}
194+
195+
fmt.Printf("Assistant: %s\n", assistantContent)
196+
fmt.Printf("(Finish Reason: %s, Tokens: %d prompt + %d completion = %d total)\n",
197+
choice.FinishReason,
198+
chatResponse.Usage.PromptTokens,
199+
chatResponse.Usage.CompletionTokens,
200+
chatResponse.Usage.TotalTokens)
201+
202+
// 将有效的助手响应添加到历史记录 (为下一轮准备)
203+
if assistantContent != "" || choice.FinishReason == "tool_calls" { // 即使无文本但有工具调用也应记录
204+
assistantMessage := ChatMessage{Role: "assistant", Content: assistantContent}
205+
// 如果是工具调用，还需要处理 tool_calls 字段，这里简化处理
206+
conversationHistory = append(conversationHistory, assistantMessage)
207+
} else {
208+
fmt.Printf("[Warning] Assistant response content was empty for turn %d.\n", turn)
209+
// 如果响应无效，可以选择不添加到历史，或添加一个标记
210+
// 这里同样移除用户消息，避免空响应影响后续
211+
conversationHistory = conversationHistory[:len(conversationHistory)-1]
212+
}
213+
} else {
214+
fmt.Printf("[Error] No choices received in response for turn %d.\n", turn)
215+
conversationHistory = conversationHistory[:len(conversationHistory)-1]
216+
}
217+
}() // 立即执行匿名函数
218+
219+
fmt.Println("--------------------------------------------------")
220+
// 可以加个短暂休眠，模拟思考时间或避免触发速率限制
221+
// time.Sleep(1 * time.Second)
222+
}
223+
224+
fmt.Println("Conversation finished after 5 turns.")
225+
}

0 commit comments

Comments
 (0)