feat(openai): 支持messages模型映射与instructions模板注入

backend/internal/config/config.go

@@ -318,6 +318,12 @@ type GatewayConfig struct {
 	// ForceCodexCLI: 强制将 OpenAI `/v1/responses` 请求按 Codex CLI 处理。
 	// 用于网关未透传/改写 User-Agent 时的兼容兜底（默认关闭，避免影响其他客户端）。
 	ForceCodexCLI bool `mapstructure:"force_codex_cli"`
+	// ForcedCodexInstructionsTemplateFile: 服务端强制附加到 Codex 顶层 instructions 的模板文件路径。
+	// 模板渲染后会直接覆盖最终 instructions；若需要保留客户端 system 转换结果，请在模板中显式引用 {{ .ExistingInstructions }}。
+	ForcedCodexInstructionsTemplateFile string `mapstructure:"forced_codex_instructions_template_file"`
+	// ForcedCodexInstructionsTemplate: 启动时从模板文件读取并缓存的模板内容。
+	// 该字段不直接参与配置反序列化，仅用于请求热路径避免重复读盘。
+	ForcedCodexInstructionsTemplate string `mapstructure:"-"`
 	// OpenAIPassthroughAllowTimeoutHeaders: OpenAI 透传模式是否放行客户端超时头
 	// 关闭（默认）可避免 x-stainless-timeout 等头导致上游提前断流。
 	OpenAIPassthroughAllowTimeoutHeaders bool `mapstructure:"openai_passthrough_allow_timeout_headers"`
@@ -983,6 +989,14 @@ func load(allowMissingJWTSecret bool) (*Config, error) {
 	cfg.Log.Environment = strings.TrimSpace(cfg.Log.Environment)
 	cfg.Log.StacktraceLevel = strings.ToLower(strings.TrimSpace(cfg.Log.StacktraceLevel))
 	cfg.Log.Output.FilePath = strings.TrimSpace(cfg.Log.Output.FilePath)
+	cfg.Gateway.ForcedCodexInstructionsTemplateFile = strings.TrimSpace(cfg.Gateway.ForcedCodexInstructionsTemplateFile)
+	if cfg.Gateway.ForcedCodexInstructionsTemplateFile != "" {
+		content, err := os.ReadFile(cfg.Gateway.ForcedCodexInstructionsTemplateFile)
+		if err != nil {
+			return nil, fmt.Errorf("read forced codex instructions template %q: %w", cfg.Gateway.ForcedCodexInstructionsTemplateFile, err)
+		}
+		cfg.Gateway.ForcedCodexInstructionsTemplate = string(content)
+	}
 
 	// 兼容旧键 gateway.openai_ws.sticky_previous_response_ttl_seconds。
 	// 新键未配置（<=0）时回退旧键；新键优先。

backend/internal/config/config_test.go

 package config
 
 import (
+	"os"
+	"path/filepath"
 	"strings"
 	"testing"
 	"time"
@@ -223,6 +225,23 @@ func TestLoadSchedulingConfigFromEnv(t *testing.T) {
 	}
 }
 
+func TestLoadForcedCodexInstructionsTemplate(t *testing.T) {
+	resetViperWithJWTSecret(t)
+
+	tempDir := t.TempDir()
+	templatePath := filepath.Join(tempDir, "codex-instructions.md.tmpl")
+	configPath := filepath.Join(tempDir, "config.yaml")
+
+	require.NoError(t, os.WriteFile(templatePath, []byte("server-prefix\n\n{{ .ExistingInstructions }}"), 0o644))
+	require.NoError(t, os.WriteFile(configPath, []byte("gateway:\n  forced_codex_instructions_template_file: \""+templatePath+"\"\n"), 0o644))
+	t.Setenv("DATA_DIR", tempDir)
+
+	cfg, err := Load()
+	require.NoError(t, err)
+	require.Equal(t, templatePath, cfg.Gateway.ForcedCodexInstructionsTemplateFile)
+	require.Equal(t, "server-prefix\n\n{{ .ExistingInstructions }}", cfg.Gateway.ForcedCodexInstructionsTemplate)
+}
+
 func TestLoadDefaultSecurityToggles(t *testing.T) {
 	resetViperWithJWTSecret(t)
 

backend/internal/handler/openai_gateway_handler.go

@@ -47,6 +47,13 @@ func resolveOpenAIForwardDefaultMappedModel(apiKey *service.APIKey, fallbackMode
 	return strings.TrimSpace(apiKey.Group.DefaultMappedModel)
 }
 
