diff --git a/internal/application/service/image_multimodal.go b/internal/application/service/image_multimodal.go
index f56b53539d..754324d742 100644
--- a/internal/application/service/image_multimodal.go
+++ b/internal/application/service/image_multimodal.go
@@ -18,6 +18,7 @@ import (
 	"github.com/Tencent/WeKnora/internal/types"
 	"github.com/Tencent/WeKnora/internal/types/interfaces"
 	secutils "github.com/Tencent/WeKnora/internal/utils"
+	hashutil "github.com/Tencent/WeKnora/internal/utils/hash"
 	"github.com/google/uuid"
 	"github.com/hibiken/asynq"
 	"github.com/redis/go-redis/v9"
diff --git a/internal/application/service/knowledge_process.go b/internal/application/service/knowledge_process.go
index a754e45298..65e3372b5b 100644
--- a/internal/application/service/knowledge_process.go
+++ b/internal/application/service/knowledge_process.go
@@ -23,6 +23,7 @@ import (
 	"github.com/Tencent/WeKnora/internal/types"
 	"github.com/Tencent/WeKnora/internal/types/interfaces"
 	secutils "github.com/Tencent/WeKnora/internal/utils"
+	hashutil "github.com/Tencent/WeKnora/internal/utils/hash"
 	"github.com/google/uuid"
 	"github.com/hibiken/asynq"
 )
@@ -382,20 +383,20 @@ func (s *knowledgeService) processChunks(ctx context.Context,
 	if hasParentChild {
 		parentDBChunks = make([]*types.Chunk, len(options.ParentChunks))
 		for i, pc := range options.ParentChunks {
-			parentDBChunks[i] = &types.Chunk{
-				ID:              uuid.New().String(),
-				TenantID:        knowledge.TenantID,
-				KnowledgeID:     knowledge.ID,
-				KnowledgeBaseID: knowledge.KnowledgeBaseID,
-				Content:         pc.Content,
-				ChunkIndex:      pc.Seq,
-				IsEnabled:       true,
-				CreatedAt:       time.Now(),
-				UpdatedAt:       time.Now(),
-				StartAt:         pc.Start,
-				EndAt:           pc.End,
-				ChunkType:       types.ChunkTypeParentText,
-			}
+		parentDBChunks[i] = &types.Chunk{
+			ID:              hashutil.StableParentChunkID(knowledge.ID, pc.Content, pc.Seq),
+			TenantID:        knowledge.TenantID,
+			KnowledgeID:     knowledge.ID,
+			KnowledgeBaseID: knowledge.KnowledgeBaseID,
+			Content:         pc.Content,
+			ChunkIndex:      pc.Seq,
+			IsEnabled:       true,
+			CreatedAt:       time.Now(),
+			UpdatedAt:       time.Now(),
+			StartAt:         pc.Start,
+			EndAt:           pc.End,
+			ChunkType:       types.ChunkTypeParentText,
+		}
 		}
 		// Set prev/next links for parent chunks
 		for i := range parentDBChunks {
@@ -420,9 +421,9 @@ func (s *knowledgeService) processChunks(ctx context.Context,
 			continue
 		}
 
-		// 创建主文本Chunk
+		// 创建主文本Chunk (stable ID for cache reuse across reparses)
 		textChunk := &types.Chunk{
-			ID:              uuid.New().String(),
+			ID:              hashutil.StableChunkID(knowledge.ID, chunkData.Content, int(chunkData.Seq)),
 			TenantID:        knowledge.TenantID,
 			KnowledgeID:     knowledge.ID,
 			KnowledgeBaseID: knowledge.KnowledgeBaseID,
diff --git a/internal/utils/hash/content.go b/internal/utils/hash/content.go
new file mode 100644
index 0000000000..5704ea0bc3
--- /dev/null
+++ b/internal/utils/hash/content.go
@@ -0,0 +1,83 @@
+package hash
+
+import (
+	"crypto/sha256"
+	"encoding/hex"
+	"fmt"
+	"regexp"
+	"strings"
+)
+
+// StableChunkID generates a deterministic, content-addressable chunk ID
+// from knowledgeID + normalized content + optional stable sequence.
