Version: 0.9.14.dev.260410
后端：
1. LLM 客户端从 newAgent/llm 提升为 infra/llm 基础设施层
- 删除 backend/newAgent/llm/（ark.go / ark_adapter.go / client.go / json.go）
- 等价迁移至 backend/infra/llm/，所有 newAgent node 与 service 统一改引用 infrallm
- 消除 newAgent 对模型客户端的私有依赖，为 memory / websearch 等多模块复用铺路
2. RAG 基础设施完成可运行态接入（factory / runtime / observer / service 四层成型）
- 新建 backend/infra/rag/factory.go / runtime.go / observe.go / observer.go /
service.go：工厂创建、运行时生命周期、轻量观测接口、检索服务门面
- 更新 infra/rag/config/config.go：补齐 Milvus / Embed / Reranker 全部配置项与默认值
- 更新 infra/rag/embed/eino_embedder.go：增强 Eino embedding 适配，支持 BaseURL / APIKey 环境变量 / 超时 /
维度等参数
- 更新 infra/rag/store/milvus_store.go：完整实现 Milvus 向量存储（建集合 / 建 Index / Upsert / Search /
Delete），支持 COSINE / L2 / IP 度量
- 更新 infra/rag/core/pipeline.go：适配 Runtime 接口，Pipeline 由 factory 注入而非手动拼装
- 更新 infra/rag/corpus/memory_corpus.go / vector_store.go：对接 Memory 模块数据源与 Store 接口扩展
3. Memory 模块从 Day1 骨架升级为 Day2 完整可运行态
- 新建 memory/module.go：统一门面 Module，对外封装 EnqueueExtract / ReadService / ManageService / WithTx /
StartWorker，启动层只依赖这一个入口
- 新建 memory/orchestrator/llm_write_orchestrator.go：LLM 驱动的记忆抽取编排器，替代原 mock 抽取
- 新建 memory/service/read_service.go：按用户开关过滤 + 轻量重排 + 访问时间刷新的读取链路
- 新建 memory/service/manage_service.go：记忆管理面能力（列出 / 软删除 / 开关读写），删除同步写审计日志
- 新建 memory/service/common.go：服务层公共工具
- 新建 memory/worker/loop.go：后台轮询循环 RunPollingLoop，定时抢占 pending 任务并推进
- 新建 memory/utils/audit.go / settings.go：审计日志构造、用户设置过滤等纯函数
- 更新 memory/model/item.go / job.go / settings.go / config.go / status.go：补齐 DTO 字段与状态常量
- 更新 memory/repo/item_repo.go / job_repo.go / audit_repo.go / settings_repo.go：补齐 CRUD 与查询能力
- 更新 memory/worker/runner.go：Runner 对接 Module 与 LLM 抽取器，任务状态机完整化
- 更新 memory/README.md：同步模块现状说明
4. newAgent 接入 Memory 读取注入与工具注册依赖预埋
- 新建 service/agentsvc/agent_memory.go：定义 MemoryReader 接口 + injectMemoryContext，在 graph
执行前统一补充记忆上下文
- 更新 service/agentsvc/agent.go：新增 memoryReader 字段与 SetMemoryReader 方法
- 更新 service/agentsvc/agent_newagent.go：调用 injectMemoryContext 注入 pinned block，检索失败仅降级不阻断主链路
- 更新 newAgent/tools/registry.go：新增 DefaultRegistryDeps（含 RAGRuntime），工具注册表支持依赖注入
5. 启动流程与事件处理器接线更新
- 更新 cmd/start.go：初始化 RAG Runtime → Memory Module → 注册事件处理器 → 启动 Worker 后台轮询
- 更新 service/events/memory_extract_requested.go：改用 memory.Module.WithTx(tx) 统一门面，事件处理器不再直接依赖
repo/service 内部包
6. 缓存插件与配置同步
- 更新 middleware/cache_deleter.go：静默忽略 MemoryJob / MemoryItem / MemoryAuditLog / MemoryUserSetting
等新模型，避免日志刷屏；清理冗余注释
- 更新 config.example.yaml：补齐 rag / memory / websearch 配置段及默认值
- 更新 go.mod / go.sum：新增 eino-ext/openai / json-patch / go-openai 依赖
前端：无 仓库：无
This commit is contained in:
Losita
380
backend/infra/rag/runtime.go
Normal file
380
backend/infra/rag/runtime.go
Normal file
@@ -0,0 +1,380 @@
|
||||
package rag
|
||||
|
||||
import (
|
||||
"context"
|
||||
"fmt"
|
||||
"strings"
|
||||
"time"
|
||||
|
||||
ragconfig "github.com/LoveLosita/smartflow/backend/infra/rag/config"
|
||||