+func resolveOpenAIMessagesDispatchMappedModel(apiKey *service.APIKey, requestedModel string) string {
+	if apiKey == nil || apiKey.Group == nil {
+		return ""
+	}
+	return strings.TrimSpace(apiKey.Group.ResolveMessagesDispatchModel(requestedModel))
+}
+
 // NewOpenAIGatewayHandler creates a new OpenAIGatewayHandler
 func NewOpenAIGatewayHandler(
 	gatewayService *service.OpenAIGatewayService,
@@ -551,6 +558,7 @@ func (h *OpenAIGatewayHandler) Messages(c *gin.Context) {
 	}
 	reqModel := modelResult.String()
 	routingModel := service.NormalizeOpenAICompatRequestedModel(reqModel)
+	preferredMappedModel := resolveOpenAIMessagesDispatchMappedModel(apiKey, reqModel)
 	reqStream := gjson.GetBytes(body, "stream").Bool()
 
 	reqLog = reqLog.With(zap.String("model", reqModel), zap.Bool("stream", reqStream))
@@ -609,17 +617,20 @@ func (h *OpenAIGatewayHandler) Messages(c *gin.Context) {
 	failedAccountIDs := make(map[int64]struct{})
 	sameAccountRetryCount := make(map[int64]int)
 	var lastFailoverErr *service.UpstreamFailoverError
+	effectiveMappedModel := preferredMappedModel
 
 	for {
-		// 清除上一次迭代的降级模型标记，避免残留影响本次迭代
-		c.Set("openai_messages_fallback_model", "")
+		currentRoutingModel := routingModel
+		if effectiveMappedModel != "" {
+			currentRoutingModel = effectiveMappedModel
+		}
 		reqLog.Debug("openai_messages.account_selecting", zap.Int("excluded_account_count", len(failedAccountIDs)))
 		selection, scheduleDecision, err := h.gatewayService.SelectAccountWithScheduler(
 			c.Request.Context(),
 			apiKey.GroupID,
 			"", // no previous_response_id
 			sessionHash,
-			routingModel,
+			currentRoutingModel,
 			failedAccountIDs,
 			service.OpenAIUpstreamTransportAny,
 		)
@@ -628,29 +639,7 @@ func (h *OpenAIGatewayHandler) Messages(c *gin.Context) {
 				zap.Error(err),
 				zap.Int("excluded_account_count", len(failedAccountIDs)),
 			)
-			// 首次调度失败 + 有默认映射模型 → 用默认模型重试
 			if len(failedAccountIDs) == 0 {
-				defaultModel := ""
-				if apiKey.Group != nil {
-					defaultModel = apiKey.Group.DefaultMappedModel
-				}
-				if defaultModel != "" && defaultModel != routingModel {
-					reqLog.Info("openai_messages.fallback_to_default_model",
-						zap.String("default_mapped_model", defaultModel),
-					)
-					selection, scheduleDecision, err = h.gatewayService.SelectAccountWithScheduler(
-						c.Request.Context(),
-						apiKey.GroupID,
-						"",
-						sessionHash,
-						defaultModel,
-						failedAccountIDs,
-						service.OpenAIUpstreamTransportAny,
-					)
-					if err == nil && selection != nil {
-						c.Set("openai_messages_fallback_model", defaultModel)
-					}
-				}
 				if err != nil {
 					h.anthropicStreamingAwareError(c, http.StatusServiceUnavailable, "api_error", "Service temporarily unavailable", streamStarted)
 					return
@@ -682,9 +671,7 @@ func (h *OpenAIGatewayHandler) Messages(c *gin.Context) {
 		service.SetOpsLatencyMs(c, service.OpsRoutingLatencyMsKey, time.Since(routingStart).Milliseconds())
 		forwardStart := time.Now()
 
