Merge branch 'HKUDS:main' into main

lightrag/__init__.py

@@ -1,5 +1,5 @@
 from .lightrag import LightRAG as LightRAG, QueryParam as QueryParam
 
-__version__ = "1.2.4"
+__version__ = "1.2.5"
 __author__ = "Zirui Guo"
 __url__ = "https://github.com/HKUDS/LightRAG"

lightrag/api/lightrag_server.py

@@ -50,9 +50,6 @@ from .auth import auth_handler
 # This update allows the user to put a different.env file for each lightrag folder
 load_dotenv(".env", override=True)
 
-# Read entity extraction cache config
-enable_llm_cache = os.getenv("ENABLE_LLM_CACHE_FOR_EXTRACT", "false").lower() == "true"
-
 # Initialize config parser
 config = configparser.ConfigParser()
 config.read("config.ini")
@@ -144,23 +141,25 @@ def create_app(args):
         try:
             # Initialize database connections
             await rag.initialize_storages()
-            await initialize_pipeline_status()
 
+            await initialize_pipeline_status()
+            pipeline_status = await get_namespace_data("pipeline_status")
+
+            should_start_autoscan = False
+            async with get_pipeline_status_lock():
                 # Auto scan documents if enabled
                 if args.auto_scan_at_startup:
-                # Check if a task is already running (with lock protection)
+                    if not pipeline_status.get("autoscanned", False):
-                pipeline_status = await get_namespace_data("pipeline_status")
+                        pipeline_status["autoscanned"] = True
-                should_start_task = False
+                        should_start_autoscan = True
-                async with get_pipeline_status_lock():
+
-                    if not pipeline_status.get("busy", False):
+            # Only run auto scan when no other process started it first
-                        should_start_task = True
+            if should_start_autoscan:
-                # Only start the task if no other task is running
-                if should_start_task:
                 # Create background task
                 task = asyncio.create_task(run_scanning_process(rag, doc_manager))
                 app.state.background_tasks.add(task)
                 task.add_done_callback(app.state.background_tasks.discard)
-                    logger.info("Auto scan task started at startup.")
+                logger.info(f"Process {os.getpid()} auto scan task started at startup.")
 
             ASCIIColors.green("\nServer is ready to accept connections! 🚀\n")
 
@@ -326,7 +325,7 @@ def create_app(args):
             vector_db_storage_cls_kwargs={
                 "cosine_better_than_threshold": args.cosine_threshold
             },
-            enable_llm_cache_for_entity_extract=enable_llm_cache,  # Read from environment variable
+            enable_llm_cache_for_entity_extract=args.enable_llm_cache_for_extract,
             embedding_cache_config={
                 "enabled": True,
                 "similarity_threshold": 0.95,
@@ -355,7 +354,7 @@ def create_app(args):
             vector_db_storage_cls_kwargs={
                 "cosine_better_than_threshold": args.cosine_threshold
             },
-            enable_llm_cache_for_entity_extract=enable_llm_cache,  # Read from environment variable
+            enable_llm_cache_for_entity_extract=args.enable_llm_cache_for_extract,
             embedding_cache_config={
                 "enabled": True,
                 "similarity_threshold": 0.95,
@@ -419,6 +418,7 @@ def create_app(args):
                 "doc_status_storage": args.doc_status_storage,
                 "graph_storage": args.graph_storage,
                 "vector_storage": args.vector_storage,
+                "enable_llm_cache_for_extract": args.enable_llm_cache_for_extract,
             },
             "update_status": update_status,
         }

lightrag/api/routers/document_routes.py

@@ -16,7 +16,11 @@ from pydantic import BaseModel, Field, field_validator
 
 from lightrag import LightRAG
 from lightrag.base import DocProcessingStatus, DocStatus
-from ..utils_api import get_api_key_dependency, get_auth_dependency
+from lightrag.api.utils_api import (
+    get_api_key_dependency,
+    global_args,
+    get_auth_dependency,
+)
 
 router = APIRouter(
     prefix="/documents",
@@ -240,6 +244,15 @@ async def pipeline_enqueue_file(rag: LightRAG, file_path: Path) -> bool:
                     )
                     return False
             case ".pdf":
+                if global_args["main_args"].document_loading_engine == "DOCLING":
+                    if not pm.is_installed("docling"):  # type: ignore
+                        pm.install("docling")
+                    from docling.document_converter import DocumentConverter
+
+                    converter = DocumentConverter()
+                    result = converter.convert(file_path)
+                    content = result.document.export_to_markdown()
+                else:
                     if not pm.is_installed("pypdf2"):  # type: ignore
                         pm.install("pypdf2")
                     from PyPDF2 import PdfReader  # type: ignore
@@ -250,6 +263,15 @@ async def pipeline_enqueue_file(rag: LightRAG, file_path: Path) -> bool:
                     for page in reader.pages:
                         content += page.extract_text() + "\n"
             case ".docx":
+                if global_args["main_args"].document_loading_engine == "DOCLING":
+                    if not pm.is_installed("docling"):  # type: ignore
+                        pm.install("docling")
+                    from docling.document_converter import DocumentConverter
+
+                    converter = DocumentConverter()
+                    result = converter.convert(file_path)
+                    content = result.document.export_to_markdown()
+                else:
                     if not pm.is_installed("python-docx"):  # type: ignore
                         pm.install("docx")
                     from docx import Document  # type: ignore
@@ -257,8 +279,19 @@ async def pipeline_enqueue_file(rag: LightRAG, file_path: Path) -> bool:
 
                     docx_file = BytesIO(file)
                     doc = Document(docx_file)
-                content = "\n".join([paragraph.text for paragraph in doc.paragraphs])
+                    content = "\n".join(
+                        [paragraph.text for paragraph in doc.paragraphs]
+                    )
             case ".pptx":
+                if global_args["main_args"].document_loading_engine == "DOCLING":
+                    if not pm.is_installed("docling"):  # type: ignore
+                        pm.install("docling")
+                    from docling.document_converter import DocumentConverter
+
+                    converter = DocumentConverter()
+                    result = converter.convert(file_path)
+                    content = result.document.export_to_markdown()
+                else:
                     if not pm.is_installed("python-pptx"):  # type: ignore
                         pm.install("pptx")
                     from pptx import Presentation  # type: ignore
@@ -271,6 +304,15 @@ async def pipeline_enqueue_file(rag: LightRAG, file_path: Path) -> bool:
                             if hasattr(shape, "text"):
                                 content += shape.text + "\n"
             case ".xlsx":
+                if global_args["main_args"].document_loading_engine == "DOCLING":
+                    if not pm.is_installed("docling"):  # type: ignore
+                        pm.install("docling")
+                    from docling.document_converter import DocumentConverter
+
+                    converter = DocumentConverter()
+                    result = converter.convert(file_path)
+                    content = result.document.export_to_markdown()
+                else:
                     if not pm.is_installed("openpyxl"):  # type: ignore
                         pm.install("openpyxl")
                     from openpyxl import load_workbook  # type: ignore
@@ -283,7 +325,8 @@ async def pipeline_enqueue_file(rag: LightRAG, file_path: Path) -> bool:
                         for row in sheet.iter_rows(values_only=True):
                             content += (
                                 "\t".join(
-                                str(cell) if cell is not None else "" for cell in row
+                                    str(cell) if cell is not None else ""
+                                    for cell in row
                                 )
                                 + "\n"
                             )

lightrag/api/routers/ollama_api.py

@@ -11,7 +11,7 @@ import asyncio
 from ascii_colors import trace_exception
 from lightrag import LightRAG, QueryParam
 from lightrag.utils import encode_string_by_tiktoken
-from ..utils_api import ollama_server_infos
+from lightrag.api.utils_api import ollama_server_infos
 
 
 # query mode according to query prefix (bypass is not LightRAG quer mode)

lightrag/api/utils_api.py

@@ -18,6 +18,8 @@ from .auth import auth_handler
 # Load environment variables
 load_dotenv(override=True)
 
+global_args = {"main_args": None}
+
 
 class OllamaServerInfos:
     # Constants for emulated Ollama model information
@@ -360,8 +362,17 @@ def parse_args(is_uvicorn_mode: bool = False) -> argparse.Namespace:
     args.chunk_size = get_env_value("CHUNK_SIZE", 1200, int)
     args.chunk_overlap_size = get_env_value("CHUNK_OVERLAP_SIZE", 100, int)
 
+    # Inject LLM cache configuration
+    args.enable_llm_cache_for_extract = get_env_value(
+        "ENABLE_LLM_CACHE_FOR_EXTRACT", False, bool
+    )
+
+    # Select Document loading tool (DOCLING, DEFAULT)
+    args.document_loading_engine = get_env_value("DOCUMENT_LOADING_ENGINE", "DEFAULT")
+
     ollama_server_infos.LIGHTRAG_MODEL = args.simulated_model_name
 
+    global_args["main_args"] = args
     return args
 
 
@@ -451,8 +462,10 @@ def display_splash_screen(args: argparse.Namespace) -> None:
     ASCIIColors.yellow(f"{args.history_turns}")
     ASCIIColors.white("    ├─ Cosine Threshold: ", end="")
     ASCIIColors.yellow(f"{args.cosine_threshold}")
-    ASCIIColors.white("    └─ Top-K: ", end="")
+    ASCIIColors.white("    ├─ Top-K: ", end="")
     ASCIIColors.yellow(f"{args.top_k}")
+    ASCIIColors.white("    └─ LLM Cache for Extraction Enabled: ", end="")
+    ASCIIColors.yellow(f"{args.enable_llm_cache_for_extract}")
 
     # System Configuration
     ASCIIColors.magenta("\n💾 Storage Configuration:")

lightrag/base.py

@@ -127,6 +127,30 @@ class BaseVectorStorage(StorageNameSpace, ABC):
     async def delete_entity_relation(self, entity_name: str) -> None:
         """Delete relations for a given entity."""
 
+    @abstractmethod
+    async def get_by_id(self, id: str) -> dict[str, Any] | None:
+        """Get vector data by its ID
+
+        Args:
+            id: The unique identifier of the vector
+
+        Returns:
+            The vector data if found, or None if not found
+        """
+        pass
+
+    @abstractmethod
+    async def get_by_ids(self, ids: list[str]) -> list[dict[str, Any]]:
+        """Get multiple vector data by their IDs
+
+        Args:
+            ids: List of unique identifiers
+
+        Returns:
+            List of vector data objects that were found
+        """
+        pass
+
 
 @dataclass
 class BaseKVStorage(StorageNameSpace, ABC):

lightrag/kg/chroma_impl.py

@@ -271,3 +271,67 @@ class ChromaVectorDBStorage(BaseVectorStorage):
         except Exception as e:
             logger.error(f"Error during prefix search in ChromaDB: {str(e)}")
             raise
+
+    async def get_by_id(self, id: str) -> dict[str, Any] | None:
+        """Get vector data by its ID
+
+        Args:
+            id: The unique identifier of the vector
+
+        Returns:
+            The vector data if found, or None if not found
+        """
+        try:
+            # Query the collection for a single vector by ID
+            result = self._collection.get(
+                ids=[id], include=["metadatas", "embeddings", "documents"]
+            )
+
+            if not result or not result["ids"] or len(result["ids"]) == 0:
+                return None
+
+            # Format the result to match the expected structure
+            return {
+                "id": result["ids"][0],
+                "vector": result["embeddings"][0],
+                "content": result["documents"][0],
+                **result["metadatas"][0],
+            }
+        except Exception as e:
+            logger.error(f"Error retrieving vector data for ID {id}: {e}")
+            return None
+
+    async def get_by_ids(self, ids: list[str]) -> list[dict[str, Any]]:
+        """Get multiple vector data by their IDs
+
+        Args:
+            ids: List of unique identifiers
+
+        Returns:
+            List of vector data objects that were found
+        """
+        if not ids:
+            return []
+
+        try:
+            # Query the collection for multiple vectors by IDs
+            result = self._collection.get(
+                ids=ids, include=["metadatas", "embeddings", "documents"]
+            )
+
+            if not result or not result["ids"] or len(result["ids"]) == 0:
+                return []
+
+            # Format the results to match the expected structure
+            return [
+                {
+                    "id": result["ids"][i],
+                    "vector": result["embeddings"][i],
+                    "content": result["documents"][i],
+                    **result["metadatas"][i],
+                }
+                for i in range(len(result["ids"]))
+            ]
+        except Exception as e:
+            logger.error(f"Error retrieving vector data for IDs {ids}: {e}")
+            return []

lightrag/kg/faiss_impl.py

@@ -394,3 +394,46 @@ class FaissVectorDBStorage(BaseVectorStorage):
 
         logger.debug(f"Found {len(matching_records)} records with prefix '{prefix}'")
         return matching_records
+
+    async def get_by_id(self, id: str) -> dict[str, Any] | None:
+        """Get vector data by its ID
+
+        Args:
+            id: The unique identifier of the vector
+
+        Returns:
+            The vector data if found, or None if not found
+        """
+        # Find the Faiss internal ID for the custom ID
+        fid = self._find_faiss_id_by_custom_id(id)
+        if fid is None:
+            return None
+
+        # Get the metadata for the found ID
+        metadata = self._id_to_meta.get(fid, {})
+        if not metadata:
+            return None
+
+        return {**metadata, "id": metadata.get("__id__")}
+
+    async def get_by_ids(self, ids: list[str]) -> list[dict[str, Any]]:
+        """Get multiple vector data by their IDs
+
+        Args:
+            ids: List of unique identifiers
+
+        Returns:
+            List of vector data objects that were found
+        """
+        if not ids:
+            return []
+
+        results = []
+        for id in ids:
+            fid = self._find_faiss_id_by_custom_id(id)
+            if fid is not None:
+                metadata = self._id_to_meta.get(fid, {})
+                if metadata:
+                    results.append({**metadata, "id": metadata.get("__id__")})
+
+        return results

lightrag/kg/json_doc_status_impl.py

@@ -15,6 +15,10 @@ from lightrag.utils import (
 from .shared_storage import (
     get_namespace_data,
     get_storage_lock,
+    get_data_init_lock,
+    get_update_flag,
+    set_all_update_flags,
+    clear_all_update_flags,
     try_initialize_namespace,
 )
 
@@ -27,20 +31,24 @@ class JsonDocStatusStorage(DocStatusStorage):
     def __post_init__(self):
         working_dir = self.global_config["working_dir"]
         self._file_name = os.path.join(working_dir, f"kv_store_{self.namespace}.json")
-        self._storage_lock = get_storage_lock()
         self._data = None
+        self._storage_lock = None
+        self.storage_updated = None
 
     async def initialize(self):
         """Initialize storage data"""
+        self._storage_lock = get_storage_lock()
+        self.storage_updated = await get_update_flag(self.namespace)
+        async with get_data_init_lock():
             # check need_init must before get_namespace_data
-        need_init = try_initialize_namespace(self.namespace)
+            need_init = await try_initialize_namespace(self.namespace)
             self._data = await get_namespace_data(self.namespace)
             if need_init:
                 loaded_data = load_json(self._file_name) or {}
                 async with self._storage_lock:
                     self._data.update(loaded_data)
                     logger.info(
-                    f"Loaded document status storage with {len(loaded_data)} records"
+                        f"Process {os.getpid()} doc status load {self.namespace} with {len(loaded_data)} records"
                     )
 
     async def filter_keys(self, keys: set[str]) -> set[str]:
@@ -87,18 +95,24 @@ class JsonDocStatusStorage(DocStatusStorage):
 
     async def index_done_callback(self) -> None:
         async with self._storage_lock:
+            if self.storage_updated.value:
                 data_dict = (
                     dict(self._data) if hasattr(self._data, "_getvalue") else self._data
                 )
+                logger.info(
+                    f"Process {os.getpid()} doc status writting {len(data_dict)} records to {self.namespace}"
+                )
                 write_json(data_dict, self._file_name)
+                await clear_all_update_flags(self.namespace)
 
     async def upsert(self, data: dict[str, dict[str, Any]]) -> None:
-        logger.info(f"Inserting {len(data)} to {self.namespace}")
         if not data:
             return
-
+        logger.info(f"Inserting {len(data)} records to {self.namespace}")
         async with self._storage_lock:
             self._data.update(data)
+            await set_all_update_flags(self.namespace)
+
         await self.index_done_callback()
 
     async def get_by_id(self, id: str) -> Union[dict[str, Any], None]:
@@ -109,9 +123,12 @@ class JsonDocStatusStorage(DocStatusStorage):
         async with self._storage_lock:
             for doc_id in doc_ids:
                 self._data.pop(doc_id, None)
+            await set_all_update_flags(self.namespace)
         await self.index_done_callback()
 
     async def drop(self) -> None:
         """Drop the storage"""
         async with self._storage_lock:
             self._data.clear()
+            await set_all_update_flags(self.namespace)
+        await self.index_done_callback()

lightrag/kg/json_kv_impl.py

@@ -13,6 +13,10 @@ from lightrag.utils import (
 from .shared_storage import (
     get_namespace_data,
     get_storage_lock,
+    get_data_init_lock,
+    get_update_flag,
+    set_all_update_flags,
+    clear_all_update_flags,
     try_initialize_namespace,
 )
 