+// Format: sha256(knowledgeID|normalizedContent|seq)[:16]  (short but collision-resistant for practical use)
+// This replaces uuid.New().String() so that identical content yields identical IDs
+// across reparse / rebuild / crash recovery, enabling cache hits on vectors, wiki refs, graph edges, etc.
+//
+// Normalization rules (to achieve high hit rate):
+//   - Trim whitespace
+//   - Collapse internal whitespace runs to single space
+//   - Lowercase (for case-insensitive matching where appropriate)
+//   - Remove common markdown artifacts that don't change semantics
+//
+// The seq parameter should be the original document order (ChunkIndex or stable position)
+// to disambiguate identical text blocks within the same document.
+func StableChunkID(knowledgeID, content string, seq int) string {
+	normalized := normalizeForID(content)
+	payload := fmt.Sprintf("%s|%s|%d", knowledgeID, normalized, seq)
+	sum := sha256.Sum256([]byte(payload))
+	return hex.EncodeToString(sum[:])[:16] // 16 hex chars = 64-bit, sufficient for per-doc uniqueness
+}
+
+// StableParentChunkID is the same as StableChunkID but marks it as parent for clarity.
+func StableParentChunkID(knowledgeID, content string, seq int) string {
+	return StableChunkID(knowledgeID, content, seq)
+}
+
+// normalizeForID produces a canonical string for hashing.
+// It is intentionally lossy in a controlled way so that semantically identical
+// chunks (minor whitespace, case, markdown noise) produce the same ID.
+func normalizeForID(s string) string {
+	s = strings.TrimSpace(s)
+	// Collapse any run of whitespace (space, tab, newline) into a single space
+	re := regexp.MustCompile(`\s+`)
+	s = re.ReplaceAllString(s, " ")
+	// Lowercase for better reuse across documents that differ only in case
+	s = strings.ToLower(s)
+	return s
+}
+
+// ContentHash produces a full 64-char SHA256 for use as ContentHash field
+// (more collision resistant, used for de-duplication / cache key).
+func ContentHash(content string) string {
+	normalized := normalizeForID(content)
+	sum := sha256.Sum256([]byte(normalized))
+	return hex.EncodeToString(sum[:])
+}
+
+// ImageContentHash produces a stable key for VLM caching: hash(imageBytes) + model + promptVersion
+func ImageContentHash(imageBytes []byte, vlmModelID, promptVersion string) string {
+	h := sha256.New()
+	h.Write(imageBytes)
+	h.Write([]byte("|"))
+	h.Write([]byte(vlmModelID))
+	h.Write([]byte("|"))
+	h.Write([]byte(promptVersion))
+	sum := h.Sum(nil)
+	return hex.EncodeToString(sum[:])
+}
+
+// WikiDocMapKey produces cache key for per-document wiki extract/summary/classify map.
+// Uses frozen (post-VLM) document content hash + granularity + model + prompt version.
+func WikiDocMapKey(docContentHash, granularity, chatModelID, promptVersion string) string {
+	payload := fmt.Sprintf("%s|%s|%s|%s", docContentHash, granularity, chatModelID, promptVersion)
+	sum := sha256.Sum256([]byte(payload))
+	return hex.EncodeToString(sum[:16])
+}
+
+// EmbeddingCacheKey produces key for embedding cache.
+func EmbeddingCacheKey(normalizedText, embeddingModelID string, dim int) string {
+	payload := fmt.Sprintf("%s|%s|%d", normalizedText, embeddingModelID, dim)
+	sum := sha256.Sum256([]byte(payload))
+	return hex.EncodeToString(sum[:16])
+}
\ No newline at end of file