-		// Forward 层需要始终拿到 group 默认映射模型，这样未命中账号级映射的
-		// Claude 兼容模型才不会在后续 Codex 规范化中意外退化到 gpt-5.1。
-		defaultMappedModel := resolveOpenAIForwardDefaultMappedModel(apiKey, c.GetString("openai_messages_fallback_model"))
+		defaultMappedModel := strings.TrimSpace(effectiveMappedModel)
 		// 应用渠道模型映射到请求体
 		forwardBody := body
 		if channelMappingMsg.Mapped {

backend/internal/handler/openai_gateway_handler_test.go

@@ -360,7 +360,7 @@ func TestResolveOpenAIForwardDefaultMappedModel(t *testing.T) {
 		require.Equal(t, "gpt-5.2", resolveOpenAIForwardDefaultMappedModel(apiKey, " gpt-5.2 "))
 	})
 
-	t.Run("uses_group_default_on_normal_path", func(t *testing.T) {
+	t.Run("uses_group_default_when_explicit_fallback_absent", func(t *testing.T) {
 		apiKey := &service.APIKey{
 			Group: &service.Group{DefaultMappedModel: "gpt-5.4"},
 		}
@@ -376,6 +376,45 @@ func TestResolveOpenAIForwardDefaultMappedModel(t *testing.T) {
 	})
 }
 
+func TestResolveOpenAIMessagesDispatchMappedModel(t *testing.T) {
+	t.Run("exact_claude_model_override_wins", func(t *testing.T) {
+		apiKey := &service.APIKey{
+			Group: &service.Group{
+				MessagesDispatchModelConfig: service.OpenAIMessagesDispatchModelConfig{
+					SonnetMappedModel: "gpt-5.2",
+					ExactModelMappings: map[string]string{
+						"claude-sonnet-4-5-20250929": "gpt-5.4-mini-high",
+					},
+				},
+			},
+		}
+		require.Equal(t, "gpt-5.4-mini", resolveOpenAIMessagesDispatchMappedModel(apiKey, "claude-sonnet-4-5-20250929"))
+	})
+
+	t.Run("uses_family_default_when_no_override", func(t *testing.T) {
+		apiKey := &service.APIKey{Group: &service.Group{}}
+		require.Equal(t, "gpt-5.4", resolveOpenAIMessagesDispatchMappedModel(apiKey, "claude-opus-4-6"))
+		require.Equal(t, "gpt-5.3-codex", resolveOpenAIMessagesDispatchMappedModel(apiKey, "claude-sonnet-4-5-20250929"))
+		require.Equal(t, "gpt-5.4-mini", resolveOpenAIMessagesDispatchMappedModel(apiKey, "claude-haiku-4-5-20251001"))
+	})
+
+	t.Run("returns_empty_for_non_claude_or_missing_group", func(t *testing.T) {
+		require.Empty(t, resolveOpenAIMessagesDispatchMappedModel(nil, "claude-sonnet-4-5-20250929"))
+		require.Empty(t, resolveOpenAIMessagesDispatchMappedModel(&service.APIKey{}, "claude-sonnet-4-5-20250929"))
+		require.Empty(t, resolveOpenAIMessagesDispatchMappedModel(&service.APIKey{Group: &service.Group{}}, "gpt-5.4"))
+	})
+
+	t.Run("does_not_fall_back_to_group_default_mapped_model", func(t *testing.T) {
+		apiKey := &service.APIKey{
+			Group: &service.Group{
+				DefaultMappedModel: "gpt-5.4",
+			},
+		}
+		require.Empty(t, resolveOpenAIMessagesDispatchMappedModel(apiKey, "gpt-5.4"))
+		require.Equal(t, "gpt-5.3-codex", resolveOpenAIMessagesDispatchMappedModel(apiKey, "claude-sonnet-4-5-20250929"))
+	})
+}
+
 func TestOpenAIResponses_MissingDependencies_ReturnsServiceUnavailable(t *testing.T) {
 	gin.SetMode(gin.TestMode)
 

backend/internal/pkg/apicompat/types.go

@@ -28,7 +28,7 @@ type AnthropicRequest struct {
 
 // AnthropicOutputConfig controls output generation parameters.
 type AnthropicOutputConfig struct {
-	Effort string `json:"effort,omitempty"` // "low" | "medium" | "high"
+	Effort string `json:"effort,omitempty"` // "low" | "medium" | "high" | "max"
 }
 