@@ -23,26 +27,63 @@ class JsonKVStorage(BaseKVStorage):
     def __post_init__(self):
         working_dir = self.global_config["working_dir"]
         self._file_name = os.path.join(working_dir, f"kv_store_{self.namespace}.json")
-        self._storage_lock = get_storage_lock()
         self._data = None
+        self._storage_lock = None
+        self.storage_updated = None
 
     async def initialize(self):
         """Initialize storage data"""
+        self._storage_lock = get_storage_lock()
+        self.storage_updated = await get_update_flag(self.namespace)
+        async with get_data_init_lock():
             # check need_init must before get_namespace_data
-        need_init = try_initialize_namespace(self.namespace)
+            need_init = await try_initialize_namespace(self.namespace)
             self._data = await get_namespace_data(self.namespace)
             if need_init:
                 loaded_data = load_json(self._file_name) or {}
                 async with self._storage_lock:
                     self._data.update(loaded_data)
-                logger.info(f"Load KV {self.namespace} with {len(loaded_data)} data")
+
+                    # Calculate data count based on namespace
+                    if self.namespace.endswith("cache"):
+                        # For cache namespaces, sum the cache entries across all cache types
+                        data_count = sum(
+                            len(first_level_dict)
+                            for first_level_dict in loaded_data.values()
+                            if isinstance(first_level_dict, dict)
+                        )
+                    else:
+                        # For non-cache namespaces, use the original count method
+                        data_count = len(loaded_data)
+
+                    logger.info(
+                        f"Process {os.getpid()} KV load {self.namespace} with {data_count} records"
+                    )
 
     async def index_done_callback(self) -> None:
         async with self._storage_lock:
+            if self.storage_updated.value:
                 data_dict = (
                     dict(self._data) if hasattr(self._data, "_getvalue") else self._data
                 )
+
+                # Calculate data count based on namespace
+                if self.namespace.endswith("cache"):
+                    # # For cache namespaces, sum the cache entries across all cache types
+                    data_count = sum(
+                        len(first_level_dict)
+                        for first_level_dict in data_dict.values()
+                        if isinstance(first_level_dict, dict)
+                    )
+                else:
+                    # For non-cache namespaces, use the original count method
+                    data_count = len(data_dict)
+
+                logger.info(
+                    f"Process {os.getpid()} KV writting {data_count} records to {self.namespace}"
+                )
                 write_json(data_dict, self._file_name)
+                await clear_all_update_flags(self.namespace)
 
     async def get_all(self) -> dict[str, Any]:
         """Get all data from storage
@@ -73,15 +114,16 @@ class JsonKVStorage(BaseKVStorage):
             return set(keys) - set(self._data.keys())
 
     async def upsert(self, data: dict[str, dict[str, Any]]) -> None:
-        logger.info(f"Inserting {len(data)} to {self.namespace}")
         if not data:
             return
+        logger.info(f"Inserting {len(data)} records to {self.namespace}")
         async with self._storage_lock:
-            left_data = {k: v for k, v in data.items() if k not in self._data}
+            self._data.update(data)
-            self._data.update(left_data)
+            await set_all_update_flags(self.namespace)
 
     async def delete(self, ids: list[str]) -> None:
         async with self._storage_lock:
             for doc_id in ids:
                 self._data.pop(doc_id, None)
+            await set_all_update_flags(self.namespace)
         await self.index_done_callback()

lightrag/kg/milvus_impl.py

@@ -233,3 +233,57 @@ class MilvusVectorDBStorage(BaseVectorStorage):
         except Exception as e:
             logger.error(f"Error searching for records with prefix '{prefix}': {e}")
             return []
+
+    async def get_by_id(self, id: str) -> dict[str, Any] | None:
+        """Get vector data by its ID
+
+        Args:
+            id: The unique identifier of the vector
+
+        Returns:
+            The vector data if found, or None if not found
+        """
+        try:
+            # Query Milvus for a specific ID
+            result = self._client.query(
+                collection_name=self.namespace,
+                filter=f'id == "{id}"',
+                output_fields=list(self.meta_fields) + ["id"],
+            )
+
+            if not result or len(result) == 0:
+                return None
+
+            return result[0]
+        except Exception as e:
+            logger.error(f"Error retrieving vector data for ID {id}: {e}")
+            return None
+
+    async def get_by_ids(self, ids: list[str]) -> list[dict[str, Any]]:
+        """Get multiple vector data by their IDs
+
+        Args:
+            ids: List of unique identifiers
+
+        Returns:
+            List of vector data objects that were found
+        """
+        if not ids:
+            return []
+
+        try:
+            # Prepare the ID filter expression
+            id_list = '", "'.join(ids)
+            filter_expr = f'id in ["{id_list}"]'
+
+            # Query Milvus with the filter
+            result = self._client.query(
+                collection_name=self.namespace,
+                filter=filter_expr,
+                output_fields=list(self.meta_fields) + ["id"],
+            )
+
+            return result or []
+        except Exception as e:
+            logger.error(f"Error retrieving vector data for IDs {ids}: {e}")
+            return []

lightrag/kg/mongo_impl.py

@@ -1073,6 +1073,59 @@ class MongoVectorDBStorage(BaseVectorStorage):
             logger.error(f"Error searching by prefix in {self.namespace}: {str(e)}")
             return []
 
+    async def get_by_id(self, id: str) -> dict[str, Any] | None:
+        """Get vector data by its ID
+
+        Args:
+            id: The unique identifier of the vector
+
+        Returns:
+            The vector data if found, or None if not found
+        """
+        try:
+            # Search for the specific ID in MongoDB
+            result = await self._data.find_one({"_id": id})
+            if result:
+                # Format the result to include id field expected by API
+                result_dict = dict(result)
+                if "_id" in result_dict and "id" not in result_dict:
+                    result_dict["id"] = result_dict["_id"]
+                return result_dict
+            return None
+        except Exception as e:
+            logger.error(f"Error retrieving vector data for ID {id}: {e}")
+            return None
+
+    async def get_by_ids(self, ids: list[str]) -> list[dict[str, Any]]:
+        """Get multiple vector data by their IDs
+
+        Args:
+            ids: List of unique identifiers
+
+        Returns:
+            List of vector data objects that were found
+        """
+        if not ids:
+            return []
+
+        try:
+            # Query MongoDB for multiple IDs
+            cursor = self._data.find({"_id": {"$in": ids}})
+            results = await cursor.to_list(length=None)
+
+            # Format results to include id field expected by API
+            formatted_results = []
+            for result in results:
+                result_dict = dict(result)
+                if "_id" in result_dict and "id" not in result_dict:
+                    result_dict["id"] = result_dict["_id"]
+                formatted_results.append(result_dict)
+
+            return formatted_results
+        except Exception as e:
+            logger.error(f"Error retrieving vector data for IDs {ids}: {e}")
+            return []
+
 
 async def get_or_create_collection(db: AsyncIOMotorDatabase, collection_name: str):
     collection_names = await db.list_collection_names()

lightrag/kg/nano_vector_db_impl.py

@@ -258,3 +258,33 @@ class NanoVectorDBStorage(BaseVectorStorage):
 
         logger.debug(f"Found {len(matching_records)} records with prefix '{prefix}'")
         return matching_records
+
+    async def get_by_id(self, id: str) -> dict[str, Any] | None:
+        """Get vector data by its ID
+
+        Args:
+            id: The unique identifier of the vector
+
+        Returns:
+            The vector data if found, or None if not found
+        """
+        client = await self._get_client()
+        result = client.get([id])
+        if result:
+            return result[0]
+        return None
+
+    async def get_by_ids(self, ids: list[str]) -> list[dict[str, Any]]:
+        """Get multiple vector data by their IDs
+
+        Args:
+            ids: List of unique identifiers
+
+        Returns:
+            List of vector data objects that were found
+        """
+        if not ids:
+            return []
+
+        client = await self._get_client()
+        return client.get(ids)

lightrag/kg/oracle_impl.py

@@ -531,6 +531,80 @@ class OracleVectorDBStorage(BaseVectorStorage):
             logger.error(f"Error searching records with prefix '{prefix}': {e}")
             return []
 
+    async def get_by_id(self, id: str) -> dict[str, Any] | None:
+        """Get vector data by its ID
+
+        Args:
+            id: The unique identifier of the vector
+
+        Returns:
+            The vector data if found, or None if not found
+        """
+        try:
+            # Determine the table name based on namespace
+            table_name = namespace_to_table_name(self.namespace)
+            if not table_name:
+                logger.error(f"Unknown namespace for ID lookup: {self.namespace}")
+                return None
+
+            # Create the appropriate ID field name based on namespace
+            id_field = "entity_id" if "NODES" in table_name else "relation_id"
+            if "CHUNKS" in table_name:
+                id_field = "chunk_id"
+
+            # Prepare and execute the query
+            query = f"""
+                SELECT * FROM {table_name}
+                WHERE {id_field} = :id AND workspace = :workspace
+            """
+            params = {"id": id, "workspace": self.db.workspace}
+
+            result = await self.db.query(query, params)
+            return result
+        except Exception as e:
+            logger.error(f"Error retrieving vector data for ID {id}: {e}")
+            return None
+
+    async def get_by_ids(self, ids: list[str]) -> list[dict[str, Any]]:
+        """Get multiple vector data by their IDs
+
+        Args:
+            ids: List of unique identifiers
+
+        Returns:
+            List of vector data objects that were found
+        """
+        if not ids:
+            return []
+
+        try:
+            # Determine the table name based on namespace
+            table_name = namespace_to_table_name(self.namespace)
+            if not table_name:
+                logger.error(f"Unknown namespace for IDs lookup: {self.namespace}")
+                return []
+
+            # Create the appropriate ID field name based on namespace
+            id_field = "entity_id" if "NODES" in table_name else "relation_id"
+            if "CHUNKS" in table_name:
+                id_field = "chunk_id"
+
+            # Format the list of IDs for SQL IN clause
+            ids_list = ", ".join([f"'{id}'" for id in ids])
+
+            # Prepare and execute the query
+            query = f"""
+                SELECT * FROM {table_name}
+                WHERE {id_field} IN ({ids_list}) AND workspace = :workspace
+            """
+            params = {"workspace": self.db.workspace}
+
+            results = await self.db.query(query, params, multirows=True)
+            return results or []
+        except Exception as e:
+            logger.error(f"Error retrieving vector data for IDs {ids}: {e}")
+            return []
+
 
 @final
 @dataclass

lightrag/kg/postgres_impl.py

@@ -621,6 +621,60 @@ class PGVectorStorage(BaseVectorStorage):
             logger.error(f"Error during prefix search for '{prefix}': {e}")
             return []
 
+    async def get_by_id(self, id: str) -> dict[str, Any] | None:
+        """Get vector data by its ID
+
+        Args:
+            id: The unique identifier of the vector
+
+        Returns:
+            The vector data if found, or None if not found
+        """
+        table_name = namespace_to_table_name(self.namespace)
+        if not table_name:
+            logger.error(f"Unknown namespace for ID lookup: {self.namespace}")
+            return None
+
+        query = f"SELECT * FROM {table_name} WHERE workspace=$1 AND id=$2"
+        params = {"workspace": self.db.workspace, "id": id}
+
+        try:
+            result = await self.db.query(query, params)
+            if result:
+                return dict(result)
+            return None
+        except Exception as e:
+            logger.error(f"Error retrieving vector data for ID {id}: {e}")
+            return None
+
+    async def get_by_ids(self, ids: list[str]) -> list[dict[str, Any]]:
+        """Get multiple vector data by their IDs
+
+        Args:
+            ids: List of unique identifiers
+
+        Returns:
+            List of vector data objects that were found
+        """
+        if not ids:
+            return []
+
+        table_name = namespace_to_table_name(self.namespace)
+        if not table_name:
+            logger.error(f"Unknown namespace for IDs lookup: {self.namespace}")
+            return []
+
+        ids_str = ",".join([f"'{id}'" for id in ids])
+        query = f"SELECT * FROM {table_name} WHERE workspace=$1 AND id IN ({ids_str})"
+        params = {"workspace": self.db.workspace}
+
+        try:
+            results = await self.db.query(query, params, multirows=True)
+            return [dict(record) for record in results]
+        except Exception as e:
+            logger.error(f"Error retrieving vector data for IDs {ids}: {e}")
+            return []
+
 
 @final
 @dataclass

lightrag/kg/shared_storage.py

@@ -7,11 +7,17 @@ from typing import Any, Dict, Optional, Union, TypeVar, Generic
 
 
 # Define a direct print function for critical logs that must be visible in all processes
-def direct_log(message, level="INFO"):
+def direct_log(message, level="INFO", enable_output: bool = True):
     """
     Log a message directly to stderr to ensure visibility in all processes,
     including the Gunicorn master process.
+
+    Args:
+        message: The message to log
+        level: Log level (default: "INFO")
+        enable_output: Whether to actually output the log (default: True)
     """
+    if enable_output:
         print(f"{level}: {message}", file=sys.stderr, flush=True)
 
 
@@ -32,55 +38,165 @@ _update_flags: Optional[Dict[str, bool]] = None  # namespace -> updated
 _storage_lock: Optional[LockType] = None
 _internal_lock: Optional[LockType] = None
 _pipeline_status_lock: Optional[LockType] = None
+_graph_db_lock: Optional[LockType] = None
+_data_init_lock: Optional[LockType] = None
 
 
 class UnifiedLock(Generic[T]):
     """Provide a unified lock interface type for asyncio.Lock and multiprocessing.Lock"""
 
-    def __init__(self, lock: Union[ProcessLock, asyncio.Lock], is_async: bool):
+    def __init__(
+        self,
+        lock: Union[ProcessLock, asyncio.Lock],
+        is_async: bool,
+        name: str = "unnamed",
+        enable_logging: bool = True,
+    ):
         self._lock = lock
         self._is_async = is_async
+        self._pid = os.getpid()  # for debug only
+        self._name = name  # for debug only
+        self._enable_logging = enable_logging  # for debug only
 
     async def __aenter__(self) -> "UnifiedLock[T]":
+        try:
+            direct_log(
+                f"== Lock == Process {self._pid}: Acquiring lock '{self._name}' (async={self._is_async})",
+                enable_output=self._enable_logging,
+            )
             if self._is_async:
                 await self._lock.acquire()
             else:
                 self._lock.acquire()
+            direct_log(
+                f"== Lock == Process {self._pid}: Lock '{self._name}' acquired (async={self._is_async})",
+                enable_output=self._enable_logging,
+            )
             return self
+        except Exception as e:
+            direct_log(
+                f"== Lock == Process {self._pid}: Failed to acquire lock '{self._name}': {e}",
+                level="ERROR",
+                enable_output=self._enable_logging,
+            )
+            raise
 
     async def __aexit__(self, exc_type, exc_val, exc_tb):
+        try:
+            direct_log(
+                f"== Lock == Process {self._pid}: Releasing lock '{self._name}' (async={self._is_async})",
+                enable_output=self._enable_logging,
+            )
             if self._is_async:
                 self._lock.release()
             else:
                 self._lock.release()
+            direct_log(
+                f"== Lock == Process {self._pid}: Lock '{self._name}' released (async={self._is_async})",
+                enable_output=self._enable_logging,
+            )
+        except Exception as e:
+            direct_log(
+                f"== Lock == Process {self._pid}: Failed to release lock '{self._name}': {e}",
+                level="ERROR",
+                enable_output=self._enable_logging,
+            )
+            raise
 
     def __enter__(self) -> "UnifiedLock[T]":
         """For backward compatibility"""
+        try:
             if self._is_async:
                 raise RuntimeError("Use 'async with' for shared_storage lock")
+            direct_log(
+                f"== Lock == Process {self._pid}: Acquiring lock '{self._name}' (sync)",
+                enable_output=self._enable_logging,
+            )
             self._lock.acquire()
+            direct_log(
+                f"== Lock == Process {self._pid}: Lock '{self._name}' acquired (sync)",
+                enable_output=self._enable_logging,
+            )
             return self
+        except Exception as e:
+            direct_log(
+                f"== Lock == Process {self._pid}: Failed to acquire lock '{self._name}' (sync): {e}",
+                level="ERROR",
+                enable_output=self._enable_logging,
+            )
+            raise
 
     def __exit__(self, exc_type, exc_val, exc_tb):
         """For backward compatibility"""
+        try:
             if self._is_async:
                 raise RuntimeError("Use 'async with' for shared_storage lock")
+            direct_log(
+                f"== Lock == Process {self._pid}: Releasing lock '{self._name}' (sync)",
+                enable_output=self._enable_logging,
+            )
             self._lock.release()
+            direct_log(
+                f"== Lock == Process {self._pid}: Lock '{self._name}' released (sync)",
+                enable_output=self._enable_logging,
+            )
+        except Exception as e:
+            direct_log(
+                f"== Lock == Process {self._pid}: Failed to release lock '{self._name}' (sync): {e}",
+                level="ERROR",
+                enable_output=self._enable_logging,
+            )
+            raise
 
 
-def get_internal_lock() -> UnifiedLock:
+def get_internal_lock(enable_logging: bool = False) -> UnifiedLock:
     """return unified storage lock for data consistency"""
-    return UnifiedLock(lock=_internal_lock, is_async=not is_multiprocess)
+    return UnifiedLock(
+        lock=_internal_lock,
+        is_async=not is_multiprocess,
+        name="internal_lock",
+        enable_logging=enable_logging,
+    )
 
 
-def get_storage_lock() -> UnifiedLock:
+def get_storage_lock(enable_logging: bool = False) -> UnifiedLock:
     """return unified storage lock for data consistency"""
-    return UnifiedLock(lock=_storage_lock, is_async=not is_multiprocess)
+    return UnifiedLock(
+        lock=_storage_lock,
+        is_async=not is_multiprocess,
+        name="storage_lock",
+        enable_logging=enable_logging,
+    )
 
 
-def get_pipeline_status_lock() -> UnifiedLock:
+def get_pipeline_status_lock(enable_logging: bool = False) -> UnifiedLock:
     """return unified storage lock for data consistency"""
-    return UnifiedLock(lock=_pipeline_status_lock, is_async=not is_multiprocess)
+    return UnifiedLock(
+        lock=_pipeline_status_lock,
+        is_async=not is_multiprocess,
+        name="pipeline_status_lock",
+        enable_logging=enable_logging,
+    )
+
+
+def get_graph_db_lock(enable_logging: bool = False) -> UnifiedLock:
+    """return unified graph database lock for ensuring atomic operations"""
+    return UnifiedLock(
+        lock=_graph_db_lock,
+        is_async=not is_multiprocess,
+        name="graph_db_lock",
+        enable_logging=enable_logging,
+    )
+
+
+def get_data_init_lock(enable_logging: bool = False) -> UnifiedLock:
+    """return unified data initialization lock for ensuring atomic data initialization"""
+    return UnifiedLock(
+        lock=_data_init_lock,
+        is_async=not is_multiprocess,
+        name="data_init_lock",
+        enable_logging=enable_logging,
+    )
 
 
 def initialize_share_data(workers: int = 1):
@@ -108,6 +224,8 @@ def initialize_share_data(workers: int = 1):
         _storage_lock, \
         _internal_lock, \
         _pipeline_status_lock, \
+        _graph_db_lock, \
+        _data_init_lock, \
         _shared_dicts, \
         _init_flags, \
         _initialized, \
@@ -120,14 +238,16 @@ def initialize_share_data(workers: int = 1):
         )
         return
 