"github.com/LoveLosita/smartflow/backend/infra/rag/core"
|
||||
"github.com/LoveLosita/smartflow/backend/infra/rag/corpus"
|
||||
)
|
||||
|
||||
type runtime struct {
|
||||
cfg ragconfig.Config
|
||||
pipeline *core.Pipeline
|
||||
memoryCorpus *corpus.MemoryCorpus
|
||||
webCorpus *corpus.WebCorpus
|
||||
observer Observer
|
||||
}
|
||||
|
||||
func newRuntime(cfg ragconfig.Config, pipeline *core.Pipeline, observer Observer) Runtime {
|
||||
if observer == nil {
|
||||
observer = NewNopObserver()
|
||||
}
|
||||
return &runtime{
|
||||
cfg: cfg,
|
||||
pipeline: pipeline,
|
||||
memoryCorpus: corpus.NewMemoryCorpus(),
|
||||
webCorpus: corpus.NewWebCorpus(),
|
||||
observer: observer,
|
||||
}
|
||||
}
|
||||
|
||||
// IngestMemory 统一承接记忆语料入库。
|
||||
func (r *runtime) IngestMemory(ctx context.Context, req MemoryIngestRequest) (*IngestResult, error) {
|
||||
items := make([]corpus.MemoryIngestItem, 0, len(req.Items))
|
||||
for _, item := range req.Items {
|
||||
items = append(items, corpus.MemoryIngestItem{
|
||||
MemoryID: item.MemoryID,
|
||||
UserID: item.UserID,
|
||||
ConversationID: item.ConversationID,
|
||||
AssistantID: item.AssistantID,
|
||||
RunID: item.RunID,
|
||||
MemoryType: item.MemoryType,
|
||||
Title: item.Title,
|
||||
Content: item.Content,
|
||||
Confidence: item.Confidence,
|
||||
Importance: item.Importance,
|
||||
SensitivityLevel: item.SensitivityLevel,
|
||||
IsExplicit: item.IsExplicit,
|
||||
Status: item.Status,
|
||||
TTLAt: item.TTLAt,
|
||||
CreatedAt: item.CreatedAt,
|
||||
})
|
||||
}
|
||||
return r.ingestWithCorpus(ctx, req.TraceID, "memory", r.memoryCorpus, items, req.Action)
|
||||
}
|
||||
|
||||
// RetrieveMemory 统一承接记忆语料检索。
|
||||
func (r *runtime) RetrieveMemory(ctx context.Context, req MemoryRetrieveRequest) (*RetrieveResult, error) {
|
||||
corpusInput := corpus.MemoryRetrieveInput{
|
||||
UserID: req.UserID,
|
||||
ConversationID: req.ConversationID,
|
||||
AssistantID: req.AssistantID,
|
||||
RunID: req.RunID,
|
||||
}
|
||||
if len(req.MemoryTypes) == 1 {
|
||||
corpusInput.MemoryType = req.MemoryTypes[0]
|
||||
}
|
||||
|
||||
result, err := r.retrieveWithCorpus(ctx, req.TraceID, "memory", r.memoryCorpus, core.RetrieveRequest{
|
||||
Query: req.Query,
|
||||
TopK: normalizeTopK(req.TopK, r.cfg.TopK),
|
||||
Threshold: normalizeThreshold(req.Threshold, r.cfg.Threshold),
|
||||
Action: normalizeAction(req.Action, "search"),
|
||||
CorpusInput: corpusInput,
|
||||
})
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
if len(req.MemoryTypes) <= 1 {
|
||||
return result, nil
|
||||
}
|
||||
|
||||
// 1. 当前底层过滤仍以等值条件为主，先保持 Runtime 做多类型二次筛选；
|
||||
// 2. 这样可以避免把 “memory_type in (...)” 的实现细节扩散到所有 Store；
|
||||
// 3. 等后续底层过滤能力统一后，再考虑把该逻辑继续下沉。
|
||||
allowed := make(map[string]struct{}, len(req.MemoryTypes))
|
||||
for _, item := range req.MemoryTypes {
|
||||
value := strings.TrimSpace(strings.ToLower(item))
|
||||
if value == "" {