 // AnthropicThinking configures extended thinking in the Anthropic API.
@@ -167,7 +167,7 @@ type ResponsesRequest struct {
 
 // ResponsesReasoning configures reasoning effort in the Responses API.
 type ResponsesReasoning struct {
-	Effort  string `json:"effort"`            // "low" | "medium" | "high"
+	Effort  string `json:"effort"`            // "low" | "medium" | "high" | "xhigh"
 	Summary string `json:"summary,omitempty"` // "auto" | "concise" | "detailed"
 }
 
@@ -345,7 +345,7 @@ type ChatCompletionsRequest struct {
 	StreamOptions       *ChatStreamOptions `json:"stream_options,omitempty"`
 	Tools               []ChatTool         `json:"tools,omitempty"`
 	ToolChoice          json.RawMessage    `json:"tool_choice,omitempty"`
-	ReasoningEffort     string             `json:"reasoning_effort,omitempty"` // "low" | "medium" | "high"
+	ReasoningEffort     string             `json:"reasoning_effort,omitempty"` // "low" | "medium" | "high" | "xhigh"
 	ServiceTier         string             `json:"service_tier,omitempty"`
 	Stop                json.RawMessage    `json:"stop,omitempty"` // string or []string
 

backend/internal/service/openai_codex_instructions_template.go

+package service
+
+import (
+	"bytes"
+	"fmt"
+	"strings"
+	"text/template"
+)
+
+type forcedCodexInstructionsTemplateData struct {
+	ExistingInstructions string
+	OriginalModel        string
+	NormalizedModel      string
+	BillingModel         string
+	UpstreamModel        string
+}
+
+func applyForcedCodexInstructionsTemplate(
+	reqBody map[string]any,
+	templateText string,
+	data forcedCodexInstructionsTemplateData,
+) (bool, error) {
+	rendered, err := renderForcedCodexInstructionsTemplate(templateText, data)
+	if err != nil {
+		return false, err
+	}
+	if rendered == "" {
+		return false, nil
+	}
+
+	existing, _ := reqBody["instructions"].(string)
+	if strings.TrimSpace(existing) == rendered {
+		return false, nil
+	}
+
+	reqBody["instructions"] = rendered
+	return true, nil
+}
+
+func renderForcedCodexInstructionsTemplate(
+	templateText string,
+	data forcedCodexInstructionsTemplateData,
+) (string, error) {
+	tmpl, err := template.New("forced_codex_instructions").Option("missingkey=zero").Parse(templateText)
+	if err != nil {
+		return "", fmt.Errorf("parse forced codex instructions template: %w", err)
+	}
+
+	var buf bytes.Buffer
+	if err := tmpl.Execute(&buf, data); err != nil {
+		return "", fmt.Errorf("render forced codex instructions template: %w", err)
+	}
+
+	return strings.TrimSpace(buf.String()), nil
+}

backend/internal/service/openai_compat_model_test.go

@@ -6,9 +6,12 @@ import (
 	"io"
 	"net/http"
 	"net/http/httptest"
+	"os"
+	"path/filepath"
 	"strings"
 	"testing"
 