-    _manager = Manager()
     _workers = workers
 
     if workers > 1:
         is_multiprocess = True
+        _manager = Manager()
         _internal_lock = _manager.Lock()
         _storage_lock = _manager.Lock()
         _pipeline_status_lock = _manager.Lock()
+        _graph_db_lock = _manager.Lock()
+        _data_init_lock = _manager.Lock()
         _shared_dicts = _manager.dict()
         _init_flags = _manager.dict()
         _update_flags = _manager.dict()
@@ -139,6 +259,8 @@ def initialize_share_data(workers: int = 1):
         _internal_lock = asyncio.Lock()
         _storage_lock = asyncio.Lock()
         _pipeline_status_lock = asyncio.Lock()
+        _graph_db_lock = asyncio.Lock()
+        _data_init_lock = asyncio.Lock()
         _shared_dicts = {}
         _init_flags = {}
         _update_flags = {}
@@ -164,6 +286,7 @@ async def initialize_pipeline_status():
         history_messages = _manager.list() if is_multiprocess else []
         pipeline_namespace.update(
             {
+                "autoscanned": False,  # Auto-scan started
                 "busy": False,  # Control concurrent processes
                 "job_name": "Default Job",  # Current job name (indexing files/indexing texts)
                 "job_start": None,  # Job start time
@@ -200,7 +323,12 @@ async def get_update_flag(namespace: str):
         if is_multiprocess and _manager is not None:
             new_update_flag = _manager.Value("b", False)
         else:
-            new_update_flag = False
+            # Create a simple mutable object to store boolean value for compatibility with mutiprocess
+            class MutableBoolean:
+                def __init__(self, initial_value=False):
+                    self.value = initial_value
+
+            new_update_flag = MutableBoolean(False)
 
         _update_flags[namespace].append(new_update_flag)
         return new_update_flag
@@ -220,7 +348,26 @@ async def set_all_update_flags(namespace: str):
             if is_multiprocess:
                 _update_flags[namespace][i].value = True
             else:
-                _update_flags[namespace][i] = True
+                # Use .value attribute instead of direct assignment
+                _update_flags[namespace][i].value = True
+
+
+async def clear_all_update_flags(namespace: str):
+    """Clear all update flag of namespace indicating all workers need to reload data from files"""
+    global _update_flags
+    if _update_flags is None:
+        raise ValueError("Try to create namespace before Shared-Data is initialized")
+
+    async with get_internal_lock():
+        if namespace not in _update_flags:
+            raise ValueError(f"Namespace {namespace} not found in update flags")
+        # Update flags for both modes
+        for i in range(len(_update_flags[namespace])):
+            if is_multiprocess:
+                _update_flags[namespace][i].value = False
+            else:
+                # Use .value attribute instead of direct assignment
+                _update_flags[namespace][i].value = False
 
 
 async def get_all_update_flags_status() -> Dict[str, list]:
@@ -247,7 +394,7 @@ async def get_all_update_flags_status() -> Dict[str, list]:
     return result
 
 
-def try_initialize_namespace(namespace: str) -> bool:
+async def try_initialize_namespace(namespace: str) -> bool:
     """
     Returns True if the current worker(process) gets initialization permission for loading data later.
     The worker does not get the permission is prohibited to load data from files.
@@ -257,6 +404,7 @@ def try_initialize_namespace(namespace: str) -> bool:
     if _init_flags is None:
         raise ValueError("Try to create nanmespace before Shared-Data is initialized")
 
+    async with get_internal_lock():
         if namespace not in _init_flags:
             _init_flags[namespace] = True
             direct_log(
@@ -266,6 +414,7 @@ def try_initialize_namespace(namespace: str) -> bool:
         direct_log(
             f"Process {os.getpid()} storage namespace already initialized: [{namespace}]"
         )
+
     return False
 
 
@@ -304,6 +453,8 @@ def finalize_share_data():
         _storage_lock, \
         _internal_lock, \
         _pipeline_status_lock, \
+        _graph_db_lock, \
+        _data_init_lock, \
         _shared_dicts, \
         _init_flags, \
         _initialized, \
@@ -369,6 +520,8 @@ def finalize_share_data():
     _storage_lock = None
     _internal_lock = None
     _pipeline_status_lock = None
+    _graph_db_lock = None
+    _data_init_lock = None
     _update_flags = None
 
     direct_log(f"Process {os.getpid()} storage data finalization complete")

lightrag/kg/tidb_impl.py

@@ -465,6 +465,100 @@ class TiDBVectorDBStorage(BaseVectorStorage):
             logger.error(f"Error searching records with prefix '{prefix}': {e}")
             return []
 
+    async def get_by_id(self, id: str) -> dict[str, Any] | None:
+        """Get vector data by its ID
+
+        Args:
+            id: The unique identifier of the vector
+
+        Returns:
+            The vector data if found, or None if not found
+        """
+        try:
+            # Determine which table to query based on namespace
+            if self.namespace == NameSpace.VECTOR_STORE_ENTITIES:
+                sql_template = """
+                    SELECT entity_id as id, name as entity_name, entity_type, description, content
+                    FROM LIGHTRAG_GRAPH_NODES
+                    WHERE entity_id = :entity_id AND workspace = :workspace
+                """
+                params = {"entity_id": id, "workspace": self.db.workspace}
+            elif self.namespace == NameSpace.VECTOR_STORE_RELATIONSHIPS:
+                sql_template = """
+                    SELECT relation_id as id, source_name as src_id, target_name as tgt_id,
+                           keywords, description, content
+                    FROM LIGHTRAG_GRAPH_EDGES
+                    WHERE relation_id = :relation_id AND workspace = :workspace
+                """
+                params = {"relation_id": id, "workspace": self.db.workspace}
+            elif self.namespace == NameSpace.VECTOR_STORE_CHUNKS:
+                sql_template = """
+                    SELECT chunk_id as id, content, tokens, chunk_order_index, full_doc_id
+                    FROM LIGHTRAG_DOC_CHUNKS
+                    WHERE chunk_id = :chunk_id AND workspace = :workspace
+                """
+                params = {"chunk_id": id, "workspace": self.db.workspace}
+            else:
+                logger.warning(
+                    f"Namespace {self.namespace} not supported for get_by_id"
+                )
+                return None
+
+            result = await self.db.query(sql_template, params=params)
+            return result
+        except Exception as e:
+            logger.error(f"Error retrieving vector data for ID {id}: {e}")
+            return None
+
+    async def get_by_ids(self, ids: list[str]) -> list[dict[str, Any]]:
+        """Get multiple vector data by their IDs
+
+        Args:
+            ids: List of unique identifiers
+
+        Returns:
+            List of vector data objects that were found
+        """
+        if not ids:
+            return []
+
+        try:
+            # Format IDs for SQL IN clause
+            ids_str = ", ".join([f"'{id}'" for id in ids])
+
+            # Determine which table to query based on namespace
+            if self.namespace == NameSpace.VECTOR_STORE_ENTITIES:
+                sql_template = f"""
+                    SELECT entity_id as id, name as entity_name, entity_type, description, content
+                    FROM LIGHTRAG_GRAPH_NODES
+                    WHERE entity_id IN ({ids_str}) AND workspace = :workspace
+                """
+            elif self.namespace == NameSpace.VECTOR_STORE_RELATIONSHIPS:
+                sql_template = f"""
+                    SELECT relation_id as id, source_name as src_id, target_name as tgt_id,
+                           keywords, description, content
+                    FROM LIGHTRAG_GRAPH_EDGES
+                    WHERE relation_id IN ({ids_str}) AND workspace = :workspace
+                """
+            elif self.namespace == NameSpace.VECTOR_STORE_CHUNKS:
+                sql_template = f"""
+                    SELECT chunk_id as id, content, tokens, chunk_order_index, full_doc_id
+                    FROM LIGHTRAG_DOC_CHUNKS
+                    WHERE chunk_id IN ({ids_str}) AND workspace = :workspace
+                """
+            else:
+                logger.warning(
+                    f"Namespace {self.namespace} not supported for get_by_ids"
+                )
+                return []
+
+            params = {"workspace": self.db.workspace}
+            results = await self.db.query(sql_template, params=params, multirows=True)
+            return results if results else []
+        except Exception as e:
+            logger.error(f"Error retrieving vector data for IDs {ids}: {e}")
+            return []
+
 
 @final
 @dataclass

lightrag/lightrag.py

@@ -30,11 +30,10 @@ from .namespace import NameSpace, make_namespace
 from .operate import (
     chunking_by_token_size,
     extract_entities,
-    extract_keywords_only,
     kg_query,
-    kg_query_with_keywords,
     mix_kg_vector_query,
     naive_query,
+    query_with_keywords,
 )
 from .prompt import GRAPH_FIELD_SEP, PROMPTS
 from .utils import (
@@ -45,6 +44,9 @@ from .utils import (
     encode_string_by_tiktoken,
     lazy_external_import,
     limit_async_func_call,
+    get_content_summary,
+    clean_text,
+    check_storage_env_vars,
     logger,
 )
 from .types import KnowledgeGraph
@@ -309,7 +311,7 @@ class LightRAG:
             # Verify storage implementation compatibility
             verify_storage_implementation(storage_type, storage_name)
             # Check environment variables
-            # self.check_storage_env_vars(storage_name)
+            check_storage_env_vars(storage_name)
 
         # Ensure vector_db_storage_cls_kwargs has required fields
         self.vector_db_storage_cls_kwargs = {
@@ -354,6 +356,9 @@ class LightRAG:
             namespace=make_namespace(
                 self.namespace_prefix, NameSpace.KV_STORE_LLM_RESPONSE_CACHE
             ),
+            global_config=asdict(
+                self
+            ),  # Add global_config to ensure cache works properly
             embedding_func=self.embedding_func,
         )
 
@@ -404,18 +409,8 @@ class LightRAG:
             embedding_func=None,
         )
 
-        if self.llm_response_cache and hasattr(
+        # Directly use llm_response_cache, don't create a new object
-            self.llm_response_cache, "global_config"
-        ):
         hashing_kv = self.llm_response_cache
-        else:
-            hashing_kv = self.key_string_value_json_storage_cls(  # type: ignore
-                namespace=make_namespace(
-                    self.namespace_prefix, NameSpace.KV_STORE_LLM_RESPONSE_CACHE
-                ),
-                global_config=asdict(self),
-                embedding_func=self.embedding_func,
-            )
 
         self.llm_model_func = limit_async_func_call(self.llm_model_max_async)(
             partial(
@@ -543,11 +538,6 @@ class LightRAG:
         storage_class = lazy_external_import(import_path, storage_name)
         return storage_class
 
-    @staticmethod
-    def clean_text(text: str) -> str:
-        """Clean text by removing null bytes (0x00) and whitespace"""
-        return text.strip().replace("\x00", "")
-
     def insert(
         self,
         input: str | list[str],
@@ -590,6 +580,7 @@ class LightRAG:
             split_by_character, split_by_character_only
         )
 
+    # TODO: deprecated, use insert instead
     def insert_custom_chunks(
         self,
         full_text: str,
@@ -601,14 +592,15 @@ class LightRAG:
             self.ainsert_custom_chunks(full_text, text_chunks, doc_id)
         )
 
+    # TODO: deprecated, use ainsert instead
     async def ainsert_custom_chunks(
         self, full_text: str, text_chunks: list[str], doc_id: str | None = None
     ) -> None:
         update_storage = False
         try:
             # Clean input texts
-            full_text = self.clean_text(full_text)
+            full_text = clean_text(full_text)
-            text_chunks = [self.clean_text(chunk) for chunk in text_chunks]
+            text_chunks = [clean_text(chunk) for chunk in text_chunks]
 
             # Process cleaned texts
             if doc_id is None:
@@ -687,7 +679,7 @@ class LightRAG:
             contents = {id_: doc for id_, doc in zip(ids, input)}
         else:
             # Clean input text and remove duplicates
-            input = list(set(self.clean_text(doc) for doc in input))
+            input = list(set(clean_text(doc) for doc in input))
             # Generate contents dict of MD5 hash IDs and documents
             contents = {compute_mdhash_id(doc, prefix="doc-"): doc for doc in input}
 
@@ -703,7 +695,7 @@ class LightRAG:
         new_docs: dict[str, Any] = {
             id_: {
                 "content": content,
-                "content_summary": self._get_content_summary(content),
+                "content_summary": get_content_summary(content),
                 "content_length": len(content),
                 "status": DocStatus.PENDING,
                 "created_at": datetime.now().isoformat(),
@@ -892,7 +884,9 @@ class LightRAG:
                                 self.chunks_vdb.upsert(chunks)
                             )
                             entity_relation_task = asyncio.create_task(
-                                self._process_entity_relation_graph(chunks)
+                                self._process_entity_relation_graph(
+                                    chunks, pipeline_status, pipeline_status_lock
+                                )
                             )
                             full_docs_task = asyncio.create_task(
                                 self.full_docs.upsert(
@@ -1007,21 +1001,27 @@ class LightRAG:
                 pipeline_status["latest_message"] = log_message
                 pipeline_status["history_messages"].append(log_message)
 
-    async def _process_entity_relation_graph(self, chunk: dict[str, Any]) -> None:
+    async def _process_entity_relation_graph(
+        self, chunk: dict[str, Any], pipeline_status=None, pipeline_status_lock=None
+    ) -> None:
         try:
             await extract_entities(
                 chunk,
                 knowledge_graph_inst=self.chunk_entity_relation_graph,
                 entity_vdb=self.entities_vdb,
                 relationships_vdb=self.relationships_vdb,
-                llm_response_cache=self.llm_response_cache,
                 global_config=asdict(self),
+                pipeline_status=pipeline_status,
+                pipeline_status_lock=pipeline_status_lock,
+                llm_response_cache=self.llm_response_cache,
             )
         except Exception as e:
             logger.error("Failed to extract entities and relationships")
             raise e
 
-    async def _insert_done(self) -> None:
+    async def _insert_done(
+        self, pipeline_status=None, pipeline_status_lock=None
+    ) -> None:
         tasks = [
             cast(StorageNameSpace, storage_inst).index_done_callback()
             for storage_inst in [  # type: ignore
@@ -1040,10 +1040,8 @@ class LightRAG:
         log_message = "All Insert done"
         logger.info(log_message)
 
-        # 获取 pipeline_status 并更新 latest_message 和 history_messages
+        if pipeline_status is not None and pipeline_status_lock is not None:
-        from lightrag.kg.shared_storage import get_namespace_data
+            async with pipeline_status_lock:
-
-        pipeline_status = await get_namespace_data("pipeline_status")
                 pipeline_status["latest_message"] = log_message
                 pipeline_status["history_messages"].append(log_message)
 
@@ -1062,7 +1060,7 @@ class LightRAG:
             all_chunks_data: dict[str, dict[str, str]] = {}
             chunk_to_source_map: dict[str, str] = {}
             for chunk_data in custom_kg.get("chunks", []):
-                chunk_content = self.clean_text(chunk_data["content"])
+                chunk_content = clean_text(chunk_data["content"])
                 source_id = chunk_data["source_id"]
                 tokens = len(
                     encode_string_by_tiktoken(
@@ -1260,16 +1258,7 @@ class LightRAG:
                 self.text_chunks,
                 param,
                 asdict(self),
-                hashing_kv=self.llm_response_cache
+                hashing_kv=self.llm_response_cache,  # Directly use llm_response_cache
-                if self.llm_response_cache
-                and hasattr(self.llm_response_cache, "global_config")
-                else self.key_string_value_json_storage_cls(
-                    namespace=make_namespace(
-                        self.namespace_prefix, NameSpace.KV_STORE_LLM_RESPONSE_CACHE
-                    ),
-                    global_config=asdict(self),
-                    embedding_func=self.embedding_func,
-                ),
                 system_prompt=system_prompt,
             )
         elif param.mode == "naive":
@@ -1279,16 +1268,7 @@ class LightRAG:
                 self.text_chunks,
                 param,
                 asdict(self),
-                hashing_kv=self.llm_response_cache
+                hashing_kv=self.llm_response_cache,  # Directly use llm_response_cache
-                if self.llm_response_cache
-                and hasattr(self.llm_response_cache, "global_config")
-                else self.key_string_value_json_storage_cls(
-                    namespace=make_namespace(
-                        self.namespace_prefix, NameSpace.KV_STORE_LLM_RESPONSE_CACHE
-                    ),
-                    global_config=asdict(self),
-                    embedding_func=self.embedding_func,
-                ),
                 system_prompt=system_prompt,
             )
         elif param.mode == "mix":
@@ -1301,16 +1281,7 @@ class LightRAG:
                 self.text_chunks,
                 param,
                 asdict(self),
-                hashing_kv=self.llm_response_cache
+                hashing_kv=self.llm_response_cache,  # Directly use llm_response_cache
-                if self.llm_response_cache
-                and hasattr(self.llm_response_cache, "global_config")
-                else self.key_string_value_json_storage_cls(
-                    namespace=make_namespace(
-                        self.namespace_prefix, NameSpace.KV_STORE_LLM_RESPONSE_CACHE
-                    ),
-                    global_config=asdict(self),
-                    embedding_func=self.embedding_func,
-                ),
                 system_prompt=system_prompt,
             )
         else:
@@ -1322,8 +1293,17 @@ class LightRAG:
         self, query: str, prompt: str, param: QueryParam = QueryParam()
     ):
         """
-        1. Extract keywords from the 'query' using new function in operate.py.
+        Query with separate keyword extraction step.
-        2. Then run the standard aquery() flow with the final prompt (formatted_question).
+
+        This method extracts keywords from the query first, then uses them for the query.
+
+        Args:
+            query: User query
+            prompt: Additional prompt for the query
+            param: Query parameters
+
+        Returns:
+            Query response
         """
         loop = always_get_an_event_loop()
         return loop.run_until_complete(
@@ -1334,100 +1314,29 @@ class LightRAG:
         self, query: str, prompt: str, param: QueryParam = QueryParam()
     ) -> str | AsyncIterator[str]:
         """
-        1. Calls extract_keywords_only to get HL/LL keywords from 'query'.
+        Async version of query_with_separate_keyword_extraction.
-        2. Then calls kg_query(...) or naive_query(...), etc. as the main query, while also injecting the newly extracted keywords if needed.
+
+        Args:
+            query: User query
+            prompt: Additional prompt for the query
+            param: Query parameters
+
+        Returns:
+            Query response or async iterator
         """
-        # ---------------------
+        response = await query_with_keywords(
-        # STEP 1: Keyword Extraction
+            query=query,
-        # ---------------------
+            prompt=prompt,
-        hl_keywords, ll_keywords = await extract_keywords_only(
-            text=query,
             param=param,
+            knowledge_graph_inst=self.chunk_entity_relation_graph,
+            entities_vdb=self.entities_vdb,
+            relationships_vdb=self.relationships_vdb,
+            chunks_vdb=self.chunks_vdb,
+            text_chunks_db=self.text_chunks,
             global_config=asdict(self),
-            hashing_kv=self.llm_response_cache
+            hashing_kv=self.llm_response_cache,
-            or self.key_string_value_json_storage_cls(
-                namespace=make_namespace(
-                    self.namespace_prefix, NameSpace.KV_STORE_LLM_RESPONSE_CACHE
-                ),
-                global_config=asdict(self),
-                embedding_func=self.embedding_func,
-            ),
         )
 