|
||||
continue
|
||||
}
|
||||
allowed[value] = struct{}{}
|
||||
}
|
||||
|
||||
filtered := make([]RetrieveHit, 0, len(result.Items))
|
||||
for _, item := range result.Items {
|
||||
memoryType := strings.TrimSpace(strings.ToLower(asString(item.Metadata["memory_type"])))
|
||||
if len(allowed) > 0 {
|
||||
if _, ok := allowed[memoryType]; !ok {
|
||||
continue
|
||||
}
|
||||
}
|
||||
filtered = append(filtered, item)
|
||||
}
|
||||
result.Items = filtered
|
||||
if req.TopK > 0 && len(result.Items) > req.TopK {
|
||||
result.Items = result.Items[:req.TopK]
|
||||
}
|
||||
return result, nil
|
||||
}
|
||||
|
||||
// IngestWeb 统一承接网页语料入库。
|
||||
func (r *runtime) IngestWeb(ctx context.Context, req WebIngestRequest) (*IngestResult, error) {
|
||||
items := make([]corpus.WebIngestItem, 0, len(req.Items))
|
||||
for _, item := range req.Items {
|
||||
items = append(items, corpus.WebIngestItem{
|
||||
URL: item.URL,
|
||||
Title: item.Title,
|
||||
Content: item.Content,
|
||||
Snippet: item.Snippet,
|
||||
Domain: item.Domain,
|
||||
QueryID: item.QueryID,
|
||||
SessionID: item.SessionID,
|
||||
PublishedAt: item.PublishedAt,
|
||||
FetchedAt: item.FetchedAt,
|
||||
SourceRank: item.SourceRank,
|
||||
})
|
||||
}
|
||||
return r.ingestWithCorpus(ctx, req.TraceID, "web", r.webCorpus, items, req.Action)
|
||||
}
|
||||
|
||||
// RetrieveWeb 统一承接网页语料检索。
|
||||
func (r *runtime) RetrieveWeb(ctx context.Context, req WebRetrieveRequest) (*RetrieveResult, error) {
|
||||
return r.retrieveWithCorpus(ctx, req.TraceID, "web", r.webCorpus, core.RetrieveRequest{
|
||||
Query: req.Query,
|
||||
TopK: normalizeTopK(req.TopK, r.cfg.TopK),
|
||||
Threshold: normalizeThreshold(req.Threshold, r.cfg.Threshold),
|
||||
Action: normalizeAction(req.Action, "search"),
|
||||
CorpusInput: corpus.WebRetrieveInput{
|
||||
QueryID: req.QueryID,
|
||||
SessionID: req.SessionID,
|
||||
Domain: req.Domain,
|
||||
},
|
||||
})
|
||||
}
|
||||
|
||||
func (r *runtime) ingestWithCorpus(
|
||||
ctx context.Context,
|
||||
traceID string,
|
||||
corpusName string,
|
||||
adapter core.CorpusAdapter,
|
||||
input any,
|
||||
action string,
|
||||
) (*IngestResult, error) {
|
||||
start := time.Now()
|
||||
if r == nil || r.pipeline == nil || adapter == nil {
|
||||
return nil, core.ErrNilDependency
|
||||
}
|
||||
|
||||
action = normalizeAction(action, "add")
|
||||
observeCtx := newObserveContext(ctx, traceID, corpusName, action)
|
||||
|
||||
docs, err := adapter.BuildIngestDocuments(observeCtx, input)
|
||||
if err != nil {
|
||||
r.observe(observeCtx, ObserveEvent{
|
||||
Level: ObserveLevelError,
|
||||
Component: "runtime",
|
||||
Operation: "ingest",
|
||||
Fields: map[string]any{
|
||||
"status": "failed",
|
||||
"latency_ms": time.Since(start).Milliseconds(),
|
||||
"phase": "build_documents",
|
||||
"error": err,
|
||||
"error_code": core.ClassifyErrorCode(err),
|
||||
