refactor: P0-P3 fixes, C1 cleanup, architecture diagrams, go.work upgrade

- P0-1: ProviderError type + ReportStatus for precise 401/403 detection
- P0-2: Remove -config flag from deploy/homeagent.service
- P2-1: 5s debounce on context.go Save()
- P2-2→C1: Delete output_set_channel entirely
- P2-3: Extract mediaDataURL/mediaChat helpers
- P2-4: Dedup defaultSources var
- P3: Delete dead packages (embed/tokenizer/container/snapshot)
- P3: Delete dead functions (messagesToMap, RunStageAll)
- CL: Update .gitignore, docs, Makefile, gojieba removal
- Config: Delete config/config.yaml, update docs
- Arch: Remove EmitOutputTo from emitResponse
- CL-1: go.work 1.19→1.21
- Docs: Add Mermaid architecture diagrams to README
- Docs: Add kernel-rebuild requires plugin-rebuild note to PLUGIN_DEV.md

internal/memory/document/document.go

@@ -37,11 +37,13 @@ type Store struct {
 	mu     sync.RWMutex
 
 	docs      map[string]*Doc
-	summaries []string // 用于训练向量化器
+	summaries []string // 用于训练向量化器，最大 10000 条
 
 	dirty bool
 }
 
+const maxSummaries = 10000
+
 func NewStore(dir string) *Store {
 	return &Store{
 		dir:    dir,
@@ -83,11 +85,15 @@ func (s *Store) Insert(doc *Doc) error {
 
 	s.docs[doc.ID] = doc
 
+	// 增量训练向量化器并加入向量索引
+	s.addSummary(doc.Summary)
 	vec := s.veczer.Vectorize(doc.Summary + " " + doc.Content)
 	s.vec.Insert(doc.ID, doc.Summary, vec, doc.Meta)
 
-	// 更新训练集
-	s.summaries = append(s.summaries, doc.Summary)
+	// 立即写盘
+	path := filepath.Join(s.dir, doc.ID+".json")
+	data, _ := json.MarshalIndent(doc, "", "  ")
+	os.WriteFile(path, data, 0644)
 
 	s.dirty = true
 	return nil
@@ -110,12 +116,13 @@ func (s *Store) ContextToDoc(source string, entries []ContextEntry) (*Doc, error
 	content := strings.Join(parts, "\n")
 	contentHash := simpleHash(content)
 
-	// 去重：检查是否已有相同 hash 的文档（在锁内完成创建/更新）
 	summary := summarizeEntries(entries)
 	tags := extractTags(entries)
 	entities := extractEntities(entries)
 
 	s.mu.Lock()
+
+	// 去重
 	for _, d := range s.docs {
 		if d.Meta != nil && d.Meta["content_hash"] == contentHash {
 			d.UpdatedAt = time.Now()
@@ -146,8 +153,20 @@ func (s *Store) ContextToDoc(source string, entries []ContextEntry) (*Doc, error
 		Meta:        map[string]string{"content_hash": contentHash},
 	}
 	s.docs[id] = doc
+
+	// 增量训练向量化器并加入向量索引
+	s.addSummary(summary)
+	vec := s.veczer.Vectorize(summary + " " + content)
+	s.vec.Insert(id, summary, vec, nil)
+
 	s.dirty = true
 	s.mu.Unlock()
+
+	// 立即写盘
+	path := filepath.Join(s.dir, id+".json")
+	data, _ := json.MarshalIndent(doc, "", "  ")
+	os.WriteFile(path, data, 0644)
+
 	return doc, nil
 }
 
@@ -166,8 +185,7 @@ func (s *Store) Consume(text string, topK int) []*Doc {
 	var docs []*Doc
 	for _, r := range results {
 		if d, ok := s.docs[r.ID]; ok {
-			delete(s.docs, r.ID)
-			s.vec.Remove(r.ID)
+			s.removeDoc(r.ID)
 			s.dirty = true
 			docs = append(docs, d)
 		}
@@ -266,14 +284,39 @@ func (s *Store) Remove(id string) {
 	defer s.mu.Unlock()
 
 	if _, ok := s.docs[id]; ok {
-		delete(s.docs, id)
-		s.vec.Remove(id)
+		s.removeDoc(id)
 		s.dirty = true
 	}
 }
 
 // ——— internal ———
 
+// addSummary 添加一条摘要到训练集，超限时截断并触发重索引。
+// 调用方必须已持有 s.mu 写锁。
+func (s *Store) addSummary(summary string) {
+	s.summaries = append(s.summaries, summary)
+	if len(s.summaries) > maxSummaries {
+		n := maxSummaries / 2
+		copy(s.summaries, s.summaries[len(s.summaries)-n:])
+		s.summaries = s.summaries[:n]
+		s.veczer.Train(s.summaries)
+		s.vec = vector.NewStore()
+		for _, doc := range s.docs {
+			vec := s.veczer.Vectorize(doc.Summary + " " + doc.Content)
+			s.vec.Insert(doc.ID, doc.Summary, vec, nil)
+		}
+	}
+}
+
+// removeDoc 从内存索引和磁盘删除文档。
+// 调用方必须已持有 s.mu 写锁。
+func (s *Store) removeDoc(id string) {
+	delete(s.docs, id)
+	s.vec.Remove(id)
+	path := filepath.Join(s.dir, id+".json")
+	os.Remove(path)
+}
+
 func (s *Store) loadAll() error {
 	entries, err := os.ReadDir(s.dir)
 	if err != nil {

internal/memory/graph.go

@@ -526,7 +526,7 @@ func (g *GraphDB) Introspect() (map[string]interface{}, error) {
 // MergeEntities 合并两个实体：将 sourceName 的所有信息合并到 targetName
 // 1. sourceName 的所有关系重新指向 targetName
 // 2. targetName 的 mention_count 增加 sourceName 的计数
-// 3. sourceName 标记为 merged
+// 3. sourceName 彻底删除（不再残留 @merged_ 实体）
 // 返回 (关系的重定向数, error)
 func (g *GraphDB) MergeEntities(sourceName, targetName string) (int, error) {
 	g.mu.Lock()
@@ -554,7 +554,7 @@ func (g *GraphDB) MergeEntities(sourceName, targetName string) (int, error) {
 		return 0, fmt.Errorf("cannot merge entity with itself")
 	}
 
