v0.2 管线扩展 + 多供应商 + 测试

.dockerignore

@@ -0,0 +1,6 @@
+.git
+.github
+docs
+*.md
+tmp/
+data/

.github/workflows/release.yml

@@ -28,5 +28,6 @@ jobs:
       - name: Create release
         uses: softprops/action-gh-release@v2
         with:
+          name: "Strait ${{ github.ref_name }}"
           files: build/*
           generate_release_notes: true

.gitignore

@@ -19,6 +19,7 @@
 # 环境变量
 .env
 .env.local
+k8s/secret.yaml
 
 # 依赖缓存
 vendor/

Dockerfile

@@ -0,0 +1,17 @@
+# 构建阶段
+FROM golang:1.26-alpine AS builder
+WORKDIR /app
+COPY go.mod go.sum ./
+RUN go mod download
+COPY . .
+RUN CGO_ENABLED=0 go build -o strait ./cmd/strait/
+
+# 运行阶段
+FROM alpine:3.20
+RUN apk add --no-cache ca-certificates
+WORKDIR /app
+COPY --from=builder /app/strait .
+COPY --from=builder /app/configs ./configs
+
+EXPOSE 8080
+CMD ["./strait"]

api/plugin.go

@@ -29,6 +29,21 @@ type ChatAdapter interface {
 	SendChatStream(ctx context.Context, payload *ChatRequest, route *RouteDecision) (<-chan *StreamChunk, error)
 }
 
+type Guard interface {
+	Plugin
+	Guard(pctx *PipelineContext) error
+}
+
+type PreProcessor interface {
+	Plugin
+	PreProcess(pctx *PipelineContext) error
+}
+
+type PostProcessor interface {
+	Plugin
+	PostProcess(pctx *PipelineContext) error
+}
+
 // Constructor 插件构造函数
 type Constructor func() Plugin
 

api/types.go

@@ -89,16 +89,28 @@ type ToolCallFunction struct {
 
 // RouteDecision 路由决策
 type RouteDecision struct {
-	Protocol string `json:"protocol"` // 调用协议：openAI / anthropic / deepseek / ollama
-	BaseURL  string `json:"base_url"` // 调用地址
-	APIKey   string `json:"api_key"`  // 认证密钥
+	Protocol string `json:"protocol"`        // 调用协议：openAI / anthropic / deepseek / ollama
+	BaseURL  string `json:"base_url"`        // 调用地址
+	APIKey   string `json:"api_key"`         // 认证密钥
+	Model    string `json:"model,omitempty"` // 目标模型名称
 }
 
 // Subject 认证后的调用方信息
 type Subject struct {
 	ID string `json:"id"` // 调用方唯一标识
 }
 
+// ─ 管线 ─
+
+// PipelineContext 管线上下文，贯穿 guard → preprocess → route → adapter → postprocess 全流程
+type PipelineContext struct {
+	Context  context.Context // 请求上下文
+	Request  any             // 请求体：*ChatRequest / *EmbeddingRequest / ...
+	Response any             // 响应体：*ChatResponse / *EmbeddingResponse / ...
+	Route    *RouteDecision  // 路由决策
+	Meta     map[string]any  // 插件间传递的元数据
+}
+
 // ─ Context ─
 
 type ctxKey string

cmd/strait/main.go

@@ -1,31 +1,59 @@
-// Strait — AI 代理网关入口。
+// main Strait — AI 代理网关入口。
 package main
 
 import (
 	"context"
-	"log"
+	"errors"
+	"fmt"
+	"log/slog"
 	"net/http"
+	"os"
+	"os/signal"
+	"syscall"
+	"time"
 
+	"strait/internal/app"
 	"strait/internal/config"
-
 	"strait/internal/hotreload"
-
-	"strait/internal/app"
+	"strait/internal/metrics"
 	"strait/internal/plugin"
 	"strait/internal/router"
 	_ "strait/plugins/adapter-ollama"
 	_ "strait/plugins/adapter-openai"
 	_ "strait/plugins/auth-static-token"
+	_ "strait/plugins/prompt-injector"
 )
 
+const banner = `
+███████╗████████╗██████╗  █████╗ ██╗████████╗
+██╔════╝╚══██╔══╝██╔══██╗██╔══██╗██║╚══██╔══╝
+███████╗   ██║   ██████╔╝███████║██║   ██║
+╚════██║   ██║   ██╔══██╗██╔══██║██║   ██║
+███████║   ██║   ██║  ██║██║  ██║██║   ██║
+╚══════╝   ╚═╝   ╚═╝  ╚═╝╚═╝  ╚═╝╚═╝   ╚═╝  v0.2
+`
+
 func main() {
+	slog.SetDefault(slog.New(slog.NewTextHandler(os.Stdout, &slog.HandlerOptions{
+		Level: slog.LevelInfo,
+	})))
+
 	loader := plugin.NewLoader(config.PluginsPath)
 	m, err := loader.Build()
 	if err != nil {
-		log.Fatal(err)
+		slog.Error("startup failed", "error", err)
+		os.Exit(1)
+	}
+
+	if os.Getenv("STRAIT_BANNER") != "false" {
+		fmt.Print(banner)
+		fmt.Println(" Plugins:")
+		fmt.Print(m.Summary())
+		fmt.Println()
 	}
 
 	scheduler := plugin.NewScheduler(m)
+	met := metrics.NewMetrics()
 
 	reload := func(filename string) error {
 		if filename == "plugins.yaml" {
@@ -49,7 +77,30 @@ func main() {
 		_ = watcher.Start(context.Background())
 	}()
 
