Version: 0.9.65.dev.260503

后端：
1. 阶段 1.5/1.6
收口 llm-service / rag-service，统一模型出口与检索基础设施入口，清退 backend/infra/llm 与 backend/infra/rag 旧实现；
2. 同步更新相关调用链与微服务迁移计划文档

backend/memory/worker/decision_flow.go

@@ -4,11 +4,11 @@ import (
 	"context"
 	"fmt"
 
-	infrarag "github.com/LoveLosita/smartflow/backend/infra/rag"
 	memorymodel "github.com/LoveLosita/smartflow/backend/memory/model"
 	memoryrepo "github.com/LoveLosita/smartflow/backend/memory/repo"
 	memoryutils "github.com/LoveLosita/smartflow/backend/memory/utils"
 	"github.com/LoveLosita/smartflow/backend/model"
+	ragservice "github.com/LoveLosita/smartflow/backend/services/rag"
 	"gorm.io/gorm"
 )
 
@@ -192,7 +192,7 @@ func (r *Runner) recallCandidates(
 ) candidateRecallResult {
 	// 1. 优先使用 Milvus 向量语义召回。
 	if r.ragRuntime != nil {
-		retrieveResult, err := r.ragRuntime.RetrieveMemory(ctx, infrarag.MemoryRetrieveRequest{
+		retrieveResult, err := r.ragRuntime.RetrieveMemory(ctx, ragservice.MemoryRetrieveRequest{
 			Query:       fact.Content,
 			TopK:        r.cfg.DecisionCandidateTopK,
 			Threshold:   r.cfg.DecisionCandidateMinScore,
@@ -235,7 +235,7 @@ func (r *Runner) recallCandidates(
 // 1. 从 DocumentID（格式 memory:{id}）解析出 mysql_id；
 // 2. 从 metadata 提取 title 和 memory_type；
 // 3. 跳过无法解析 DocumentID 的结果。
-func (r *Runner) buildCandidatesFromRAG(hits []infrarag.RetrieveHit) []memorymodel.CandidateSnapshot {
+func (r *Runner) buildCandidatesFromRAG(hits []ragservice.RetrieveHit) []memorymodel.CandidateSnapshot {
 	candidates := make([]memorymodel.CandidateSnapshot, 0, len(hits))
 	for _, hit := range hits {
 		memoryID := parseMemoryID(hit.DocumentID)

backend/memory/worker/runner.go

@@ -9,7 +9,6 @@ import (
 	"strings"
 	"time"
 
-	infrarag "github.com/LoveLosita/smartflow/backend/infra/rag"
 	memorymodel "github.com/LoveLosita/smartflow/backend/memory/model"
 	memoryobserve "github.com/LoveLosita/smartflow/backend/memory/observe"
 	memoryorchestrator "github.com/LoveLosita/smartflow/backend/memory/orchestrator"
@@ -17,6 +16,7 @@ import (
 	memoryutils "github.com/LoveLosita/smartflow/backend/memory/utils"
 	memoryvectorsync "github.com/LoveLosita/smartflow/backend/memory/vectorsync"
 	"github.com/LoveLosita/smartflow/backend/model"
+	ragservice "github.com/LoveLosita/smartflow/backend/services/rag"
 	"gorm.io/gorm"
 )
 
@@ -41,7 +41,7 @@ type Runner struct {
 	auditRepo    *memoryrepo.AuditRepo
 	settingsRepo *memoryrepo.SettingsRepo
 	extractor    Extractor
-	ragRuntime   infrarag.Runtime
+	ragRuntime   ragservice.Runtime
 	logger       *log.Logger
 	vectorSyncer *memoryvectorsync.Syncer
 	observer     memoryobserve.Observer
@@ -63,7 +63,7 @@ func NewRunner(
 	auditRepo *memoryrepo.AuditRepo,
 	settingsRepo *memoryrepo.SettingsRepo,
 	extractor Extractor,
-	ragRuntime infrarag.Runtime,
+	ragRuntime ragservice.Runtime,
 	cfg memorymodel.Config,
 	decisionOrchestrator *memoryorchestrator.LLMDecisionOrchestrator,
 	vectorSyncer *memoryvectorsync.Syncer,