-        param.hl_keywords = hl_keywords
-        param.ll_keywords = ll_keywords
-
-        # ---------------------
-        # STEP 2: Final Query Logic
-        # ---------------------
-
-        # Create a new string with the prompt and the keywords
-        ll_keywords_str = ", ".join(ll_keywords)
-        hl_keywords_str = ", ".join(hl_keywords)
-        formatted_question = f"{prompt}\n\n### Keywords:\nHigh-level: {hl_keywords_str}\nLow-level: {ll_keywords_str}\n\n### Query:\n{query}"
-
-        if param.mode in ["local", "global", "hybrid"]:
-            response = await kg_query_with_keywords(
-                formatted_question,
-                self.chunk_entity_relation_graph,
-                self.entities_vdb,
-                self.relationships_vdb,
-                self.text_chunks,
-                param,
-                asdict(self),
-                hashing_kv=self.llm_response_cache
-                if self.llm_response_cache
-                and hasattr(self.llm_response_cache, "global_config")
-                else self.key_string_value_json_storage_cls(
-                    namespace=make_namespace(
-                        self.namespace_prefix, NameSpace.KV_STORE_LLM_RESPONSE_CACHE
-                    ),
-                    global_config=asdict(self),
-                    embedding_func=self.embedding_func,
-                ),
-            )
-        elif param.mode == "naive":
-            response = await naive_query(
-                formatted_question,
-                self.chunks_vdb,
-                self.text_chunks,
-                param,
-                asdict(self),
-                hashing_kv=self.llm_response_cache
-                if self.llm_response_cache
-                and hasattr(self.llm_response_cache, "global_config")
-                else self.key_string_value_json_storage_cls(
-                    namespace=make_namespace(
-                        self.namespace_prefix, NameSpace.KV_STORE_LLM_RESPONSE_CACHE
-                    ),
-                    global_config=asdict(self),
-                    embedding_func=self.embedding_func,
-                ),
-            )
-        elif param.mode == "mix":
-            response = await mix_kg_vector_query(
-                formatted_question,
-                self.chunk_entity_relation_graph,
-                self.entities_vdb,
-                self.relationships_vdb,
-                self.chunks_vdb,
-                self.text_chunks,
-                param,
-                asdict(self),
-                hashing_kv=self.llm_response_cache
-                if self.llm_response_cache
-                and hasattr(self.llm_response_cache, "global_config")
-                else self.key_string_value_json_storage_cls(
-                    namespace=make_namespace(
-                        self.namespace_prefix, NameSpace.KV_STORE_LLM_RESPONSE_CACHE
-                    ),
-                    global_config=asdict(self),
-                    embedding_func=self.embedding_func,
-                ),
-            )
-        else:
-            raise ValueError(f"Unknown mode {param.mode}")
-
         await self._query_done()
         return response
 
@@ -1525,21 +1434,6 @@ class LightRAG:
             ]
         )
 
-    def _get_content_summary(self, content: str, max_length: int = 100) -> str:
-        """Get summary of document content
-
-        Args:
-            content: Original document content
-            max_length: Maximum length of summary
-
-        Returns:
-            Truncated content with ellipsis if needed
-        """
-        content = content.strip()
-        if len(content) <= max_length:
-            return content
-        return content[:max_length] + "..."
-
     async def get_processing_status(self) -> dict[str, int]:
         """Get current document processing status counts
 
@@ -1816,19 +1710,7 @@ class LightRAG:
     async def get_entity_info(
         self, entity_name: str, include_vector_data: bool = False
     ) -> dict[str, str | None | dict[str, str]]:
-        """Get detailed information of an entity
+        """Get detailed information of an entity"""
-
-        Args:
-            entity_name: Entity name (no need for quotes)
-            include_vector_data: Whether to include data from the vector database
-
-        Returns:
-            dict: A dictionary containing entity information, including:
-                - entity_name: Entity name
-                - source_id: Source document ID
-                - graph_data: Complete node data from the graph database
-                - vector_data: (optional) Data from the vector database
-        """
 
         # Get information from the graph
         node_data = await self.chunk_entity_relation_graph.get_node(entity_name)
@@ -1843,29 +1725,15 @@ class LightRAG:
         # Optional: Get vector database information
         if include_vector_data:
             entity_id = compute_mdhash_id(entity_name, prefix="ent-")
-            vector_data = self.entities_vdb._client.get([entity_id])
+            vector_data = await self.entities_vdb.get_by_id(entity_id)
-            result["vector_data"] = vector_data[0] if vector_data else None
+            result["vector_data"] = vector_data
 
         return result
 
     async def get_relation_info(
         self, src_entity: str, tgt_entity: str, include_vector_data: bool = False
     ) -> dict[str, str | None | dict[str, str]]:
-        """Get detailed information of a relationship
+        """Get detailed information of a relationship"""
-
-        Args:
-            src_entity: Source entity name (no need for quotes)
-            tgt_entity: Target entity name (no need for quotes)
-            include_vector_data: Whether to include data from the vector database
-
-        Returns:
-            dict: A dictionary containing relationship information, including:
-                - src_entity: Source entity name
-                - tgt_entity: Target entity name
-                - source_id: Source document ID
-                - graph_data: Complete edge data from the graph database
-                - vector_data: (optional) Data from the vector database
-        """
 
         # Get information from the graph
         edge_data = await self.chunk_entity_relation_graph.get_edge(
@@ -1883,8 +1751,8 @@ class LightRAG:
         # Optional: Get vector database information
         if include_vector_data:
             rel_id = compute_mdhash_id(src_entity + tgt_entity, prefix="rel-")
-            vector_data = self.relationships_vdb._client.get([rel_id])
+            vector_data = await self.relationships_vdb.get_by_id(rel_id)
-            result["vector_data"] = vector_data[0] if vector_data else None
+            result["vector_data"] = vector_data
 
         return result
 
@@ -2682,6 +2550,12 @@ class LightRAG:
 
             # 9. Delete source entities
             for entity_name in source_entities:
+                if entity_name == target_entity:
+                    logger.info(
+                        f"Skipping deletion of '{entity_name}' as it's also the target entity"
+                    )
+                    continue
+
                 # Delete entity node from knowledge graph
                 await self.chunk_entity_relation_graph.delete_node(entity_name)
 

lightrag/llm/azure_openai.py

@@ -55,6 +55,7 @@ async def azure_openai_complete_if_cache(
 
     openai_async_client = AsyncAzureOpenAI(
         azure_endpoint=os.getenv("AZURE_OPENAI_ENDPOINT"),
+        azure_deployment=model,
         api_key=os.getenv("AZURE_OPENAI_API_KEY"),
         api_version=os.getenv("AZURE_OPENAI_API_VERSION"),
     )
@@ -136,6 +137,7 @@ async def azure_openai_embed(
 
     openai_async_client = AsyncAzureOpenAI(
         azure_endpoint=os.getenv("AZURE_OPENAI_ENDPOINT"),
+        azure_deployment=model,
         api_key=os.getenv("AZURE_OPENAI_API_KEY"),
         api_version=os.getenv("AZURE_OPENAI_API_VERSION"),
     )

lightrag/operate.py

@@ -3,6 +3,7 @@ from __future__ import annotations
 import asyncio
 import json
 import re
+import os
 from typing import Any, AsyncIterator
 from collections import Counter, defaultdict
 
@@ -140,18 +141,36 @@ async def _handle_single_entity_extraction(
 ):
     if len(record_attributes) < 4 or record_attributes[0] != '"entity"':
         return None
-    # add this record as a node in the G
+
+    # Clean and validate entity name
     entity_name = clean_str(record_attributes[1]).strip('"')
     if not entity_name.strip():
+        logger.warning(
+            f"Entity extraction error: empty entity name in: {record_attributes}"
+        )
         return None
+
+    # Clean and validate entity type
     entity_type = clean_str(record_attributes[2]).strip('"')
+    if not entity_type.strip() or entity_type.startswith('("'):
+        logger.warning(
+            f"Entity extraction error: invalid entity type in: {record_attributes}"
+        )
+        return None
+
+    # Clean and validate description
     entity_description = clean_str(record_attributes[3]).strip('"')
-    entity_source_id = chunk_key
+    if not entity_description.strip():
+        logger.warning(
+            f"Entity extraction error: empty description for entity '{entity_name}' of type '{entity_type}'"
+        )
+        return None
+
     return dict(
         entity_name=entity_name,
         entity_type=entity_type,
         description=entity_description,
-        source_id=entity_source_id,
+        source_id=chunk_key,
         metadata={"created_at": time.time()},
     )
 
@@ -220,6 +239,7 @@ async def _merge_nodes_then_upsert(
         entity_name, description, global_config
     )
     node_data = dict(
+        entity_id=entity_name,
         entity_type=entity_type,
         description=description,
         source_id=source_id,
@@ -301,6 +321,7 @@ async def _merge_edges_then_upsert(
             await knowledge_graph_inst.upsert_node(
                 need_insert_id,
                 node_data={
+                    "entity_id": need_insert_id,
                     "source_id": source_id,
                     "description": description,
                     "entity_type": "UNKNOWN",
@@ -337,11 +358,10 @@ async def extract_entities(
     entity_vdb: BaseVectorStorage,
     relationships_vdb: BaseVectorStorage,
     global_config: dict[str, str],
+    pipeline_status: dict = None,
+    pipeline_status_lock=None,
     llm_response_cache: BaseKVStorage | None = None,
 ) -> None:
-    from lightrag.kg.shared_storage import get_namespace_data
-
-    pipeline_status = await get_namespace_data("pipeline_status")
     use_llm_func: callable = global_config["llm_model_func"]
     entity_extract_max_gleaning = global_config["entity_extract_max_gleaning"]
     enable_llm_cache_for_entity_extract: bool = global_config[
@@ -400,6 +420,7 @@ async def extract_entities(
             else:
                 _prompt = input_text
 
+            # TODO： add cache_type="extract"
             arg_hash = compute_args_hash(_prompt)
             cached_return, _1, _2, _3 = await handle_cache(
                 llm_response_cache,
@@ -407,7 +428,6 @@ async def extract_entities(
                 _prompt,
                 "default",
                 cache_type="extract",
-                force_llm_cache=True,
             )
             if cached_return:
                 logger.debug(f"Found cache for {arg_hash}")
@@ -436,47 +456,22 @@ async def extract_entities(
         else:
             return await use_llm_func(input_text)
 
-    async def _process_single_content(chunk_key_dp: tuple[str, TextChunkSchema]):
+    async def _process_extraction_result(result: str, chunk_key: str):
-        """ "Prpocess a single chunk
+        """Process a single extraction result (either initial or gleaning)
         Args:
-            chunk_key_dp (tuple[str, TextChunkSchema]):
+            result (str): The extraction result to process
-                ("chunck-xxxxxx", {"tokens": int, "content": str, "full_doc_id": str, "chunk_order_index": int})
+            chunk_key (str): The chunk key for source tracking
+        Returns:
+            tuple: (nodes_dict, edges_dict) containing the extracted entities and relationships
         """
-        nonlocal processed_chunks
+        maybe_nodes = defaultdict(list)
-        chunk_key = chunk_key_dp[0]
+        maybe_edges = defaultdict(list)
-        chunk_dp = chunk_key_dp[1]
-        content = chunk_dp["content"]
-        # hint_prompt = entity_extract_prompt.format(**context_base, input_text=content)
-        hint_prompt = entity_extract_prompt.format(
-            **context_base, input_text="{input_text}"
-        ).format(**context_base, input_text=content)
-
-        final_result = await _user_llm_func_with_cache(hint_prompt)
-        history = pack_user_ass_to_openai_messages(hint_prompt, final_result)
-        for now_glean_index in range(entity_extract_max_gleaning):
-            glean_result = await _user_llm_func_with_cache(
-                continue_prompt, history_messages=history
-            )
-
-            history += pack_user_ass_to_openai_messages(continue_prompt, glean_result)
-            final_result += glean_result
-            if now_glean_index == entity_extract_max_gleaning - 1:
-                break
-
-            if_loop_result: str = await _user_llm_func_with_cache(
-                if_loop_prompt, history_messages=history
-            )
-            if_loop_result = if_loop_result.strip().strip('"').strip("'").lower()
-            if if_loop_result != "yes":
-                break
 
         records = split_string_by_multi_markers(
-            final_result,
+            result,
             [context_base["record_delimiter"], context_base["completion_delimiter"]],
         )
 
-        maybe_nodes = defaultdict(list)
-        maybe_edges = defaultdict(list)
         for record in records:
             record = re.search(r"\((.*)\)", record)
             if record is None:
@@ -485,6 +480,7 @@ async def extract_entities(
             record_attributes = split_string_by_multi_markers(
                 record, [context_base["tuple_delimiter"]]
             )
+
             if_entities = await _handle_single_entity_extraction(
                 record_attributes, chunk_key
             )
@@ -499,11 +495,69 @@ async def extract_entities(
                 maybe_edges[(if_relation["src_id"], if_relation["tgt_id"])].append(
                     if_relation
                 )
+
+        return maybe_nodes, maybe_edges
+
+    async def _process_single_content(chunk_key_dp: tuple[str, TextChunkSchema]):
+        """Process a single chunk
+        Args:
+            chunk_key_dp (tuple[str, TextChunkSchema]):
+                ("chunk-xxxxxx", {"tokens": int, "content": str, "full_doc_id": str, "chunk_order_index": int})
+        """
+        nonlocal processed_chunks
+        chunk_key = chunk_key_dp[0]
+        chunk_dp = chunk_key_dp[1]
+        content = chunk_dp["content"]
+
+        # Get initial extraction
+        hint_prompt = entity_extract_prompt.format(
+            **context_base, input_text="{input_text}"
+        ).format(**context_base, input_text=content)
+
+        final_result = await _user_llm_func_with_cache(hint_prompt)
+        history = pack_user_ass_to_openai_messages(hint_prompt, final_result)
+
+        # Process initial extraction
+        maybe_nodes, maybe_edges = await _process_extraction_result(
+            final_result, chunk_key
+        )
+
+        # Process additional gleaning results
+        for now_glean_index in range(entity_extract_max_gleaning):
+            glean_result = await _user_llm_func_with_cache(
+                continue_prompt, history_messages=history
+            )
+
+            history += pack_user_ass_to_openai_messages(continue_prompt, glean_result)
+
+            # Process gleaning result separately
+            glean_nodes, glean_edges = await _process_extraction_result(
+                glean_result, chunk_key
+            )
+
+            # Merge results
+            for entity_name, entities in glean_nodes.items():
+                maybe_nodes[entity_name].extend(entities)
+            for edge_key, edges in glean_edges.items():
+                maybe_edges[edge_key].extend(edges)
+
+            if now_glean_index == entity_extract_max_gleaning - 1:
+                break
+
+            if_loop_result: str = await _user_llm_func_with_cache(
+                if_loop_prompt, history_messages=history
+            )
+            if_loop_result = if_loop_result.strip().strip('"').strip("'").lower()
+            if if_loop_result != "yes":
+                break
+
         processed_chunks += 1
         entities_count = len(maybe_nodes)
         relations_count = len(maybe_edges)
         log_message = f"  Chunk {processed_chunks}/{total_chunks}: extracted {entities_count} entities and {relations_count} relationships (deduplicated)"
         logger.info(log_message)
+        if pipeline_status is not None:
+            async with pipeline_status_lock:
                 pipeline_status["latest_message"] = log_message
                 pipeline_status["history_messages"].append(log_message)
         return dict(maybe_nodes), dict(maybe_edges)
@@ -519,6 +573,12 @@ async def extract_entities(
         for k, v in m_edges.items():
             maybe_edges[tuple(sorted(k))].extend(v)
 