"input_count": estimateInputCount(input),
|
||||
},
|
||||
})
|
||||
return nil, err
|
||||
}
|
||||
|
||||
docIDs := make([]string, 0, len(docs))
|
||||
for _, doc := range docs {
|
||||
docIDs = append(docIDs, doc.ID)
|
||||
}
|
||||
|
||||
result, err := r.pipeline.IngestDocuments(observeCtx, adapter.Name(), docs, core.IngestOption{
|
||||
Chunk: core.ChunkOption{
|
||||
ChunkSize: r.cfg.ChunkSize,
|
||||
ChunkOverlap: r.cfg.ChunkOverlap,
|
||||
},
|
||||
Action: action,
|
||||
})
|
||||
if err != nil {
|
||||
r.observe(observeCtx, ObserveEvent{
|
||||
Level: ObserveLevelError,
|
||||
Component: "runtime",
|
||||
Operation: "ingest",
|
||||
Fields: map[string]any{
|
||||
"status": "failed",
|
||||
"latency_ms": time.Since(start).Milliseconds(),
|
||||
"document_count": len(docs),
|
||||
"error": err,
|
||||
"error_code": core.ClassifyErrorCode(err),
|
||||
},
|
||||
})
|
||||
return nil, err
|
||||
}
|
||||
|
||||
r.observe(observeCtx, ObserveEvent{
|
||||
Level: ObserveLevelInfo,
|
||||
Component: "runtime",
|
||||
Operation: "ingest",
|
||||
Fields: map[string]any{
|
||||
"status": "success",
|
||||
"latency_ms": time.Since(start).Milliseconds(),
|
||||
"document_count": result.DocumentCount,
|
||||
"chunk_count": result.ChunkCount,
|
||||
},
|
||||
})
|
||||
return &IngestResult{
|
||||
DocumentCount: result.DocumentCount,
|
||||
ChunkCount: result.ChunkCount,
|
||||
DocumentIDs: docIDs,
|
||||
}, nil
|
||||
}
|
||||
|
||||
func (r *runtime) retrieveWithCorpus(
|
||||
ctx context.Context,
|
||||
traceID string,
|
||||
corpusName string,
|
||||
adapter core.CorpusAdapter,
|
||||
req core.RetrieveRequest,
|
||||
) (*RetrieveResult, error) {
|
||||
start := time.Now()
|
||||
if r == nil || r.pipeline == nil || adapter == nil {
|
||||
return nil, core.ErrNilDependency
|
||||
}
|
||||
|
||||
action := normalizeAction(req.Action, "search")
|
||||
req.Action = action
|
||||
observeCtx := newObserveContext(ctx, traceID, corpusName, action)
|
||||
|
||||
timeoutCtx := observeCtx
|
||||
cancel := func() {}
|
||||
if r.cfg.RetrieveTimeoutMS > 0 {
|
||||
timeoutCtx, cancel = context.WithTimeout(observeCtx, time.Duration(r.cfg.RetrieveTimeoutMS)*time.Millisecond)
|
||||
}
|
||||
defer cancel()
|
||||
|
||||
result, err := r.pipeline.Retrieve(timeoutCtx, adapter, req)
|
||||
if err != nil {
|
||||
r.observe(observeCtx, ObserveEvent{
|
||||
Level: ObserveLevelError,
|
||||
Component: "runtime",
|
||||
Operation: "retrieve",
|
||||
Fields: map[string]any{
|
||||
"status": "failed",
|
||||
"latency_ms": time.Since(start).Milliseconds(),
|
||||
"query_len": len(strings.TrimSpace(req.Query)),
|
||||
"top_k": req.TopK,
|
||||
"threshold": req.Threshold,
|
||||
"error": err,
|
||||
"error_code": core.ClassifyErrorCode(err),
|
||||
},
|
||||
})
|
||||
return nil, err
|
||||
}
|
||||
|
||||
items := make([]RetrieveHit, 0, len(result.Items))
|
||||
for _, item := range result.Items {
|
||||