+	"github.com/Wei-Shaw/sub2api/internal/config"
 	"github.com/Wei-Shaw/sub2api/internal/pkg/apicompat"
 	"github.com/gin-gonic/gin"
 	"github.com/stretchr/testify/require"
@@ -127,3 +130,101 @@ func TestForwardAsAnthropic_NormalizesRoutingAndEffortForGpt54XHigh(t *testing.T
 	t.Logf("upstream body: %s", string(upstream.lastBody))
 	t.Logf("response body: %s", rec.Body.String())
 }
+
+func TestForwardAsAnthropic_ForcedCodexInstructionsTemplatePrependsRenderedInstructions(t *testing.T) {
+	t.Parallel()
+	gin.SetMode(gin.TestMode)
+
+	templateDir := t.TempDir()
+	templatePath := filepath.Join(templateDir, "codex-instructions.md.tmpl")
+	require.NoError(t, os.WriteFile(templatePath, []byte("server-prefix\n\n{{ .ExistingInstructions }}"), 0o644))
+
+	rec := httptest.NewRecorder()
+	c, _ := gin.CreateTestContext(rec)
+	body := []byte(`{"model":"gpt-5.4","max_tokens":16,"system":"client-system","messages":[{"role":"user","content":"hello"}],"stream":false}`)
+	c.Request = httptest.NewRequest(http.MethodPost, "/v1/messages", bytes.NewReader(body))
+	c.Request.Header.Set("Content-Type", "application/json")
+
+	upstreamBody := strings.Join([]string{
+		`data: {"type":"response.completed","response":{"id":"resp_1","object":"response","model":"gpt-5.4","status":"completed","output":[{"type":"message","id":"msg_1","role":"assistant","status":"completed","content":[{"type":"output_text","text":"ok"}]}],"usage":{"input_tokens":5,"output_tokens":2,"total_tokens":7}}}`,
+		"",
+		"data: [DONE]",
+		"",
+	}, "\n")
+	upstream := &httpUpstreamRecorder{resp: &http.Response{
+		StatusCode: http.StatusOK,
+		Header:     http.Header{"Content-Type": []string{"text/event-stream"}, "x-request-id": []string{"rid_forced"}},
+		Body:       io.NopCloser(strings.NewReader(upstreamBody)),
+	}}
+
+	svc := &OpenAIGatewayService{
+		cfg: &config.Config{Gateway: config.GatewayConfig{
+			ForcedCodexInstructionsTemplateFile: templatePath,
+			ForcedCodexInstructionsTemplate:     "server-prefix\n\n{{ .ExistingInstructions }}",
+		}},
+		httpUpstream: upstream,
+	}
+	account := &Account{
+		ID:          1,
+		Name:        "openai-oauth",
+		Platform:    PlatformOpenAI,
+		Type:        AccountTypeOAuth,
+		Concurrency: 1,
+		Credentials: map[string]any{
+			"access_token":       "oauth-token",
+			"chatgpt_account_id": "chatgpt-acc",
+		},
+	}
+
+	result, err := svc.ForwardAsAnthropic(context.Background(), c, account, body, "", "gpt-5.1")
+	require.NoError(t, err)
+	require.NotNil(t, result)
+	require.Equal(t, "server-prefix\n\nclient-system", gjson.GetBytes(upstream.lastBody, "instructions").String())
+}
+
+func TestForwardAsAnthropic_ForcedCodexInstructionsTemplateUsesCachedTemplateContent(t *testing.T) {
+	t.Parallel()
+	gin.SetMode(gin.TestMode)
+
+	rec := httptest.NewRecorder()
+	c, _ := gin.CreateTestContext(rec)
+	body := []byte(`{"model":"gpt-5.4","max_tokens":16,"system":"client-system","messages":[{"role":"user","content":"hello"}],"stream":false}`)
+	c.Request = httptest.NewRequest(http.MethodPost, "/v1/messages", bytes.NewReader(body))
+	c.Request.Header.Set("Content-Type", "application/json")
+
+	upstreamBody := strings.Join([]string{
+		`data: {"type":"response.completed","response":{"id":"resp_1","object":"response","model":"gpt-5.4","status":"completed","output":[{"type":"message","id":"msg_1","role":"assistant","status":"completed","content":[{"type":"output_text","text":"ok"}]}],"usage":{"input_tokens":5,"output_tokens":2,"total_tokens":7}}}`,
+		"",
+		"data: [DONE]",
+		"",
+	}, "\n")
+	upstream := &httpUpstreamRecorder{resp: &http.Response{
+		StatusCode: http.StatusOK,
+		Header:     http.Header{"Content-Type": []string{"text/event-stream"}, "x-request-id": []string{"rid_forced_cached"}},
+		Body:       io.NopCloser(strings.NewReader(upstreamBody)),
+	}}
+
+	svc := &OpenAIGatewayService{
+		cfg: &config.Config{Gateway: config.GatewayConfig{
+			ForcedCodexInstructionsTemplateFile: "/path/that/should/not/be/read.tmpl",
+			ForcedCodexInstructionsTemplate:     "cached-prefix\n\n{{ .ExistingInstructions }}",
+		}},
+		httpUpstream: upstream,
+	}
+	account := &Account{
+		ID:          1,
+		Name:        "openai-oauth",
+		Platform:    PlatformOpenAI,
+		Type:        AccountTypeOAuth,
+		Concurrency: 1,
+		Credentials: map[string]any{
+			"access_token":       "oauth-token",
+			"chatgpt_account_id": "chatgpt-acc",
+		},
+	}
+
+	result, err := svc.ForwardAsAnthropic(context.Background(), c, account, body, "", "gpt-5.1")
+	require.NoError(t, err)
+	require.NotNil(t, result)
+	require.Equal(t, "cached-prefix\n\nclient-system", gjson.GetBytes(upstream.lastBody, "instructions").String())
+}

