refactor: 将客服后端重构为宿主可嵌入模块

- 注入数据库、运行时配置、统一响应、文件存储和平台 AI 能力，补充业务读写工具与客户快捷操作契约。

- 移除模块内重复的组织、客户、工单、标签、技能、旧工作流、MCP 和迁移实现，将身份权限与业务主体交由宿主管理。

- 使用 libSQL 重构向量存储，并完善图片消息、访客身份、排队调度、企业微信和支持聊天页面。

- 统一 HTTP、DTO 与 WebSocket 的 snake_case 协议，补齐模块初始化、业务动作和公共载荷等回归测试。

internal/ai/rag/retrieve.go

@@ -2,10 +2,12 @@ package rag
 
 import (
 	"context"
+	"errors"
 	"fmt"
 	"log/slog"
 	"strings"
 
+	"code.tczkiot.com/wlw/ai-agent/internal/ai"
 	"code.tczkiot.com/wlw/ai-agent/internal/ai/rag/vectordb"
 	"code.tczkiot.com/wlw/ai-agent/internal/models"
 	"code.tczkiot.com/wlw/ai-agent/internal/pkg/enums"
@@ -15,6 +17,7 @@ import (
 )
 
 type retrieve struct {
+	rerankResults func(context.Context, string, []RetrieveResult, int) ([]RetrieveResult, error)
 }
 
 var Retrieve = &retrieve{}
@@ -117,14 +120,22 @@ func (s *retrieve) RetrieveWithRerank(ctx context.Context, req RetrieveRequest,
 	if err != nil {
 		return nil, err
 	}
+	return s.ApplyRerank(ctx, req.Query, results, rerankLimit)
+}
 
-	if len(results) <= rerankLimit {
+// ApplyRerank reranks an existing vector result set. Keeping rerank separate
+// from retrieval prevents callers from generating and billing the query
+// embedding a second time.
+func (s *retrieve) ApplyRerank(ctx context.Context, query string, results []RetrieveResult, rerankLimit int) ([]RetrieveResult, error) {
+	if rerankLimit <= 0 || len(results) <= rerankLimit {
 		return results, nil
 	}
 
-	rerankedResults, err := s.rerank(ctx, req.Query, results, rerankLimit)
+	rerankedResults, err := s.rerank(ctx, query, results, rerankLimit)
 	if err != nil {
-		slog.Warn("Rerank failed, returning original results", "error", err)
+		if !errors.Is(err, ai.ErrPlatformModelUnsupported) {
+			slog.Warn("Rerank failed, returning original results", "error", err)
+		}
 		if len(results) > rerankLimit {
 			return results[:rerankLimit], nil
 		}
@@ -135,6 +146,9 @@ func (s *retrieve) RetrieveWithRerank(ctx context.Context, req RetrieveRequest,
 }
 
 func (s *retrieve) rerank(ctx context.Context, query string, results []RetrieveResult, limit int) ([]RetrieveResult, error) {
+	if s.rerankResults != nil {
+		return s.rerankResults(ctx, query, results, limit)
+	}
 	return Rerank.RerankResults(ctx, query, results, limit)
 }
 
@@ -222,9 +236,9 @@ func (s *retrieve) loadRetrievableKnowledgeBases(ids []int64) []models.Knowledge
 }
 
 type KnowledgeBaseStats struct {
-	KnowledgeBaseID int64 `json:"knowledgeBaseId"`
-	DocumentCount   int64 `json:"documentCount"`
-	PublishedCount  int64 `json:"publishedCount"`
-	ChunkCount      int64 `json:"chunkCount"`
-	VectorCount     int   `json:"vectorCount"`
+	KnowledgeBaseID int64 `json:"knowledge_base_id"`
+	DocumentCount   int64 `json:"document_count"`
+	PublishedCount  int64 `json:"published_count"`
+	ChunkCount      int64 `json:"chunk_count"`
+	VectorCount     int   `json:"vector_count"`
 }