+    from .kg.shared_storage import get_graph_db_lock
+
+    graph_db_lock = get_graph_db_lock(enable_logging=False)
+
+    # Ensure that nodes and edges are merged and upserted atomically
+    async with graph_db_lock:
         all_entities_data = await asyncio.gather(
             *[
                 _merge_nodes_then_upsert(k, v, knowledge_graph_inst, global_config)
@@ -528,7 +588,9 @@ async def extract_entities(
 
         all_relationships_data = await asyncio.gather(
             *[
-            _merge_edges_then_upsert(k[0], k[1], v, knowledge_graph_inst, global_config)
+                _merge_edges_then_upsert(
+                    k[0], k[1], v, knowledge_graph_inst, global_config
+                )
                 for k, v in maybe_edges.items()
             ]
         )
@@ -536,6 +598,8 @@ async def extract_entities(
     if not (all_entities_data or all_relationships_data):
         log_message = "Didn't extract any entities and relationships."
         logger.info(log_message)
+        if pipeline_status is not None:
+            async with pipeline_status_lock:
                 pipeline_status["latest_message"] = log_message
                 pipeline_status["history_messages"].append(log_message)
         return
@@ -543,16 +607,22 @@ async def extract_entities(
     if not all_entities_data:
         log_message = "Didn't extract any entities"
         logger.info(log_message)
+        if pipeline_status is not None:
+            async with pipeline_status_lock:
                 pipeline_status["latest_message"] = log_message
                 pipeline_status["history_messages"].append(log_message)
     if not all_relationships_data:
         log_message = "Didn't extract any relationships"
         logger.info(log_message)
+        if pipeline_status is not None:
+            async with pipeline_status_lock:
                 pipeline_status["latest_message"] = log_message
                 pipeline_status["history_messages"].append(log_message)
 
     log_message = f"Extracted {len(all_entities_data)} entities and {len(all_relationships_data)} relationships (deduplicated)"
     logger.info(log_message)
+    if pipeline_status is not None:
+        async with pipeline_status_lock:
             pipeline_status["latest_message"] = log_message
             pipeline_status["history_messages"].append(log_message)
     verbose_debug(
@@ -1020,6 +1090,7 @@ async def _build_query_context(
     text_chunks_db: BaseKVStorage,
     query_param: QueryParam,
 ):
+    logger.info(f"Process {os.getpid()} buidling query context...")
     if query_param.mode == "local":
         entities_context, relations_context, text_units_context = await _get_node_data(
             ll_keywords,
@@ -1845,3 +1916,90 @@ async def kg_query_with_keywords(
         )
 
     return response
+
+
+async def query_with_keywords(
+    query: str,
+    prompt: str,
+    param: QueryParam,
+    knowledge_graph_inst: BaseGraphStorage,
+    entities_vdb: BaseVectorStorage,
+    relationships_vdb: BaseVectorStorage,
+    chunks_vdb: BaseVectorStorage,
+    text_chunks_db: BaseKVStorage,
+    global_config: dict[str, str],
+    hashing_kv: BaseKVStorage | None = None,
+) -> str | AsyncIterator[str]:
+    """
+    Extract keywords from the query and then use them for retrieving information.
+
+    1. Extracts high-level and low-level keywords from the query
+    2. Formats the query with the extracted keywords and prompt
+    3. Uses the appropriate query method based on param.mode
+
+    Args:
+        query: The user's query
+        prompt: Additional prompt to prepend to the query
+        param: Query parameters
+        knowledge_graph_inst: Knowledge graph storage
+        entities_vdb: Entities vector database
+        relationships_vdb: Relationships vector database
+        chunks_vdb: Document chunks vector database
+        text_chunks_db: Text chunks storage
+        global_config: Global configuration
+        hashing_kv: Cache storage
+
+    Returns:
+        Query response or async iterator
+    """
+    # Extract keywords
+    hl_keywords, ll_keywords = await extract_keywords_only(
+        text=query,
+        param=param,
+        global_config=global_config,
+        hashing_kv=hashing_kv,
+    )
+
+    param.hl_keywords = hl_keywords
+    param.ll_keywords = ll_keywords
+
+    # Create a new string with the prompt and the keywords
+    ll_keywords_str = ", ".join(ll_keywords)
+    hl_keywords_str = ", ".join(hl_keywords)
+    formatted_question = f"{prompt}\n\n### Keywords:\nHigh-level: {hl_keywords_str}\nLow-level: {ll_keywords_str}\n\n### Query:\n{query}"
+
+    # Use appropriate query method based on mode
+    if param.mode in ["local", "global", "hybrid"]:
+        return await kg_query_with_keywords(
+            formatted_question,
+            knowledge_graph_inst,
+            entities_vdb,
+            relationships_vdb,
+            text_chunks_db,
+            param,
+            global_config,
+            hashing_kv=hashing_kv,
+        )
+    elif param.mode == "naive":
+        return await naive_query(
+            formatted_question,
+            chunks_vdb,
+            text_chunks_db,
+            param,
+            global_config,
+            hashing_kv=hashing_kv,
+        )
+    elif param.mode == "mix":
+        return await mix_kg_vector_query(
+            formatted_question,
+            knowledge_graph_inst,
+            entities_vdb,
+            relationships_vdb,
+            chunks_vdb,
+            text_chunks_db,
+            param,
+            global_config,
+            hashing_kv=hashing_kv,
+        )
+    else:
+        raise ValueError(f"Unknown mode {param.mode}")

lightrag/prompt.py

@@ -236,7 +236,7 @@ Given the query and conversation history, list both high-level and low-level key
 ---Instructions---
 
 - Consider both the current query and relevant conversation history when extracting keywords
-- Output the keywords in JSON format
+- Output the keywords in JSON format, it will be parsed by a JSON parser, do not add any extra content in output
 - The JSON should have two keys:
   - "high_level_keywords" for overarching concepts or themes
   - "low_level_keywords" for specific entities or details

lightrag/utils.py

@@ -633,15 +633,15 @@ async def handle_cache(
     prompt,
     mode="default",
     cache_type=None,
-    force_llm_cache=False,
 ):
     """Generic cache handling function"""
-    if hashing_kv is None or not (
+    if hashing_kv is None:
-        force_llm_cache or hashing_kv.global_config.get("enable_llm_cache")
+        return None, None, None, None
-    ):
+
+    if mode != "default":  # handle cache for all type of query
+        if not hashing_kv.global_config.get("enable_llm_cache"):
             return None, None, None, None
 
-    if mode != "default":
         # Get embedding cache configuration
         embedding_cache_config = hashing_kv.global_config.get(
             "embedding_cache_config",
@@ -651,8 +651,7 @@ async def handle_cache(
         use_llm_check = embedding_cache_config.get("use_llm_check", False)
 
         quantized = min_val = max_val = None
-        if is_embedding_cache_enabled:
+        if is_embedding_cache_enabled:  # Use embedding simularity to match cache
-            # Use embedding cache
             current_embedding = await hashing_kv.embedding_func([prompt])
             llm_model_func = hashing_kv.global_config.get("llm_model_func")
             quantized, min_val, max_val = quantize_embedding(current_embedding[0])
@@ -667,24 +666,29 @@ async def handle_cache(
                 cache_type=cache_type,
             )
             if best_cached_response is not None:
-                logger.info(f"Embedding cached hit(mode:{mode} type:{cache_type})")
+                logger.debug(f"Embedding cached hit(mode:{mode} type:{cache_type})")
                 return best_cached_response, None, None, None
             else:
                 # if caching keyword embedding is enabled, return the quantized embedding for saving it latter
-                logger.info(f"Embedding cached missed(mode:{mode} type:{cache_type})")
+                logger.debug(f"Embedding cached missed(mode:{mode} type:{cache_type})")
                 return None, quantized, min_val, max_val
 
-    # For default mode or is_embedding_cache_enabled is False, use regular cache
+    else:  # handle cache for entity extraction
-    # default mode is for extract_entities or naive query
+        if not hashing_kv.global_config.get("enable_llm_cache_for_entity_extract"):
+            return None, None, None, None
+
+    # Here is the conditions of code reaching this point:
+    #     1. All query mode: enable_llm_cache is True and embedding simularity is not enabled
+    #     2. Entity extract: enable_llm_cache_for_entity_extract is True
     if exists_func(hashing_kv, "get_by_mode_and_id"):
         mode_cache = await hashing_kv.get_by_mode_and_id(mode, args_hash) or {}
     else:
         mode_cache = await hashing_kv.get_by_id(mode) or {}
     if args_hash in mode_cache:
-        logger.info(f"Non-embedding cached hit(mode:{mode} type:{cache_type})")
+        logger.debug(f"Non-embedding cached hit(mode:{mode} type:{cache_type})")
         return mode_cache[args_hash]["return"], None, None, None
 
-    logger.info(f"Non-embedding cached missed(mode:{mode} type:{cache_type})")
+    logger.debug(f"Non-embedding cached missed(mode:{mode} type:{cache_type})")
     return None, None, None, None
 
 
@@ -701,9 +705,22 @@ class CacheData:
 
 
 async def save_to_cache(hashing_kv, cache_data: CacheData):
-    if hashing_kv is None or hasattr(cache_data.content, "__aiter__"):
+    """Save data to cache, with improved handling for streaming responses and duplicate content.
+
+    Args:
+        hashing_kv: The key-value storage for caching
+        cache_data: The cache data to save
+    """
+    # Skip if storage is None or content is a streaming response
+    if hashing_kv is None or not cache_data.content:
         return
 
+    # If content is a streaming response, don't cache it
+    if hasattr(cache_data.content, "__aiter__"):
+        logger.debug("Streaming response detected, skipping cache")
+        return
+
+    # Get existing cache data
     if exists_func(hashing_kv, "get_by_mode_and_id"):
         mode_cache = (
             await hashing_kv.get_by_mode_and_id(cache_data.mode, cache_data.args_hash)
@@ -712,6 +729,16 @@ async def save_to_cache(hashing_kv, cache_data: CacheData):
     else:
         mode_cache = await hashing_kv.get_by_id(cache_data.mode) or {}
 
+    # Check if we already have identical content cached
+    if cache_data.args_hash in mode_cache:
+        existing_content = mode_cache[cache_data.args_hash].get("return")
+        if existing_content == cache_data.content:
+            logger.info(
+                f"Cache content unchanged for {cache_data.args_hash}, skipping update"
+            )
+            return
+
+    # Update cache with new content
     mode_cache[cache_data.args_hash] = {
         "return": cache_data.content,
         "cache_type": cache_data.cache_type,
@@ -726,6 +753,7 @@ async def save_to_cache(hashing_kv, cache_data: CacheData):
         "original_prompt": cache_data.prompt,
     }
 
+    # Only upsert if there's actual new content
     await hashing_kv.upsert({cache_data.mode: mode_cache})
 
 
@@ -862,3 +890,52 @@ def lazy_external_import(module_name: str, class_name: str) -> Callable[..., Any
         return cls(*args, **kwargs)
 
     return import_class
+
+
+def get_content_summary(content: str, max_length: int = 100) -> str:
+    """Get summary of document content
+
+    Args:
+        content: Original document content
+        max_length: Maximum length of summary
+
+    Returns:
+        Truncated content with ellipsis if needed
+    """
+    content = content.strip()
+    if len(content) <= max_length:
+        return content
+    return content[:max_length] + "..."
+
+
+def clean_text(text: str) -> str:
+    """Clean text by removing null bytes (0x00) and whitespace
+
+    Args:
+        text: Input text to clean
+
+    Returns:
+        Cleaned text
+    """
+    return text.strip().replace("\x00", "")
+
+
+def check_storage_env_vars(storage_name: str) -> None:
+    """Check if all required environment variables for storage implementation exist
+
+    Args:
+        storage_name: Storage implementation name
+
+    Raises:
+        ValueError: If required environment variables are missing
+    """
+    from lightrag.kg import STORAGE_ENV_REQUIREMENTS
+
+    required_vars = STORAGE_ENV_REQUIREMENTS.get(storage_name, [])
+    missing_vars = [var for var in required_vars if var not in os.environ]
+
+    if missing_vars:
+        raise ValueError(
+            f"Storage implementation '{storage_name}' requires the following "
+            f"environment variables: {', '.join(missing_vars)}"
+        )

lightrag_webui/bun.lock

@@ -34,11 +34,13 @@
         "cmdk": "^1.0.4",
         "graphology": "^0.26.0",
         "graphology-generators": "^0.11.2",
+        "i18next": "^24.2.2",
         "lucide-react": "^0.475.0",
         "minisearch": "^7.1.2",
         "react": "^19.0.0",
         "react-dom": "^19.0.0",
         "react-dropzone": "^14.3.6",
+        "react-i18next": "^15.4.1",
         "react-markdown": "^9.1.0",
         "react-number-format": "^5.4.3",
         "react-syntax-highlighter": "^15.6.1",
@@ -765,8 +767,12 @@
 
     "hoist-non-react-statics": ["hoist-non-react-statics@3.3.2", "", { "dependencies": { "react-is": "^16.7.0" } }, "sha512-/gGivxi8JPKWNm/W0jSmzcMPpfpPLc3dY/6GxhX2hQ9iGj3aDfklV4ET7NjKpSinLpJ5vafa9iiGIEZg10SfBw=="],
 
+    "html-parse-stringify": ["html-parse-stringify@3.0.1", "", { "dependencies": { "void-elements": "3.1.0" } }, "sha512-KknJ50kTInJ7qIScF3jeaFRpMpE8/lfiTdzf/twXyPBLAGrLRTmkz3AdTnKeh40X8k9L2fdYwEp/42WGXIRGcg=="],
+
     "html-url-attributes": ["html-url-attributes@3.0.1", "", {}, "sha512-ol6UPyBWqsrO6EJySPz2O7ZSr856WDrEzM5zMqp+FJJLGMW35cLYmmZnl0vztAZxRUoNZJFTCohfjuIJ8I4QBQ=="],
 
+    "i18next": ["i18next@24.2.2", "", { "dependencies": { "@babel/runtime": "^7.23.2" }, "peerDependencies": { "typescript": "^5" }, "optionalPeers": ["typescript"] }, "sha512-NE6i86lBCKRYZa5TaUDkU5S4HFgLIEJRLr3Whf2psgaxBleQ2LC1YW1Vc+SCgkAW7VEzndT6al6+CzegSUHcTQ=="],
+
     "ignore": ["ignore@5.3.2", "", {}, "sha512-hsBTNUqQTDwkWtcdYI2i06Y/nUBEsNEDJKjWdigLvegy8kDuJAS8uRlpkkcQpyEXL0Z/pjDy5HBmMjRCJ2gq+g=="],
 
     "import-fresh": ["import-fresh@3.3.1", "", { "dependencies": { "parent-module": "^1.0.0", "resolve-from": "^4.0.0" } }, "sha512-TR3KfrTZTYLPB6jUjfx6MF9WcWrHL9su5TObK4ZkYgBdWKPOFoSoQIdEuTuR82pmtxH2spWG9h6etwfr1pLBqQ=="],
@@ -1093,6 +1099,8 @@
 
     "react-dropzone": ["react-dropzone@14.3.6", "", { "dependencies": { "attr-accept": "^2.2.4", "file-selector": "^2.1.0", "prop-types": "^15.8.1" }, "peerDependencies": { "react": ">= 16.8 || 18.0.0" } }, "sha512-U792j+x0rcwH/U/Slv/OBNU/LGFYbDLHKKiJoPhNaOianayZevCt4Y5S0CraPssH/6/wT6xhKDfzdXUgCBS0HQ=="],
 
+    "react-i18next": ["react-i18next@15.4.1", "", { "dependencies": { "@babel/runtime": "^7.25.0", "html-parse-stringify": "^3.0.1" }, "peerDependencies": { "i18next": ">= 23.2.3", "react": ">= 16.8.0" } }, "sha512-ahGab+IaSgZmNPYXdV1n+OYky95TGpFwnKRflX/16dY04DsYYKHtVLjeny7sBSCREEcoMbAgSkFiGLF5g5Oofw=="],
+
     "react-is": ["react-is@16.13.1", "", {}, "sha512-24e6ynE2H+OKt4kqsOvNd8kBpV65zoxbA4BVsEOB3ARVWQki/DHzaUoC5KuON/BiccDaCCTZBuOcfZs70kR8bQ=="],
 
     "react-markdown": ["react-markdown@9.1.0", "", { "dependencies": { "@types/hast": "^3.0.0", "@types/mdast": "^4.0.0", "devlop": "^1.0.0", "hast-util-to-jsx-runtime": "^2.0.0", "html-url-attributes": "^3.0.0", "mdast-util-to-hast": "^13.0.0", "remark-parse": "^11.0.0", "remark-rehype": "^11.0.0", "unified": "^11.0.0", "unist-util-visit": "^5.0.0", "vfile": "^6.0.0" }, "peerDependencies": { "@types/react": ">=18", "react": ">=18" } }, "sha512-xaijuJB0kzGiUdG7nc2MOMDUDBWPyGAjZtUrow9XxUeua8IqeP+VlIfAZ3bphpcLTnSZXz6z9jcVC/TCwbfgdw=="],
@@ -1271,6 +1279,8 @@
 
     "vite": ["vite@6.1.1", "", { "dependencies": { "esbuild": "^0.24.2", "postcss": "^8.5.2", "rollup": "^4.30.1" }, "optionalDependencies": { "fsevents": "~2.3.3" }, "peerDependencies": { "@types/node": "^18.0.0 || ^20.0.0 || >=22.0.0", "jiti": ">=1.21.0", "less": "*", "lightningcss": "^1.21.0", "sass": "*", "sass-embedded": "*", "stylus": "*", "sugarss": "*", "terser": "^5.16.0", "tsx": "^4.8.1", "yaml": "^2.4.2" }, "optionalPeers": ["@types/node", "jiti", "less", "lightningcss", "sass", "sass-embedded", "stylus", "sugarss", "terser", "tsx", "yaml"], "bin": { "vite": "bin/vite.js" } }, "sha512-4GgM54XrwRfrOp297aIYspIti66k56v16ZnqHvrIM7mG+HjDlAwS7p+Srr7J6fGvEdOJ5JcQ/D9T7HhtdXDTzA=="],
 
+    "void-elements": ["void-elements@3.1.0", "", {}, "sha512-Dhxzh5HZuiHQhbvTW9AMetFfBHDMYpo23Uo9btPXgdYP+3T5S+p+jgNy7spra+veYhBP2dCSgxR/i2Y02h5/6w=="],
+
     "which": ["which@2.0.2", "", { "dependencies": { "isexe": "^2.0.0" }, "bin": { "node-which": "./bin/node-which" } }, "sha512-BLI3Tl1TW3Pvl70l3yq3Y64i+awpwXqsGBYWkkqMtnbXgrMD+yj7rhW0kuEDxzJaYXGjEW5ogapKNMEKNMjibA=="],
 
     "which-boxed-primitive": ["which-boxed-primitive@1.1.1", "", { "dependencies": { "is-bigint": "^1.1.0", "is-boolean-object": "^1.2.1", "is-number-object": "^1.1.1", "is-string": "^1.1.1", "is-symbol": "^1.1.1" } }, "sha512-TbX3mj8n0odCBFVlY8AxkqcHASw3L60jIuF8jFP78az3C2YhmGvqbHBpAjTRH2/xqYunrJ9g1jSyjCjpoWzIAA=="],

lightrag_webui/package.json

@@ -43,11 +43,13 @@
     "cmdk": "^1.0.4",
     "graphology": "^0.26.0",
     "graphology-generators": "^0.11.2",
+    "i18next": "^24.2.2",
     "lucide-react": "^0.475.0",
     "minisearch": "^7.1.2",
     "react": "^19.0.0",
     "react-dom": "^19.0.0",
     "react-dropzone": "^14.3.6",
+    "react-i18next": "^15.4.1",
     "react-markdown": "^9.1.0",
     "react-number-format": "^5.4.3",
     "react-syntax-highlighter": "^15.6.1",

lightrag_webui/src/components/ThemeToggle.tsx

@@ -3,6 +3,7 @@ import useTheme from '@/hooks/useTheme'
 import { MoonIcon, SunIcon } from 'lucide-react'
 import { useCallback } from 'react'
 import { controlButtonVariant } from '@/lib/constants'
+import { useTranslation } from 'react-i18next'
 