backend/internal/service/openai_gateway_messages.go

@@ -86,6 +86,24 @@ func (s *OpenAIGatewayService) ForwardAsAnthropic(
 			return nil, fmt.Errorf("unmarshal for codex transform: %w", err)
 		}
 		codexResult := applyCodexOAuthTransform(reqBody, false, false)
+		forcedTemplateText := ""
+		if s.cfg != nil {
+			forcedTemplateText = s.cfg.Gateway.ForcedCodexInstructionsTemplate
+		}
+		templateUpstreamModel := upstreamModel
+		if codexResult.NormalizedModel != "" {
+			templateUpstreamModel = codexResult.NormalizedModel
+		}
+		existingInstructions, _ := reqBody["instructions"].(string)
+		if _, err := applyForcedCodexInstructionsTemplate(reqBody, forcedTemplateText, forcedCodexInstructionsTemplateData{
+			ExistingInstructions: strings.TrimSpace(existingInstructions),
+			OriginalModel:        originalModel,
+			NormalizedModel:      normalizedModel,
+			BillingModel:         billingModel,
+			UpstreamModel:        templateUpstreamModel,
+		}); err != nil {
+			return nil, err
+		}
 		if codexResult.NormalizedModel != "" {
 			upstreamModel = codexResult.NormalizedModel
 		}

deploy/codex-instructions.md.tmpl

+You are Codex, based on GPT-5. You are running as a coding agent in the Codex CLI on a user's computer.
+
+{{ if .ExistingInstructions }}
+{{ .ExistingInstructions }}
+{{ end }}