items = append(items, RetrieveHit{
|
||||
ChunkID: item.ChunkID,
|
||||
DocumentID: item.DocumentID,
|
||||
Text: item.Text,
|
||||
Score: item.Score,
|
||||
Metadata: cloneMap(item.Metadata),
|
||||
})
|
||||
}
|
||||
|
||||
r.observe(observeCtx, ObserveEvent{
|
||||
Level: ObserveLevelInfo,
|
||||
Component: "runtime",
|
||||
Operation: "retrieve",
|
||||
Fields: map[string]any{
|
||||
"status": "success",
|
||||
"latency_ms": time.Since(start).Milliseconds(),
|
||||
"query_len": len(strings.TrimSpace(req.Query)),
|
||||
"top_k": req.TopK,
|
||||
"threshold": req.Threshold,
|
||||
"raw_count": result.RawCount,
|
||||
"hit_count": len(result.Items),
|
||||
"fallback_used": result.FallbackUsed,
|
||||
"fallback_reason": result.FallbackReason,
|
||||
},
|
||||
})
|
||||
return &RetrieveResult{
|
||||
Items: items,
|
||||
RawCount: result.RawCount,
|
||||
FallbackUsed: result.FallbackUsed,
|
||||
FallbackReason: result.FallbackReason,
|
||||
}, nil
|
||||
}
|
||||
|
||||
func (r *runtime) observe(ctx context.Context, event ObserveEvent) {
|
||||
if r == nil || r.observer == nil {
|
||||
return
|
||||
}
|
||||
r.observer.Observe(ctx, event)
|
||||
}
|
||||
|
||||
func newObserveContext(ctx context.Context, traceID string, corpusName string, action string) context.Context {
|
||||
fields := map[string]any{
|
||||
"corpus": corpusName,
|
||||
"action": action,
|
||||
}
|
||||
if traceID = strings.TrimSpace(traceID); traceID != "" {
|
||||
fields["trace_id"] = traceID
|
||||
}
|
||||
return core.WithObserveFields(ctx, fields)
|
||||
}
|
||||
|
||||
func estimateInputCount(input any) int {
|
||||
switch value := input.(type) {
|
||||
case []corpus.MemoryIngestItem:
|
||||
return len(value)
|
||||
case []corpus.WebIngestItem:
|
||||
return len(value)
|
||||
default:
|
||||
return 0
|
||||
}
|
||||
}
|
||||
|
||||
func normalizeAction(action string, fallback string) string {
|
||||
action = strings.TrimSpace(action)
|
||||
if action == "" {
|
||||
return fallback
|
||||
}
|
||||
return action
|
||||
}
|
||||
|
||||
func normalizeTopK(topK int, fallback int) int {
|
||||
if topK > 0 {
|
||||
return topK
|
||||
}
|
||||
if fallback > 0 {
|
||||
return fallback
|
||||
}
|
||||
return 8
|
||||
}
|
||||
|
||||
func normalizeThreshold(threshold float64, fallback float64) float64 {
|
||||
if threshold >= 0 {
|
||||
return threshold
|
||||
}
|
||||
if fallback >= 0 {
|
||||
return fallback
|
||||
}
|
||||
return 0
|
||||
}
|
||||
|
||||
func cloneMap(src map[string]any) map[string]any {
|
||||
if len(src) == 0 {
|
||||
return map[string]any{}
|
||||
}
|
||||
dst := make(map[string]any, len(src))
|
||||
for key, value := range src {
|
||||
dst[key] = value
|
||||
}
|
||||
return dst
|
||||
}
|
||||
|
||||
func asString(v any) string {
|
||||
if v == nil {
|
||||
return ""
|
||||
}
|
||||
return strings.TrimSpace(fmt.Sprintf("%v", v))
|
||||
}
|
||||
Reference in New Issue
Block a user