-	// 重定向 source → target 的关系（作为 source）
+	// 重定向 source → target 的活跃关系（作为 source）
 	res, err := tx.Exec(
 		`UPDATE relations SET source_id = ?, updated_at = CURRENT_TIMESTAMP
 		 WHERE source_id = ? AND status = 'active'`,
@@ -565,7 +565,7 @@ func (g *GraphDB) MergeEntities(sourceName, targetName string) (int, error) {
 	}
 	redirectedSource, _ := res.RowsAffected()
 
-	// 重定向 source → target 的关系（作为 target）
+	// 重定向 source → target 的活跃关系（作为 target）
 	res, err = tx.Exec(
 		`UPDATE relations SET target_id = ?, updated_at = CURRENT_TIMESTAMP
 		 WHERE target_id = ? AND status = 'active'`,
@@ -586,6 +586,12 @@ func (g *GraphDB) MergeEntities(sourceName, targetName string) (int, error) {
 		return 0, err
 	}
 
+	// 清理 source 残留的非活跃关系（archived/deleted），否则外键约束阻止删除实体
+	_, err = tx.Exec(`DELETE FROM relations WHERE source_id = ? OR target_id = ?`, sourceID, sourceID)
+	if err != nil {
+		return 0, err
+	}
+
 	// 更新 target 的 mention_count
 	_, err = tx.Exec(
 		`UPDATE entities SET mention_count = ?, updated_at = CURRENT_TIMESTAMP WHERE id = ?`,
@@ -595,14 +601,8 @@ func (g *GraphDB) MergeEntities(sourceName, targetName string) (int, error) {
 		return 0, err
 	}
 
-	// 标记 source 为 merged（改名避免 UNIQUE 冲突）
-	_, err = tx.Exec(
-		`UPDATE entities SET name = ? || '@merged_' || ?,
-			mention_count = 0,
-			updated_at = CURRENT_TIMESTAMP
-		 WHERE id = ?`,
-		sourceName, time.Now().Format("20060102150405"), sourceID,
-	)
+	// 彻底删除 source 实体（所有关系已重定向，自引用已删除）
+	_, err = tx.Exec(`DELETE FROM entities WHERE id = ?`, sourceID)
 	if err != nil {
 		return 0, err
 	}
@@ -615,6 +615,36 @@ func (g *GraphDB) MergeEntities(sourceName, targetName string) (int, error) {
 	return total, nil
 }
 
+// DeleteEntity 彻底删除一个实体及其所有关联关系。
+func (g *GraphDB) DeleteEntity(name string) error {
+	g.mu.Lock()
+	defer g.mu.Unlock()
+
+	tx, err := g.db.Begin()
+	if err != nil {
+		return err
+	}
+	defer tx.Rollback()
+
+	var id int64
+	err = tx.QueryRow("SELECT id FROM entities WHERE name = ?", name).Scan(&id)
+	if err != nil {
+		return fmt.Errorf("entity '%s' not found: %w", name, err)
+	}
+
+	_, err = tx.Exec(`DELETE FROM relations WHERE source_id = ? OR target_id = ?`, id, id)
+	if err != nil {
+		return err
+	}
+
+	_, err = tx.Exec(`DELETE FROM entities WHERE id = ?`, id)
+	if err != nil {
+		return err
+	}
+
+	return tx.Commit()
+}
+
 func (g *GraphDB) Archive(days int) (int, error) {
 	g.mu.Lock()
 	defer g.mu.Unlock()

internal/memory/pipeline/pipeline.go

@@ -88,7 +88,7 @@ func (d *Distiller) flush() {
 	if len(d.records) == 0 {
 		return
 	}
-	path := filepath.Join(d.rawPath, fmt.Sprintf("raw_%d.jsonl", time.Now().UnixNano()))
+	path := filepath.Join(d.rawPath, fmt.Sprintf("raw_%d.tsv", time.Now().UnixNano()))
 	f, err := os.Create(path)
 	if err != nil {
 		log.Printf("[memory] flush error: %v", err)
@@ -113,7 +113,8 @@ func (d *Distiller) loadExisting() {
 	}
 	var files []fileInfo
 	for _, entry := range entries {
-		if filepath.Ext(entry.Name()) != ".jsonl" {
+		ext := filepath.Ext(entry.Name())
+		if ext != ".tsv" && ext != ".jsonl" {
 			continue
 		}
 		info, err := entry.Info()
@@ -296,8 +297,8 @@ func extractName(s string) string {
 	}{
 		{"我叫", ""},
 		{"我的名字是", ""},
-		{"我是", ""},
 		{"名字是", ""},
+		{"我是", ""},
 	}
 	s = strings.TrimSpace(s)
 	for _, p := range patterns {
@@ -313,6 +314,10 @@ func extractName(s string) string {
 					candidate = candidate[:idx]
 				}
 			}
+			// "我是张三"(姓名) vs "我是一个程序员"(职业)：名字通常 ≤4 字符
+			if p.prefix == "我是" && len([]rune(candidate)) > 4 {
+				continue
+			}
 			if len(candidate) > 0 && len(candidate) < 20 {
 				return candidate
 			}