deploy/config.example.yaml

@@ -202,6 +202,32 @@ gateway:
   #
   # 注意：开启后会影响所有客户端的行为（不仅限于 VS Code / Codex CLI），请谨慎开启。
   force_codex_cli: false
+  # Optional: template file used to build the final top-level Codex `instructions`.
+  # 可选：用于构建最终 Codex 顶层 `instructions` 的模板文件路径。
+  #
+  # This is applied on the `/v1/messages -> Responses/Codex` conversion path,
+  # after Claude `system` has already been normalized into Codex `instructions`.
+  # 该模板作用于 `/v1/messages -> Responses/Codex` 转换链路，且发生在 Claude `system`
+  # 已经被归一化为 Codex `instructions` 之后。
+  #
+  # The template can reference:
+  # 模板可引用：
+  # - {{ .ExistingInstructions }} : converted client instructions/system
+  # - {{ .OriginalModel }}        : original requested model
+  # - {{ .NormalizedModel }}      : normalized routing model
+  # - {{ .BillingModel }}         : billing model
+  # - {{ .UpstreamModel }}        : final upstream model
+  #
+  # If you want to preserve client system prompts, keep {{ .ExistingInstructions }}
+  # somewhere in the template. If omitted, the template output fully replaces it.
+  # 如需保留客户端 system 提示词，请在模板中显式包含 {{ .ExistingInstructions }}。
+  # 若省略，则模板输出会完全覆盖它。
+  #
+  # Docker users can mount a host file to /app/data/codex-instructions.md.tmpl
+  # and point this field there.
+  # Docker 用户可将宿主机文件挂载到 /app/data/codex-instructions.md.tmpl，
+  # 然后把本字段指向该路径。
+  forced_codex_instructions_template_file: ""
   # OpenAI 透传模式是否放行客户端超时头（如 x-stainless-timeout）
   # 默认 false：过滤超时头，降低上游提前断流风险。
   openai_passthrough_allow_timeout_headers: false
@@ -347,12 +373,6 @@ gateway:
     # Enable batch load calculation for scheduling
     # 启用调度批量负载计算
     load_batch_enabled: true
-    # Snapshot bucket MGET chunk size
-    # 调度快照分桶读取时的 MGET 分块大小
-    snapshot_mget_chunk_size: 128
-    # Snapshot bucket write chunk size
-    # 调度快照重建写入时的分块大小
-    snapshot_write_chunk_size: 256
     # Slot cleanup interval (duration)
     # 并发槽位清理周期（时间段）
     slot_cleanup_interval: 30s

deploy/docker-compose.yml

@@ -31,6 +31,10 @@ services:
       # Optional: Mount custom config.yaml (uncomment and create the file first)
       # Copy config.example.yaml to config.yaml, modify it, then uncomment:
       # - ./config.yaml:/app/data/config.yaml
+      # Optional: Mount a custom Codex instructions template file, then point
+      # gateway.forced_codex_instructions_template_file at /app/data/codex-instructions.md.tmpl
+      # in config.yaml.
+      # - ./codex-instructions.md.tmpl:/app/data/codex-instructions.md.tmpl:ro
     environment:
       # =======================================================================
       # Auto Setup (REQUIRED for Docker deployment)
@@ -146,7 +150,17 @@ services:
     networks:
       - sub2api-network
     healthcheck:
-      test: ["CMD", "wget", "-q", "-T", "5", "-O", "/dev/null", "http://localhost:8080/health"]
+      test:
+        [
+          "CMD",
+          "wget",
+          "-q",
+          "-T",
+          "5",
+          "-O",
+          "/dev/null",
+          "http://localhost:8080/health",
+        ]
       interval: 30s
       timeout: 10s
       retries: 3
@@ -177,11 +191,17 @@ services:
     networks:
       - sub2api-network
     healthcheck:
-      test: ["CMD-SHELL", "pg_isready -U ${POSTGRES_USER:-sub2api} -d ${POSTGRES_DB:-sub2api}"]
+      test:
+        [
+          "CMD-SHELL",
+          "pg_isready -U ${POSTGRES_USER:-sub2api} -d ${POSTGRES_DB:-sub2api}",
+        ]
       interval: 10s
       timeout: 5s
       retries: 5
       start_period: 10s
+    ports:
+      - 5432:5432
     # 注意：不暴露端口到宿主机，应用通过内部网络连接
     # 如需调试，可临时添加：ports: ["127.0.0.1:5433:5432"]
 
@@ -199,12 +219,12 @@ services:
     volumes:
       - redis_data:/data
     command: >
-        sh -c '
-          redis-server
-          --save 60 1
-          --appendonly yes
-          --appendfsync everysec
-          ${REDIS_PASSWORD:+--requirepass "$REDIS_PASSWORD"}'
+      sh -c '
+        redis-server
+        --save 60 1
+        --appendonly yes
+        --appendfsync everysec
+        ${REDIS_PASSWORD:+--requirepass "$REDIS_PASSWORD"}'
     environment:
       - TZ=${TZ:-Asia/Shanghai}
       # REDISCLI_AUTH is used by redis-cli for authentication (safer than -a flag)
@@ -217,7 +237,8 @@ services:
       timeout: 5s
       retries: 5
       start_period: 5s
-
+    ports:
+      - 6379:6379
 # =============================================================================
 # Volumes
 # =============================================================================