-	server := app.NewServer(scheduler)
-	log.Println("strait listening on :8080")
-	log.Fatal(http.ListenAndServe(":8080", server.Handler()))
+	// 启动 HTTP 服务
+	server := app.NewServer(scheduler, met)
+	srv := &http.Server{
+		Addr:    ":8080",
+		Handler: server.Handler(),
+	}
+
+	ctx, stop := signal.NotifyContext(context.Background(), os.Interrupt, syscall.SIGTERM)
+	defer stop()
+
+	go func() {
+		slog.Info("strait listening on :8080")
+		if err := srv.ListenAndServe(); err != nil && !errors.Is(err, http.ErrServerClosed) {
+			slog.Error("server crashed", "error", err)
+			os.Exit(1)
+		}
+	}()
+
+	<-ctx.Done()
+	slog.Info("shutting down...")
+	shutdownCtx, cancel := context.WithTimeout(context.Background(), 10*time.Second)
+	defer cancel()
+	if err := srv.Shutdown(shutdownCtx); err != nil {
+		slog.Error("shutdown failed", "error", err)
+	}
+	slog.Info("strait stopped")
 }

configs/plugins.yaml

@@ -9,4 +9,8 @@ plugins:
       token: sk-admin-init
       subject: admin
   - id: adapter-ollama
-    type: adapter
+    type: adapter
+  - id: prompt-injector
+    type: preprocessor
+    config:
+      system_prompt: "You are a helpful assistant."

configs/routes.yaml

@@ -2,9 +2,19 @@ routes:
   - id: deepseek-chat
     match:
       model: deepseek-chat
+    strategy: weight
     target:
       provider: deepseek-main
-      model: deepseek-chat
+      model: deepseek-reasoner
+    targets:
+      - provider: deepseek-main
+        model: deepseek-chat
+        priority: 1
+        weight: 3
+      - provider: ollama-local
+        model: qwen2.5:0.5b
+        priority: 1
+        weight: 1
 
   - id: deepseek-reasoner
     match:
@@ -16,6 +26,9 @@ routes:
   - id: ollama-qwen
     match:
       model: qwen2.5:0.5b
-    target:
-      provider: ollama-local
-      model: qwen2.5:0.5b
+    strategy: priority
+    targets:
+      - provider: ollama-local
+        model: qwen2.5:0.5b
+        priority: 1
+        weight: 1

docker-compose.yml

@@ -0,0 +1,8 @@
+services:
+  strait:
+    build: .
+    ports:
+      - "8080:8080"
+    volumes:
+      - ./configs:/app/configs
+    restart: unless-stopped

docs/deployment.md

@@ -1,20 +1,97 @@
 # 部署指南
 
-## MySQL
+## Docker
+
+### 构建镜像
 
 ```bash
-docker-compose up -d
+docker build -t strait .
+```
+
+### 运行
+
+```bash
+docker run -d \
+  -p 8080:8080 \
+  -e DEEPSEEK_API_KEY=your-key \
+  -v $(pwd)/configs:/app/configs \
+  strait
 ```
 
-## Docker Compose
+### Docker Compose
 
 ```bash
+# 设置环境变量
+export DEEPSEEK_API_KEY=your-key
+
+# 启动
 docker-compose up -d
 ```
 
+`docker-compose.yml` 会自动挂载 `configs/` 目录并注入环境变量。
+
+---
+
 ## Kubernetes
 
+### 1. 创建 Secret（存放 API Key）
+
+参考 `k8s/secret.example.yaml`：
+
+```bash
+kubectl create secret generic strait-secret \
+  --from-literal=DEEPSEEK_API_KEY=your-key
+```
+
+### 2. 创建 ConfigMap（挂载配置文件）
+
+```bash
+kubectl create configmap strait-config \
+  --from-file=configs/
+```
+
+### 3. 部署
+
 ```bash
-helm repo add strait https://yourorg.github.io/helm-charts
-helm install strait strait/strait
+kubectl apply -f k8s/service.yaml
+kubectl apply -f k8s/deployment.yaml
 ```
+
+### 4. 验证
+
+```bash
+# 检查 Pod 状态
+kubectl get pods -l app=strait
+
+# 健康检查
+kubectl port-forward svc/strait 8080:8080
+curl http://localhost:8080/health
+```
+
+### 探针说明
+
+| 探针 | 路径 | 说明 |
+|------|------|------|
+| liveness | `/health` | 存活检测，失败则重启容器 |
+| readiness | `/ready` | 就绪检测，失败则从 Service 摘除 |
+
+---
+
+## 监控
+
+### Prometheus Metrics
+
+服务暴露 `/metrics` 端点，返回 Prometheus 格式的请求计数：
+
+```bash
+curl http://localhost:8080/metrics
+```
+
+---
+
+## 配置热重载
+
+修改 `configs/` 目录下的 YAML 文件后自动重载，无需重启：
+
+- `providers.yaml` / `routes.yaml` → 路由重载
+- `plugins.yaml` → 插件重载