M4a：神识全量网关能力——OpenAI 兼容适配器（元炁共用）、候选路由+凭证/绑定级冷却+首字超时切道+粘性会话、用量计量与 /me/usage，故障注入测试四场景

shenshi/internal/transform/openai.go

@@ -0,0 +1,138 @@
+// OpenAI 兼容出站适配器：IR → OpenAI chat/completions（元炁亦走此协议，type=local 复用）。
+package transform
+
+import (
+	"bufio"
+	"bytes"
+	"context"
+	"encoding/json"
+	"fmt"
+	"net/http"
+	"strings"
+	"time"
+
+	"hunjing/shenshi/internal/ir"
+)
+
+// UpstreamError 上游错误：携带 HTTP 状态，供路由层分类冷却（凭证级/绑定级）。
+type UpstreamError struct {
+	StatusCode int
+	Message    string
+}
+
+func (e *UpstreamError) Error() string {
+	return fmt.Sprintf("上游 %d: %s", e.StatusCode, e.Message)
+}
+
+// OpenAICompatAdaptor OpenAI 兼容端点适配器（云厂商兼容面 + 元炁本地端点）。
+type OpenAICompatAdaptor struct {
+	EndpointName string // 端点标识（usage 日志与冷却键）
+	BaseURL      string
+	APIKey       string
+	// UpstreamModel 该端点承载的上游模型名（逻辑模型 → 端点绑定的组成部分）
+	UpstreamModel string
+	Client        *http.Client
+}
+
+// Name 实现 Adaptor。
+func (a *OpenAICompatAdaptor) Name() string { return a.EndpointName }
+
+type oaMessage struct {
+	Role    string `json:"role"`
+	Content string `json:"content"`
+}
+
+type oaRequest struct {
+	Model       string      `json:"model"`
+	Messages    []oaMessage `json:"messages"`
+	Stream      bool        `json:"stream"`
+	Temperature *float64    `json:"temperature,omitempty"`
+	MaxTokens   *int        `json:"max_tokens,omitempty"`
+}
+
+type oaChunk struct {
+	Choices []struct {
+		Delta struct {
+			Content string `json:"content"`
+		} `json:"delta"`
+		FinishReason *string `json:"finish_reason"`
+	} `json:"choices"`
+	Usage *struct {
+		PromptTokens     int `json:"prompt_tokens"`
+		CompletionTokens int `json:"completion_tokens"`
+	} `json:"usage"`
+}
+
+// Stream 实现 Adaptor：SSE 逐帧转 IR Delta。首字节超时由调用方 context 控制。
+func (a *OpenAICompatAdaptor) Stream(ctx context.Context, req *ir.ChatRequest, out chan<- ir.Delta) error {
+	body := oaRequest{
+		Model:    a.UpstreamModel,
+		Messages: make([]oaMessage, 0, len(req.Messages)),
+		Stream:   true,
+	}
+	for _, m := range req.Messages {
+		body.Messages = append(body.Messages, oaMessage{Role: m.Role, Content: m.Content})
+	}
+	body.Temperature = req.Temperature
+	body.MaxTokens = req.MaxTokens
+
+	payload, err := json.Marshal(body)
+	if err != nil {
+		return err
+	}
+	httpReq, err := http.NewRequestWithContext(ctx, http.MethodPost, strings.TrimRight(a.BaseURL, "/")+"/v1/chat/completions", bytes.NewReader(payload))
+	if err != nil {
+		return err
+	}
+	httpReq.Header.Set("content-type", "application/json")
+	if a.APIKey != "" {
+		httpReq.Header.Set("authorization", "Bearer "+a.APIKey)
+	}
+
+	client := a.Client
+	if client == nil {
+		client = &http.Client{Timeout: 120 * time.Second}
+	}
+	resp, err := client.Do(httpReq)
+	if err != nil {
+		return &UpstreamError{StatusCode: 0, Message: err.Error()} // 0 = 网络/超时（绑定级冷却）
+	}
+	defer func() { _ = resp.Body.Close() }()
+	if resp.StatusCode >= 400 {
+		buf := new(bytes.Buffer)
+		_, _ = buf.ReadFrom(resp.Body)
+		return &UpstreamError{StatusCode: resp.StatusCode, Message: buf.String()}
+	}
+
+	scanner := bufio.NewScanner(resp.Body)
+	scanner.Buffer(make([]byte, 0, 64*1024), 1024*1024)
+	for scanner.Scan() {
+		line := scanner.Text()
+		if !strings.HasPrefix(line, "data:") {
+			continue
+		}
+		data := strings.TrimSpace(strings.TrimPrefix(line, "data:"))
+		if data == "" || data == "[DONE]" {
+			continue
+		}
+		var chunk oaChunk
+		if err := json.Unmarshal([]byte(data), &chunk); err != nil {
+			continue
+		}
+		if len(chunk.Choices) > 0 {
+			c := chunk.Choices[0]
+			if c.Delta.Content != "" {
+				out <- ir.Delta{Delta: c.Delta.Content}
+			}
+			if c.FinishReason != nil {
+				finish := *c.FinishReason
+				usagePrompt, usageCompletion := 0, 0
+				if chunk.Usage != nil {
+					usagePrompt, usageCompletion = chunk.Usage.PromptTokens, chunk.Usage.CompletionTokens
+				}
+				out <- ir.Delta{FinishReason: finish, UsagePromptTokens: usagePrompt, UsageCompletionTokens: usageCompletion}
+			}
+		}
+	}
+	return scanner.Err()
+}