 /**
  * Component that toggles the theme between light and dark.
@@ -11,13 +12,14 @@ export default function ThemeToggle() {
   const { theme, setTheme } = useTheme()
   const setLight = useCallback(() => setTheme('light'), [setTheme])
   const setDark = useCallback(() => setTheme('dark'), [setTheme])
+  const { t } = useTranslation()
 
   if (theme === 'dark') {
     return (
       <Button
         onClick={setLight}
         variant={controlButtonVariant}
-        tooltip="Switch to light theme"
+        tooltip={t('header.themeToggle.switchToLight')}
         size="icon"
         side="bottom"
       >
@@ -29,7 +31,7 @@ export default function ThemeToggle() {
     <Button
       onClick={setDark}
       variant={controlButtonVariant}
-      tooltip="Switch to dark theme"
+      tooltip={t('header.themeToggle.switchToDark')}
       size="icon"
       side="bottom"
     >

lightrag_webui/src/components/documents/ClearDocumentsDialog.tsx

@@ -13,38 +13,40 @@ import { errorMessage } from '@/lib/utils'
 import { clearDocuments } from '@/api/lightrag'
 
 import { EraserIcon } from 'lucide-react'
+import { useTranslation } from 'react-i18next'
 
 export default function ClearDocumentsDialog() {
+  const { t } = useTranslation()
   const [open, setOpen] = useState(false)
 
   const handleClear = useCallback(async () => {
     try {
       const result = await clearDocuments()
       if (result.status === 'success') {
-        toast.success('Documents cleared successfully')
+        toast.success(t('documentPanel.clearDocuments.success'))
         setOpen(false)
       } else {
-        toast.error(`Clear Documents Failed:\n${result.message}`)
+        toast.error(t('documentPanel.clearDocuments.failed', { message: result.message }))
       }
     } catch (err) {
-      toast.error('Clear Documents Failed:\n' + errorMessage(err))
+      toast.error(t('documentPanel.clearDocuments.error', { error: errorMessage(err) }))
     }
   }, [setOpen])
 
   return (
     <Dialog open={open} onOpenChange={setOpen}>
       <DialogTrigger asChild>
-        <Button variant="outline" side="bottom" tooltip='Clear documents' size="sm">
+        <Button variant="outline" side="bottom" tooltip={t('documentPanel.clearDocuments.tooltip')} size="sm">
-          <EraserIcon/> Clear
+          <EraserIcon/> {t('documentPanel.clearDocuments.button')}
         </Button>
       </DialogTrigger>
       <DialogContent className="sm:max-w-xl" onCloseAutoFocus={(e) => e.preventDefault()}>
         <DialogHeader>
-          <DialogTitle>Clear documents</DialogTitle>
+          <DialogTitle>{t('documentPanel.clearDocuments.title')}</DialogTitle>
-          <DialogDescription>Do you really want to clear all documents?</DialogDescription>
+          <DialogDescription>{t('documentPanel.clearDocuments.confirm')}</DialogDescription>
         </DialogHeader>
         <Button variant="destructive" onClick={handleClear}>
-          YES
+        {t('documentPanel.clearDocuments.confirmButton')}
         </Button>
       </DialogContent>
     </Dialog>

lightrag_webui/src/components/documents/UploadDocumentsDialog.tsx

@@ -14,8 +14,10 @@ import { errorMessage } from '@/lib/utils'
 import { uploadDocument } from '@/api/lightrag'
 
 import { UploadIcon } from 'lucide-react'
+import { useTranslation } from 'react-i18next'
 
 export default function UploadDocumentsDialog() {
+  const { t } = useTranslation()
   const [open, setOpen] = useState(false)
   const [isUploading, setIsUploading] = useState(false)
   const [progresses, setProgresses] = useState<Record<string, number>>({})
@@ -29,24 +31,24 @@ export default function UploadDocumentsDialog() {
           filesToUpload.map(async (file) => {
             try {
               const result = await uploadDocument(file, (percentCompleted: number) => {
-                console.debug(`Uploading ${file.name}: ${percentCompleted}%`)
+                console.debug(t('documentPanel.uploadDocuments.uploading', { name: file.name, percent: percentCompleted }))
                 setProgresses((pre) => ({
                   ...pre,
                   [file.name]: percentCompleted
                 }))
               })
               if (result.status === 'success') {
-                toast.success(`Upload Success:\n${file.name} uploaded successfully`)
+                toast.success(t('documentPanel.uploadDocuments.success', { name: file.name }))
               } else {
-                toast.error(`Upload Failed:\n${file.name}\n${result.message}`)
+                toast.error(t('documentPanel.uploadDocuments.failed', { name: file.name, message: result.message }))
               }
             } catch (err) {
-              toast.error(`Upload Failed:\n${file.name}\n${errorMessage(err)}`)
+              toast.error(t('documentPanel.uploadDocuments.error', { name: file.name, error: errorMessage(err) }))
             }
           })
         )
       } catch (err) {
-        toast.error('Upload Failed\n' + errorMessage(err))
+        toast.error(t('documentPanel.uploadDocuments.generalError', { error: errorMessage(err) }))
       } finally {
         setIsUploading(false)
         // setOpen(false)
@@ -66,21 +68,21 @@ export default function UploadDocumentsDialog() {
       }}
     >
       <DialogTrigger asChild>
-        <Button variant="default" side="bottom" tooltip="Upload documents" size="sm">
+        <Button variant="default" side="bottom" tooltip={t('documentPanel.uploadDocuments.tooltip')} size="sm">
-          <UploadIcon /> Upload
+          <UploadIcon /> {t('documentPanel.uploadDocuments.button')}
         </Button>
       </DialogTrigger>
       <DialogContent className="sm:max-w-xl" onCloseAutoFocus={(e) => e.preventDefault()}>
         <DialogHeader>
-          <DialogTitle>Upload documents</DialogTitle>
+          <DialogTitle>{t('documentPanel.uploadDocuments.title')}</DialogTitle>
           <DialogDescription>
-            Drag and drop your documents here or click to browse.
+            {t('documentPanel.uploadDocuments.description')}
           </DialogDescription>
         </DialogHeader>
         <FileUploader
           maxFileCount={Infinity}
           maxSize={200 * 1024 * 1024}
-          description="supported types: TXT, MD, DOCX, PDF, PPTX, RTF, ODT, EPUB, HTML, HTM, TEX, JSON, XML, YAML, YML, CSV, LOG, CONF, INI, PROPERTIES, SQL, BAT, SH, C, CPP, PY, JAVA, JS, TS, SWIFT, GO, RB, PHP, CSS, SCSS, LESS"
+          description={t('documentPanel.uploadDocuments.fileTypes')}
           onUpload={handleDocumentsUpload}
           progresses={progresses}
           disabled={isUploading}

lightrag_webui/src/components/graph/FullScreenControl.tsx

@@ -2,21 +2,23 @@ import { useFullScreen } from '@react-sigma/core'
 import { MaximizeIcon, MinimizeIcon } from 'lucide-react'
 import { controlButtonVariant } from '@/lib/constants'
 import Button from '@/components/ui/Button'
+import { useTranslation } from 'react-i18next'
 
 /**
  * Component that toggles full screen mode.
  */
 const FullScreenControl = () => {
   const { isFullScreen, toggle } = useFullScreen()
+  const { t } = useTranslation()
 
   return (
     <>
       {isFullScreen ? (
-        <Button variant={controlButtonVariant} onClick={toggle} tooltip="Windowed" size="icon">
+        <Button variant={controlButtonVariant} onClick={toggle} tooltip={t('graphPanel.sideBar.fullScreenControl.windowed')} size="icon">
           <MinimizeIcon />
         </Button>
       ) : (
-        <Button variant={controlButtonVariant} onClick={toggle} tooltip="Full Screen" size="icon">
+        <Button variant={controlButtonVariant} onClick={toggle} tooltip={t('graphPanel.sideBar.fullScreenControl.fullScreen')} size="icon">
           <MaximizeIcon />
         </Button>
       )}

lightrag_webui/src/components/graph/GraphLabels.tsx

@@ -5,6 +5,7 @@ import { useSettingsStore } from '@/stores/settings'
 import { useGraphStore } from '@/stores/graph'
 import { labelListLimit } from '@/lib/constants'
 import MiniSearch from 'minisearch'
+import { useTranslation } from 'react-i18next'
 
 const lastGraph: any = {
   graph: null,
@@ -13,6 +14,7 @@ const lastGraph: any = {
 }
 
 const GraphLabels = () => {
+  const { t } = useTranslation()
   const label = useSettingsStore.use.queryLabel()
   const graph = useGraphStore.use.sigmaGraph()
 
@@ -69,7 +71,7 @@ const GraphLabels = () => {
 
       return result.length <= labelListLimit
         ? result
-        : [...result.slice(0, labelListLimit), `And ${result.length - labelListLimit} others`]
+        : [...result.slice(0, labelListLimit), t('graphLabels.andOthers', { count: result.length - labelListLimit })]
     },
     [getSearchEngine]
   )
@@ -84,14 +86,14 @@ const GraphLabels = () => {
       className="ml-2"
       triggerClassName="max-h-8"
       searchInputClassName="max-h-8"
-      triggerTooltip="Select query label"
+      triggerTooltip={t('graphPanel.graphLabels.selectTooltip')}
       fetcher={fetchData}
       renderOption={(item) => <div>{item}</div>}
       getOptionValue={(item) => item}
       getDisplayValue={(item) => <div>{item}</div>}
       notFound={<div className="py-6 text-center text-sm">No labels found</div>}
-      label="Label"
+      label={t('graphPanel.graphLabels.label')}
-      placeholder="Search labels..."
+      placeholder={t('graphPanel.graphLabels.placeholder')}
       value={label !== null ? label : ''}
       onChange={setQueryLabel}
     />

lightrag_webui/src/components/graph/GraphSearch.tsx

@@ -9,6 +9,7 @@ import { AsyncSearch } from '@/components/ui/AsyncSearch'
 import { searchResultLimit } from '@/lib/constants'
 import { useGraphStore } from '@/stores/graph'
 import MiniSearch from 'minisearch'
+import { useTranslation } from 'react-i18next'
 
 interface OptionItem {
   id: string
@@ -44,6 +45,7 @@ export const GraphSearchInput = ({
   onFocus?: GraphSearchInputProps['onFocus']
   value?: GraphSearchInputProps['value']
 }) => {
+  const { t } = useTranslation()
   const graph = useGraphStore.use.sigmaGraph()
 
   const searchEngine = useMemo(() => {
@@ -97,7 +99,7 @@ export const GraphSearchInput = ({
           {
             type: 'message',
             id: messageId,
-            message: `And ${result.length - searchResultLimit} others`
+            message: t('graphPanel.search.message', { count: result.length - searchResultLimit })
           }
         ]
     },
@@ -118,7 +120,7 @@ export const GraphSearchInput = ({
         if (id !== messageId && onFocus) onFocus(id ? { id, type: 'nodes' } : null)
       }}
       label={'item'}
-      placeholder="Search nodes..."
+      placeholder={t('graphPanel.search.placeholder')}
     />
   )
 }

lightrag_webui/src/components/graph/LayoutsControl.tsx

@@ -16,6 +16,7 @@ import { controlButtonVariant } from '@/lib/constants'
 import { useSettingsStore } from '@/stores/settings'
 
 import { GripIcon, PlayIcon, PauseIcon } from 'lucide-react'
+import { useTranslation } from 'react-i18next'
 
 type LayoutName =
   | 'Circular'
@@ -28,6 +29,7 @@ type LayoutName =
 const WorkerLayoutControl = ({ layout, autoRunFor }: WorkerLayoutControlProps) => {
   const sigma = useSigma()
   const { stop, start, isRunning } = layout
+  const { t } = useTranslation()
 
   /**
    * Init component when Sigma or component settings change.
@@ -61,7 +63,7 @@ const WorkerLayoutControl = ({ layout, autoRunFor }: WorkerLayoutControlProps) =
     <Button
       size="icon"
       onClick={() => (isRunning ? stop() : start())}
-      tooltip={isRunning ? 'Stop the layout animation' : 'Start the layout animation'}
+      tooltip={isRunning ? t('graphPanel.sideBar.layoutsControl.stopAnimation') : t('graphPanel.sideBar.layoutsControl.startAnimation')}
       variant={controlButtonVariant}
     >
       {isRunning ? <PauseIcon /> : <PlayIcon />}
@@ -74,6 +76,7 @@ const WorkerLayoutControl = ({ layout, autoRunFor }: WorkerLayoutControlProps) =
  */
 const LayoutsControl = () => {
   const sigma = useSigma()
+  const { t } = useTranslation()
   const [layout, setLayout] = useState<LayoutName>('Circular')
   const [opened, setOpened] = useState<boolean>(false)
 
@@ -149,7 +152,7 @@ const LayoutsControl = () => {
               size="icon"
               variant={controlButtonVariant}
               onClick={() => setOpened((e: boolean) => !e)}
-              tooltip="Layout Graph"
+              tooltip={t('graphPanel.sideBar.layoutsControl.layoutGraph')}
             >
               <GripIcon />
             </Button>
@@ -166,7 +169,7 @@ const LayoutsControl = () => {
                       key={name}
                       className="cursor-pointer text-xs"
                     >
-                      {name}
+                      {t(`graphPanel.sideBar.layoutsControl.layouts.${name}`)}
                     </CommandItem>
                   ))}
                 </CommandGroup>

lightrag_webui/src/components/graph/PropertiesView.tsx

@@ -2,6 +2,7 @@ import { useEffect, useState } from 'react'
 import { useGraphStore, RawNodeType, RawEdgeType } from '@/stores/graph'
 import Text from '@/components/ui/Text'
 import useLightragGraph from '@/hooks/useLightragGraph'
+import { useTranslation } from 'react-i18next'
 
 /**
  * Component that view properties of elements in graph.
@@ -147,21 +148,22 @@ const PropertyRow = ({
 }
 
 const NodePropertiesView = ({ node }: { node: NodeType }) => {
+  const { t } = useTranslation()
   return (
     <div className="flex flex-col gap-2">
-      <label className="text-md pl-1 font-bold tracking-wide text-sky-300">Node</label>
+      <label className="text-md pl-1 font-bold tracking-wide text-sky-300">{t('graphPanel.propertiesView.node.title')}</label>
       <div className="bg-primary/5 max-h-96 overflow-auto rounded p-1">
-        <PropertyRow name={'Id'} value={node.id} />
+        <PropertyRow name={t('graphPanel.propertiesView.node.id')} value={node.id} />
         <PropertyRow
-          name={'Labels'}
+          name={t('graphPanel.propertiesView.node.labels')}
           value={node.labels.join(', ')}
           onClick={() => {
             useGraphStore.getState().setSelectedNode(node.id, true)
           }}
         />
-        <PropertyRow name={'Degree'} value={node.degree} />
+        <PropertyRow name={t('graphPanel.propertiesView.node.degree')} value={node.degree} />
       </div>
-      <label className="text-md pl-1 font-bold tracking-wide text-yellow-400/90">Properties</label>
+      <label className="text-md pl-1 font-bold tracking-wide text-yellow-400/90">{t('graphPanel.propertiesView.node.properties')}</label>
       <div className="bg-primary/5 max-h-96 overflow-auto rounded p-1">
         {Object.keys(node.properties)
           .sort()
@@ -172,7 +174,7 @@ const NodePropertiesView = ({ node }: { node: NodeType }) => {
       {node.relationships.length > 0 && (
         <>
           <label className="text-md pl-1 font-bold tracking-wide text-teal-600/90">
-            Relationships
+          {t('graphPanel.propertiesView.node.relationships')}
           </label>
           <div className="bg-primary/5 max-h-96 overflow-auto rounded p-1">
             {node.relationships.map(({ type, id, label }) => {
@@ -195,28 +197,29 @@ const NodePropertiesView = ({ node }: { node: NodeType }) => {
 }
 
 const EdgePropertiesView = ({ edge }: { edge: EdgeType }) => {
+  const { t } = useTranslation()
   return (
     <div className="flex flex-col gap-2">
-      <label className="text-md pl-1 font-bold tracking-wide text-teal-600">Relationship</label>
+      <label className="text-md pl-1 font-bold tracking-wide text-teal-600">{t('graphPanel.propertiesView.edge.title')}</label>
       <div className="bg-primary/5 max-h-96 overflow-auto rounded p-1">
-        <PropertyRow name={'Id'} value={edge.id} />
+        <PropertyRow name={t('graphPanel.propertiesView.edge.id')} value={edge.id} />
-        {edge.type && <PropertyRow name={'Type'} value={edge.type} />}
+        {edge.type && <PropertyRow name={t('graphPanel.propertiesView.edge.type')} value={edge.type} />}
         <PropertyRow
-          name={'Source'}
+          name={t('graphPanel.propertiesView.edge.source')}
           value={edge.sourceNode ? edge.sourceNode.labels.join(', ') : edge.source}
           onClick={() => {
             useGraphStore.getState().setSelectedNode(edge.source, true)
           }}
         />
         <PropertyRow
-          name={'Target'}
+          name={t('graphPanel.propertiesView.edge.target')}
           value={edge.targetNode ? edge.targetNode.labels.join(', ') : edge.target}
           onClick={() => {
             useGraphStore.getState().setSelectedNode(edge.target, true)
           }}
         />
       </div>
-      <label className="text-md pl-1 font-bold tracking-wide text-yellow-400/90">Properties</label>
+      <label className="text-md pl-1 font-bold tracking-wide text-yellow-400/90">{t('graphPanel.propertiesView.edge.properties')}</label>
       <div className="bg-primary/5 max-h-96 overflow-auto rounded p-1">
         {Object.keys(edge.properties)
           .sort()

lightrag_webui/src/components/graph/Settings.tsx

@@ -10,6 +10,7 @@ import { useSettingsStore } from '@/stores/settings'
 import { useBackendState } from '@/stores/state'
 
 import { SettingsIcon } from 'lucide-react'
+import { useTranslation } from "react-i18next";
 
 /**
  * Component that displays a checkbox with a label.
@@ -204,10 +205,12 @@ export default function Settings() {
     [setTempApiKey]
   )
 
+  const { t } = useTranslation();
+
   return (
     <Popover open={opened} onOpenChange={setOpened}>
       <PopoverTrigger asChild>
-        <Button variant={controlButtonVariant} tooltip="Settings" size="icon">
+        <Button variant={controlButtonVariant} tooltip={t("graphPanel.sideBar.settings.settings")} size="icon">
           <SettingsIcon />
         </Button>
       </PopoverTrigger>
@@ -221,7 +224,7 @@ export default function Settings() {
           <LabeledCheckBox
             checked={enableHealthCheck}
             onCheckedChange={setEnableHealthCheck}
-            label="Health Check"
+            label={t("graphPanel.sideBar.settings.healthCheck")}
           />
 
           <Separator />
@@ -229,12 +232,12 @@ export default function Settings() {
           <LabeledCheckBox
             checked={showPropertyPanel}
             onCheckedChange={setShowPropertyPanel}
-            label="Show Property Panel"
+            label={t("graphPanel.sideBar.settings.showPropertyPanel")}
           />
           <LabeledCheckBox
             checked={showNodeSearchBar}
             onCheckedChange={setShowNodeSearchBar}
-            label="Show Search Bar"
+            label={t("graphPanel.sideBar.settings.showSearchBar")}
           />
 
           <Separator />
@@ -242,12 +245,12 @@ export default function Settings() {
           <LabeledCheckBox
             checked={showNodeLabel}
             onCheckedChange={setShowNodeLabel}
-            label="Show Node Label"
+            label={t("graphPanel.sideBar.settings.showNodeLabel")}
           />
           <LabeledCheckBox
             checked={enableNodeDrag}
             onCheckedChange={setEnableNodeDrag}
-            label="Node Draggable"
+            label={t("graphPanel.sideBar.settings.nodeDraggable")}
           />
 
           <Separator />
@@ -255,51 +258,50 @@ export default function Settings() {
           <LabeledCheckBox
             checked={showEdgeLabel}
             onCheckedChange={setShowEdgeLabel}
-            label="Show Edge Label"
+            label={t("graphPanel.sideBar.settings.showEdgeLabel")}
           />
           <LabeledCheckBox
             checked={enableHideUnselectedEdges}
             onCheckedChange={setEnableHideUnselectedEdges}
-            label="Hide Unselected Edges"
+            label={t("graphPanel.sideBar.settings.hideUnselectedEdges")}
           />
           <LabeledCheckBox
             checked={enableEdgeEvents}
             onCheckedChange={setEnableEdgeEvents}
-            label="Edge Events"
+            label={t("graphPanel.sideBar.settings.edgeEvents")}
           />
 
           <Separator />
           <LabeledNumberInput
-            label="Max Query Depth"
+            label={t("graphPanel.sideBar.settings.maxQueryDepth")}
             min={1}
             value={graphQueryMaxDepth}
             onEditFinished={setGraphQueryMaxDepth}
           />
           <LabeledNumberInput
-            label="Minimum Degree"
+            label={t("graphPanel.sideBar.settings.minDegree")}
             min={0}
             value={graphMinDegree}
             onEditFinished={setGraphMinDegree}
           />
           <LabeledNumberInput
-            label="Max Layout Iterations"
+            label={t("graphPanel.sideBar.settings.maxLayoutIterations")}
             min={1}
             max={20}
             value={graphLayoutMaxIterations}
             onEditFinished={setGraphLayoutMaxIterations}
           />
-
           <Separator />
 
           <div className="flex flex-col gap-2">
-            <label className="text-sm font-medium">API Key</label>
+            <label className="text-sm font-medium">{t("graphPanel.sideBar.settings.apiKey")}</label>
             <form className="flex h-6 gap-2" onSubmit={(e) => e.preventDefault()}>
               <div className="w-0 flex-1">
                 <Input
                   type="password"
                   value={tempApiKey}
                   onChange={handleTempApiKeyChange}
-                  placeholder="Enter your API key"
+                  placeholder={t("graphPanel.sideBar.settings.enterYourAPIkey")}
                   className="max-h-full w-full min-w-0"
                   autoComplete="off"
                 />
@@ -310,7 +312,7 @@ export default function Settings() {
                 size="sm"
                 className="max-h-full shrink-0"
               >
-                Save
+                {t("graphPanel.sideBar.settings.save")}
               </Button>
             </form>
           </div>

lightrag_webui/src/components/graph/StatusCard.tsx

@@ -1,58 +1,60 @@
 import { LightragStatus } from '@/api/lightrag'
+import { useTranslation } from 'react-i18next'
 
 const StatusCard = ({ status }: { status: LightragStatus | null }) => {
+  const { t } = useTranslation()
   if (!status) {
-    return <div className="text-muted-foreground text-sm">Status information unavailable</div>
+    return <div className="text-muted-foreground text-sm">{t('graphPanel.statusCard.unavailable')}</div>
   }
 
   return (
     <div className="min-w-[300px] space-y-3 text-sm">
       <div className="space-y-1">
-        <h4 className="font-medium">Storage Info</h4>
+        <h4 className="font-medium">{t('graphPanel.statusCard.storageInfo')}</h4>
         <div className="text-muted-foreground grid grid-cols-2 gap-1">
-          <span>Working Directory:</span>
+          <span>{t('graphPanel.statusCard.workingDirectory')}:</span>
           <span className="truncate">{status.working_directory}</span>
-          <span>Input Directory:</span>
+          <span>{t('graphPanel.statusCard.inputDirectory')}:</span>
           <span className="truncate">{status.input_directory}</span>
         </div>
       </div>
 
       <div className="space-y-1">
-        <h4 className="font-medium">LLM Configuration</h4>
+        <h4 className="font-medium">{t('graphPanel.statusCard.llmConfig')}</h4>
         <div className="text-muted-foreground grid grid-cols-2 gap-1">
-          <span>LLM Binding:</span>
+          <span>{t('graphPanel.statusCard.llmBinding')}:</span>
           <span>{status.configuration.llm_binding}</span>
-          <span>LLM Binding Host:</span>
+          <span>{t('graphPanel.statusCard.llmBindingHost')}:</span>
           <span>{status.configuration.llm_binding_host}</span>
-          <span>LLM Model:</span>
+          <span>{t('graphPanel.statusCard.llmModel')}:</span>
           <span>{status.configuration.llm_model}</span>
-          <span>Max Tokens:</span>
+          <span>{t('graphPanel.statusCard.maxTokens')}:</span>
           <span>{status.configuration.max_tokens}</span>
         </div>
       </div>
 
       <div className="space-y-1">
-        <h4 className="font-medium">Embedding Configuration</h4>
+        <h4 className="font-medium">{t('graphPanel.statusCard.embeddingConfig')}</h4>
         <div className="text-muted-foreground grid grid-cols-2 gap-1">
-          <span>Embedding Binding:</span>
+          <span>{t('graphPanel.statusCard.embeddingBinding')}:</span>
           <span>{status.configuration.embedding_binding}</span>
-          <span>Embedding Binding Host:</span>
+          <span>{t('graphPanel.statusCard.embeddingBindingHost')}:</span>
           <span>{status.configuration.embedding_binding_host}</span>
-          <span>Embedding Model:</span>
+          <span>{t('graphPanel.statusCard.embeddingModel')}:</span>
           <span>{status.configuration.embedding_model}</span>
         </div>
       </div>
 
       <div className="space-y-1">
-        <h4 className="font-medium">Storage Configuration</h4>
+        <h4 className="font-medium">{t('graphPanel.statusCard.storageConfig')}</h4>
         <div className="text-muted-foreground grid grid-cols-2 gap-1">
-          <span>KV Storage:</span>
+          <span>{t('graphPanel.statusCard.kvStorage')}:</span>
           <span>{status.configuration.kv_storage}</span>
-          <span>Doc Status Storage:</span>
+          <span>{t('graphPanel.statusCard.docStatusStorage')}:</span>
           <span>{status.configuration.doc_status_storage}</span>
-          <span>Graph Storage:</span>
+          <span>{t('graphPanel.statusCard.graphStorage')}:</span>
           <span>{status.configuration.graph_storage}</span>
-          <span>Vector Storage:</span>
+          <span>{t('graphPanel.statusCard.vectorStorage')}:</span>
           <span>{status.configuration.vector_storage}</span>
         </div>
       </div>

lightrag_webui/src/components/graph/StatusIndicator.tsx

@@ -3,8 +3,10 @@ import { useBackendState } from '@/stores/state'
 import { useEffect, useState } from 'react'
 import { Popover, PopoverContent, PopoverTrigger } from '@/components/ui/Popover'
 import StatusCard from '@/components/graph/StatusCard'
+import { useTranslation } from 'react-i18next'
 
 const StatusIndicator = () => {
+  const { t } = useTranslation()
   const health = useBackendState.use.health()
   const lastCheckTime = useBackendState.use.lastCheckTime()
   const status = useBackendState.use.status()
@@ -33,7 +35,7 @@ const StatusIndicator = () => {
               )}
             />
             <span className="text-muted-foreground text-xs">
-              {health ? 'Connected' : 'Disconnected'}
+              {health ? t('graphPanel.statusIndicator.connected') : t('graphPanel.statusIndicator.disconnected')}
             </span>
           </div>
         </PopoverTrigger>

lightrag_webui/src/components/graph/ZoomControl.tsx

@@ -3,12 +3,14 @@ import { useCallback } from 'react'
 import Button from '@/components/ui/Button'
 import { ZoomInIcon, ZoomOutIcon, FullscreenIcon } from 'lucide-react'
 import { controlButtonVariant } from '@/lib/constants'
+import { useTranslation } from "react-i18next";
 
 /**
  * Component that provides zoom controls for the graph viewer.
  */
 const ZoomControl = () => {
   const { zoomIn, zoomOut, reset } = useCamera({ duration: 200, factor: 1.5 })
+  const { t } = useTranslation();
 
   const handleZoomIn = useCallback(() => zoomIn(), [zoomIn])
   const handleZoomOut = useCallback(() => zoomOut(), [zoomOut])
@@ -16,16 +18,16 @@ const ZoomControl = () => {
 
   return (
     <>
-      <Button variant={controlButtonVariant} onClick={handleZoomIn} tooltip="Zoom In" size="icon">
+      <Button variant={controlButtonVariant} onClick={handleZoomIn} tooltip={t("graphPanel.sideBar.zoomControl.zoomIn")} size="icon">
         <ZoomInIcon />
       </Button>
-      <Button variant={controlButtonVariant} onClick={handleZoomOut} tooltip="Zoom Out" size="icon">
+      <Button variant={controlButtonVariant} onClick={handleZoomOut} tooltip={t("graphPanel.sideBar.zoomControl.zoomOut")} size="icon">
         <ZoomOutIcon />
       </Button>
       <Button
         variant={controlButtonVariant}
         onClick={handleResetZoom}
-        tooltip="Reset Zoom"
+        tooltip={t("graphPanel.sideBar.zoomControl.resetZoom")}
         size="icon"
       >
         <FullscreenIcon />

lightrag_webui/src/components/retrieval/ChatMessage.tsx

@@ -15,18 +15,21 @@ import { Prism as SyntaxHighlighter } from 'react-syntax-highlighter'
 import { oneLight, oneDark } from 'react-syntax-highlighter/dist/cjs/styles/prism'
 
 import { LoaderIcon, CopyIcon } from 'lucide-react'
+import { useTranslation } from 'react-i18next'
 
 export type MessageWithError = Message & {
   isError?: boolean
 }
 
 export const ChatMessage = ({ message }: { message: MessageWithError }) => {
+  const { t } = useTranslation()
+
   const handleCopyMarkdown = useCallback(async () => {
     if (message.content) {
       try {
         await navigator.clipboard.writeText(message.content)
       } catch (err) {
-        console.error('Failed to copy:', err)
+        console.error(t('chat.copyError'), err)
       }
     }
   }, [message])
@@ -57,7 +60,7 @@ export const ChatMessage = ({ message }: { message: MessageWithError }) => {
           <Button
             onClick={handleCopyMarkdown}
             className="absolute right-0 bottom-0 size-6 rounded-md opacity-20 transition-opacity hover:opacity-100"
-            tooltip="Copy to clipboard"
+            tooltip={t('retrievePanel.chatMessage.copyTooltip')}
             variant="default"
             size="icon"
           >

lightrag_webui/src/components/retrieval/QuerySettings.tsx

@@ -14,8 +14,10 @@ import {
   SelectValue
 } from '@/components/ui/Select'
 import { useSettingsStore } from '@/stores/settings'
+import { useTranslation } from 'react-i18next'
 
 export default function QuerySettings() {
+  const { t } = useTranslation()
   const querySettings = useSettingsStore((state) => state.querySettings)
 
   const handleChange = useCallback((key: keyof QueryRequest, value: any) => {
@@ -25,8 +27,8 @@ export default function QuerySettings() {
   return (
     <Card className="flex shrink-0 flex-col">
       <CardHeader className="px-4 pt-4 pb-2">
-        <CardTitle>Parameters</CardTitle>
+        <CardTitle>{t('retrievePanel.querySettings.parametersTitle')}</CardTitle>
-        <CardDescription>Configure your query parameters</CardDescription>
+        <CardDescription>{t('retrievePanel.querySettings.parametersDescription')}</CardDescription>
       </CardHeader>
       <CardContent className="m-0 flex grow flex-col p-0 text-xs">
         <div className="relative size-full">
@@ -35,8 +37,8 @@ export default function QuerySettings() {
             <>
               <Text
                 className="ml-1"
-                text="Query Mode"
+                text={t('retrievePanel.querySettings.queryMode')}
-                tooltip="Select the retrieval strategy:\n• Naive: Basic search without advanced techniques\n• Local: Context-dependent information retrieval\n• Global: Utilizes global knowledge base\n• Hybrid: Combines local and global retrieval\n• Mix: Integrates knowledge graph with vector retrieval"
+                tooltip={t('retrievePanel.querySettings.queryModeTooltip')}
                 side="left"
               />
               <Select
@@ -48,11 +50,11 @@ export default function QuerySettings() {
                 </SelectTrigger>
                 <SelectContent>
                   <SelectGroup>
-                    <SelectItem value="naive">Naive</SelectItem>
+                    <SelectItem value="naive">{t('retrievePanel.querySettings.queryModeOptions.naive')}</SelectItem>
-                    <SelectItem value="local">Local</SelectItem>
+                    <SelectItem value="local">{t('retrievePanel.querySettings.queryModeOptions.local')}</SelectItem>
-                    <SelectItem value="global">Global</SelectItem>
+                    <SelectItem value="global">{t('retrievePanel.querySettings.queryModeOptions.global')}</SelectItem>
-                    <SelectItem value="hybrid">Hybrid</SelectItem>
+                    <SelectItem value="hybrid">{t('retrievePanel.querySettings.queryModeOptions.hybrid')}</SelectItem>
-                    <SelectItem value="mix">Mix</SelectItem>
+                    <SelectItem value="mix">{t('retrievePanel.querySettings.queryModeOptions.mix')}</SelectItem>
                   </SelectGroup>
                 </SelectContent>
               </Select>
@@ -62,8 +64,8 @@ export default function QuerySettings() {
             <>
               <Text
                 className="ml-1"
-                text="Response Format"
+                text={t('retrievePanel.querySettings.responseFormat')}
-                tooltip="Defines the response format. Examples:\n• Multiple Paragraphs\n• Single Paragraph\n• Bullet Points"
+                tooltip={t('retrievePanel.querySettings.responseFormatTooltip')}
                 side="left"
               />
               <Select
@@ -75,9 +77,9 @@ export default function QuerySettings() {
                 </SelectTrigger>
                 <SelectContent>
                   <SelectGroup>
-                    <SelectItem value="Multiple Paragraphs">Multiple Paragraphs</SelectItem>
+                    <SelectItem value="Multiple Paragraphs">{t('retrievePanel.querySettings.responseFormatOptions.multipleParagraphs')}</SelectItem>
-                    <SelectItem value="Single Paragraph">Single Paragraph</SelectItem>
+                    <SelectItem value="Single Paragraph">{t('retrievePanel.querySettings.responseFormatOptions.singleParagraph')}</SelectItem>
-                    <SelectItem value="Bullet Points">Bullet Points</SelectItem>
+                    <SelectItem value="Bullet Points">{t('retrievePanel.querySettings.responseFormatOptions.bulletPoints')}</SelectItem>
                   </SelectGroup>
                 </SelectContent>
               </Select>
@@ -87,8 +89,8 @@ export default function QuerySettings() {
             <>
               <Text
                 className="ml-1"
-                text="Top K Results"
+                text={t('retrievePanel.querySettings.topK')}
-                tooltip="Number of top items to retrieve. Represents entities in 'local' mode and relationships in 'global' mode"
+                tooltip={t('retrievePanel.querySettings.topKTooltip')}
                 side="left"
               />
               <NumberInput
@@ -97,7 +99,7 @@ export default function QuerySettings() {
                 value={querySettings.top_k}
                 onValueChange={(v) => handleChange('top_k', v)}
                 min={1}
-                placeholder="Number of results"
+                placeholder={t('retrievePanel.querySettings.topKPlaceholder')}
               />
             </>
 
@@ -106,8 +108,8 @@ export default function QuerySettings() {
               <>
                 <Text
                   className="ml-1"
-                  text="Max Tokens for Text Unit"
+                  text={t('retrievePanel.querySettings.maxTokensTextUnit')}
-                  tooltip="Maximum number of tokens allowed for each retrieved text chunk"
+                  tooltip={t('retrievePanel.querySettings.maxTokensTextUnitTooltip')}
                   side="left"
                 />
                 <NumberInput
@@ -116,14 +118,14 @@ export default function QuerySettings() {
                   value={querySettings.max_token_for_text_unit}
                   onValueChange={(v) => handleChange('max_token_for_text_unit', v)}
                   min={1}
-                  placeholder="Max tokens for text unit"
+                  placeholder={t('retrievePanel.querySettings.maxTokensTextUnit')}
                 />
               </>
 
               <>
                 <Text
-                  text="Max Tokens for Global Context"
+                  text={t('retrievePanel.querySettings.maxTokensGlobalContext')}
-                  tooltip="Maximum number of tokens allocated for relationship descriptions in global retrieval"
+                  tooltip={t('retrievePanel.querySettings.maxTokensGlobalContextTooltip')}
                   side="left"
                 />
                 <NumberInput
@@ -132,15 +134,15 @@ export default function QuerySettings() {
                   value={querySettings.max_token_for_global_context}
                   onValueChange={(v) => handleChange('max_token_for_global_context', v)}
                   min={1}
-                  placeholder="Max tokens for global context"
+                  placeholder={t('retrievePanel.querySettings.maxTokensGlobalContext')}
                 />
               </>
 
               <>
                 <Text
                   className="ml-1"
-                  text="Max Tokens for Local Context"
+                  text={t('retrievePanel.querySettings.maxTokensLocalContext')}
-                  tooltip="Maximum number of tokens allocated for entity descriptions in local retrieval"
+                  tooltip={t('retrievePanel.querySettings.maxTokensLocalContextTooltip')}
                   side="left"
                 />
                 <NumberInput
@@ -149,7 +151,7 @@ export default function QuerySettings() {
                   value={querySettings.max_token_for_local_context}
                   onValueChange={(v) => handleChange('max_token_for_local_context', v)}
                   min={1}
-                  placeholder="Max tokens for local context"
+                  placeholder={t('retrievePanel.querySettings.maxTokensLocalContext')}
                 />
               </>
             </>
@@ -158,8 +160,8 @@ export default function QuerySettings() {
             <>
               <Text
                 className="ml-1"
-                text="History Turns"
+                text={t('retrievePanel.querySettings.historyTurns')}
-                tooltip="Number of complete conversation turns (user-assistant pairs) to consider in the response context"
+                tooltip={t('retrievePanel.querySettings.historyTurnsTooltip')}
                 side="left"
               />
               <NumberInput
@@ -170,7 +172,7 @@ export default function QuerySettings() {
                 value={querySettings.history_turns}
                 onValueChange={(v) => handleChange('history_turns', v)}
                 min={0}
-                placeholder="Number of history turns"
+                placeholder={t('retrievePanel.querySettings.historyTurnsPlaceholder')}
               />
             </>
 
@@ -179,8 +181,8 @@ export default function QuerySettings() {
               <>
                 <Text
                   className="ml-1"
-                  text="High-Level Keywords"
+                  text={t('retrievePanel.querySettings.hlKeywords')}
-                  tooltip="List of high-level keywords to prioritize in retrieval. Separate with commas"
+                  tooltip={t('retrievePanel.querySettings.hlKeywordsTooltip')}
                   side="left"
                 />
                 <Input
@@ -194,15 +196,15 @@ export default function QuerySettings() {
                       .filter((k) => k !== '')
                     handleChange('hl_keywords', keywords)
                   }}
-                  placeholder="Enter keywords"
+                  placeholder={t('retrievePanel.querySettings.hlkeywordsPlaceHolder')}
                 />
               </>
 
               <>
                 <Text
                   className="ml-1"
-                  text="Low-Level Keywords"
+                  text={t('retrievePanel.querySettings.llKeywords')}
-                  tooltip="List of low-level keywords to refine retrieval focus. Separate with commas"
+                  tooltip={t('retrievePanel.querySettings.llKeywordsTooltip')}
                   side="left"
                 />
                 <Input
@@ -216,7 +218,7 @@ export default function QuerySettings() {
                       .filter((k) => k !== '')
                     handleChange('ll_keywords', keywords)
                   }}
-                  placeholder="Enter keywords"
+                  placeholder={t('retrievePanel.querySettings.hlkeywordsPlaceHolder')}
                 />
               </>
             </>
@@ -226,8 +228,8 @@ export default function QuerySettings() {
               <div className="flex items-center gap-2">
                 <Text
                   className="ml-1"
-                  text="Only Need Context"
+                  text={t('retrievePanel.querySettings.onlyNeedContext')}
-                  tooltip="If True, only returns the retrieved context without generating a response"
+                  tooltip={t('retrievePanel.querySettings.onlyNeedContextTooltip')}
                   side="left"
                 />
                 <div className="grow" />
@@ -242,8 +244,8 @@ export default function QuerySettings() {
               <div className="flex items-center gap-2">
                 <Text
                   className="ml-1"
-                  text="Only Need Prompt"
+                  text={t('retrievePanel.querySettings.onlyNeedPrompt')}
-                  tooltip="If True, only returns the generated prompt without producing a response"
+                  tooltip={t('retrievePanel.querySettings.onlyNeedPromptTooltip')}
                   side="left"
                 />
                 <div className="grow" />
@@ -258,8 +260,8 @@ export default function QuerySettings() {
               <div className="flex items-center gap-2">
                 <Text
                   className="ml-1"
-                  text="Stream Response"
+                  text={t('retrievePanel.querySettings.streamResponse')}
-                  tooltip="If True, enables streaming output for real-time responses"
+                  tooltip={t('retrievePanel.querySettings.streamResponseTooltip')}
                   side="left"
                 />
                 <div className="grow" />

lightrag_webui/src/features/DocumentManager.tsx

@@ -1,4 +1,5 @@
 import { useState, useEffect, useCallback } from 'react'
+import { useTranslation } from 'react-i18next'
 import Button from '@/components/ui/Button'
 import {
   Table,
@@ -22,6 +23,7 @@ import { useBackendState } from '@/stores/state'
 import { RefreshCwIcon } from 'lucide-react'
 
 export default function DocumentManager() {
+  const { t } = useTranslation()
   const health = useBackendState.use.health()
   const [docs, setDocs] = useState<DocsStatusesResponse | null>(null)
 
@@ -44,7 +46,7 @@ export default function DocumentManager() {
         setDocs(null)
       }
     } catch (err) {
-      toast.error('Failed to load documents\n' + errorMessage(err))
+      toast.error(t('documentPanel.documentManager.errors.loadFailed', { error: errorMessage(err) }))
     }
   }, [setDocs])
 
@@ -57,7 +59,7 @@ export default function DocumentManager() {
       const { status } = await scanNewDocuments()
       toast.message(status)
     } catch (err) {
-      toast.error('Failed to load documents\n' + errorMessage(err))
+      toast.error(t('documentPanel.documentManager.errors.scanFailed', { error: errorMessage(err) }))
     }
   }, [])
 
@@ -69,7 +71,7 @@ export default function DocumentManager() {
       try {
         await fetchDocuments()
       } catch (err) {
-        toast.error('Failed to get scan progress\n' + errorMessage(err))
+        toast.error(t('documentPanel.documentManager.errors.scanProgressFailed', { error: errorMessage(err) }))
       }
     }, 5000)
     return () => clearInterval(interval)
@@ -78,7 +80,7 @@ export default function DocumentManager() {
   return (
     <Card className="!size-full !rounded-none !border-none">
       <CardHeader>
-        <CardTitle className="text-lg">Document Management</CardTitle>
+        <CardTitle className="text-lg">{t('documentPanel.documentManager.title')}</CardTitle>
       </CardHeader>
       <CardContent className="space-y-4">
         <div className="flex gap-2">
@@ -86,10 +88,10 @@ export default function DocumentManager() {
             variant="outline"
             onClick={scanDocuments}
             side="bottom"
-            tooltip="Scan documents"
+            tooltip={t('documentPanel.documentManager.scanTooltip')}
             size="sm"
           >
-            <RefreshCwIcon /> Scan
+            <RefreshCwIcon /> {t('documentPanel.documentManager.scanButton')}
           </Button>
           <div className="flex-1" />
           <ClearDocumentsDialog />
@@ -98,29 +100,29 @@ export default function DocumentManager() {
 
         <Card>
           <CardHeader>
-            <CardTitle>Uploaded documents</CardTitle>
+            <CardTitle>{t('documentPanel.documentManager.uploadedTitle')}</CardTitle>
-            <CardDescription>view the uploaded documents here</CardDescription>
+            <CardDescription>{t('documentPanel.documentManager.uploadedDescription')}</CardDescription>
           </CardHeader>
 
           <CardContent>
             {!docs && (
               <EmptyCard
-                title="No documents uploaded"
+                title={t('documentPanel.documentManager.emptyTitle')}
-                description="upload documents to see them here"
+                description={t('documentPanel.documentManager.emptyDescription')}
               />
             )}
             {docs && (
               <Table>
                 <TableHeader>
                   <TableRow>
-                    <TableHead>ID</TableHead>
+                    <TableHead>{t('documentPanel.documentManager.columns.id')}</TableHead>
-                    <TableHead>Summary</TableHead>
+                    <TableHead>{t('documentPanel.documentManager.columns.summary')}</TableHead>
-                    <TableHead>Status</TableHead>
+                    <TableHead>{t('documentPanel.documentManager.columns.status')}</TableHead>
-                    <TableHead>Length</TableHead>
+                    <TableHead>{t('documentPanel.documentManager.columns.length')}</TableHead>
-                    <TableHead>Chunks</TableHead>
+                    <TableHead>{t('documentPanel.documentManager.columns.chunks')}</TableHead>
-                    <TableHead>Created</TableHead>
+                    <TableHead>{t('documentPanel.documentManager.columns.created')}</TableHead>
-                    <TableHead>Updated</TableHead>
+                    <TableHead>{t('documentPanel.documentManager.columns.updated')}</TableHead>
-                    <TableHead>Metadata</TableHead>
+                    <TableHead>{t('documentPanel.documentManager.columns.metadata')}</TableHead>
                   </TableRow>
                 </TableHeader>
                 <TableBody className="text-sm">
@@ -137,13 +139,13 @@ export default function DocumentManager() {
                         </TableCell>
                         <TableCell>
                           {status === 'processed' && (
-                            <span className="text-green-600">Completed</span>
+                            <span className="text-green-600">{t('documentPanel.documentManager.status.completed')}</span>
                           )}
                           {status === 'processing' && (
-                            <span className="text-blue-600">Processing</span>
+                            <span className="text-blue-600">{t('documentPanel.documentManager.status.processing')}</span>
                           )}
-                          {status === 'pending' && <span className="text-yellow-600">Pending</span>}
+                          {status === 'pending' && <span className="text-yellow-600">{t('documentPanel.documentManager.status.pending')}</span>}
-                          {status === 'failed' && <span className="text-red-600">Failed</span>}
+                          {status === 'failed' && <span className="text-red-600">{t('documentPanel.documentManager.status.failed')}</span>}
                           {doc.error && (
                             <span className="ml-2 text-red-500" title={doc.error}>
                               ⚠️

lightrag_webui/src/features/RetrievalTesting.tsx

@@ -8,8 +8,10 @@ import { useDebounce } from '@/hooks/useDebounce'
 import QuerySettings from '@/components/retrieval/QuerySettings'
 import { ChatMessage, MessageWithError } from '@/components/retrieval/ChatMessage'
 import { EraserIcon, SendIcon } from 'lucide-react'
+import { useTranslation } from 'react-i18next'
 
 export default function RetrievalTesting() {
+  const { t } = useTranslation()
   const [messages, setMessages] = useState<MessageWithError[]>(
     () => useSettingsStore.getState().retrievalHistory || []
   )
@@ -89,7 +91,7 @@ export default function RetrievalTesting() {
         }
       } catch (err) {
         // Handle error
-        updateAssistantMessage(`Error: Failed to get response\n${errorMessage(err)}`, true)
+        updateAssistantMessage(`${t('retrievePanel.retrieval.error')}\n${errorMessage(err)}`, true)
       } finally {
         // Clear loading and add messages to state
         setIsLoading(false)
@@ -98,7 +100,7 @@ export default function RetrievalTesting() {
           .setRetrievalHistory([...prevMessages, userMessage, assistantMessage])
       }
     },
-    [inputValue, isLoading, messages, setMessages]
+    [inputValue, isLoading, messages, setMessages, t]
   )
 
   const debouncedMessages = useDebounce(messages, 100)
@@ -117,7 +119,7 @@ export default function RetrievalTesting() {
             <div className="flex min-h-0 flex-1 flex-col gap-2">
               {messages.length === 0 ? (
                 <div className="text-muted-foreground flex h-full items-center justify-center text-lg">
-                  Start a retrieval by typing your query below
+                  {t('retrievePanel.retrieval.startPrompt')}
                 </div>
               ) : (
                 messages.map((message, idx) => (
@@ -143,18 +145,18 @@ export default function RetrievalTesting() {
             size="sm"
           >
             <EraserIcon />
-            Clear
+            {t('retrievePanel.retrieval.clear')}
           </Button>
           <Input
             className="flex-1"
             value={inputValue}
             onChange={(e) => setInputValue(e.target.value)}
-            placeholder="Type your query..."
+            placeholder={t('retrievePanel.retrieval.placeholder')}
             disabled={isLoading}
           />
           <Button type="submit" variant="default" disabled={isLoading} size="sm">
             <SendIcon />
-            Send
+            {t('retrievePanel.retrieval.send')}
           </Button>
         </form>
       </div>

lightrag_webui/src/features/SiteHeader.tsx

@@ -4,6 +4,7 @@ import ThemeToggle from '@/components/ThemeToggle'
 import { TabsList, TabsTrigger } from '@/components/ui/Tabs'
 import { useSettingsStore } from '@/stores/settings'
 import { cn } from '@/lib/utils'
+import { useTranslation } from 'react-i18next'
 
 import { ZapIcon, GithubIcon } from 'lucide-react'
 
@@ -29,21 +30,22 @@ function NavigationTab({ value, currentTab, children }: NavigationTabProps) {
 
 function TabsNavigation() {
   const currentTab = useSettingsStore.use.currentTab()
+  const { t } = useTranslation()
 
   return (
     <div className="flex h-8 self-center">
       <TabsList className="h-full gap-2">
         <NavigationTab value="documents" currentTab={currentTab}>
-          Documents
+          {t('header.documents')}
         </NavigationTab>
         <NavigationTab value="knowledge-graph" currentTab={currentTab}>
-          Knowledge Graph
+          {t('header.knowledgeGraph')}
         </NavigationTab>
         <NavigationTab value="retrieval" currentTab={currentTab}>
-          Retrieval
+          {t('header.retrieval')}
         </NavigationTab>
         <NavigationTab value="api" currentTab={currentTab}>
-          API
+          {t('header.api')}
         </NavigationTab>
       </TabsList>
     </div>
@@ -51,6 +53,7 @@ function TabsNavigation() {
 }
 
 export default function SiteHeader() {
+  const { t } = useTranslation()
   return (
     <header className="border-border/40 bg-background/95 supports-[backdrop-filter]:bg-background/60 sticky top-0 z-50 flex h-10 w-full border-b px-4 backdrop-blur">
       <a href="/" className="mr-6 flex items-center gap-2">
@@ -64,7 +67,7 @@ export default function SiteHeader() {
       </div>
 
       <nav className="flex items-center">
-        <Button variant="ghost" size="icon" side="bottom" tooltip="Project Repository">
+        <Button variant="ghost" size="icon" side="bottom" tooltip={t('header.projectRepository')}>
           <a href={SiteInfo.github} target="_blank" rel="noopener noreferrer">
             <GithubIcon className="size-4" aria-hidden="true" />
           </a>

lightrag_webui/src/i18n.js

@@ -0,0 +1,21 @@
+import i18n from "i18next";
+import { initReactI18next } from "react-i18next";
+
+import en from "./locales/en.json";
+import zh from "./locales/zh.json";
+
+i18n
+  .use(initReactI18next)
+  .init({
+    resources: {
+      en: { translation: en },
+      zh: { translation: zh }
+    },
+    lng: "en", // default
+    fallbackLng: "en",
+    interpolation: {
+      escapeValue: false
+    }
+  });
+
+export default i18n;

lightrag_webui/src/locales/en.json

@@ -0,0 +1,234 @@
+{
+  "header": {
+    "documents": "Documents",
+    "knowledgeGraph": "Knowledge Graph",
+    "retrieval": "Retrieval",
+    "api": "API",
+    "projectRepository": "Project Repository",
+    "themeToggle": {
+      "switchToLight": "Switch to light theme",
+      "switchToDark": "Switch to dark theme"
+    }
+  },
+  "documentPanel": {
+    "clearDocuments": {
+      "button": "Clear",
+      "tooltip": "Clear documents",
+      "title": "Clear Documents",
+      "confirm": "Do you really want to clear all documents?",
+      "confirmButton": "YES",
+      "success": "Documents cleared successfully",
+      "failed": "Clear Documents Failed:\n{{message}}",
+      "error": "Clear Documents Failed:\n{{error}}"
+    },
+    "uploadDocuments": {
+      "button": "Upload",
+      "tooltip": "Upload documents",
+      "title": "Upload Documents",
+      "description": "Drag and drop your documents here or click to browse.",
+      "uploading": "Uploading {{name}}: {{percent}}%",
+      "success": "Upload Success:\n{{name}} uploaded successfully",
+      "failed": "Upload Failed:\n{{name}}\n{{message}}",
+      "error": "Upload Failed:\n{{name}}\n{{error}}",
+      "generalError": "Upload Failed\n{{error}}",
+      "fileTypes": "Supported types: TXT, MD, DOCX, PDF, PPTX, RTF, ODT, EPUB, HTML, HTM, TEX, JSON, XML, YAML, YML, CSV, LOG, CONF, INI, PROPERTIES, SQL, BAT, SH, C, CPP, PY, JAVA, JS, TS, SWIFT, GO, RB, PHP, CSS, SCSS, LESS"
+    },
+    "documentManager": {
+      "title": "Document Management",
+      "scanButton": "Scan",
+      "scanTooltip": "Scan documents",
+      "uploadedTitle": "Uploaded Documents",
+      "uploadedDescription": "List of uploaded documents and their statuses.",
+      "emptyTitle": "No Documents",
+      "emptyDescription": "There are no uploaded documents yet.",
+      "columns": {
+        "id": "ID",
+        "summary": "Summary",
+        "status": "Status",
+        "length": "Length",
+        "chunks": "Chunks",
+        "created": "Created",
+        "updated": "Updated",
+        "metadata": "Metadata"
+      },
+      "status": {
+        "completed": "Completed",
+        "processing": "Processing",
+        "pending": "Pending",
+        "failed": "Failed"
+      },
+      "errors": {
+        "loadFailed": "Failed to load documents\n{{error}}",
+        "scanFailed": "Failed to scan documents\n{{error}}",
+        "scanProgressFailed": "Failed to get scan progress\n{{error}}"
+      }
+    }
+  },
+  "graphPanel": {
+    "sideBar": {
+      "settings": {
+        "settings": "Settings",
+        "healthCheck": "Health Check",
+        "showPropertyPanel": "Show Property Panel",
+        "showSearchBar": "Show Search Bar",
+        "showNodeLabel": "Show Node Label",
+        "nodeDraggable": "Node Draggable",
+        "showEdgeLabel": "Show Edge Label",
+        "hideUnselectedEdges": "Hide Unselected Edges",
+        "edgeEvents": "Edge Events",
+        "maxQueryDepth": "Max Query Depth",
+        "minDegree": "Minimum Degree",
+        "maxLayoutIterations": "Max Layout Iterations",
+        "apiKey": "API Key",
+        "enterYourAPIkey": "Enter your API key",
+        "save": "Save"
+      },
+
+      "zoomControl": {
+        "zoomIn": "Zoom In",
+        "zoomOut": "Zoom Out",
+        "resetZoom": "Reset Zoom"
+      },
+
+      "layoutsControl": {
+        "startAnimation": "Start the layout animation",
+        "stopAnimation": "Stop the layout animation",
+        "layoutGraph": "Layout Graph",
+        "layouts": {
+          "Circular": "Circular",
+          "Circlepack": "Circlepack",
+          "Random": "Random",
+          "Noverlaps": "Noverlaps",
+          "Force Directed": "Force Directed",
+          "Force Atlas": "Force Atlas"
+        }
+      },
+
+      "fullScreenControl": {
+        "fullScreen": "Full Screen",
+        "windowed": "Windowed"
+      }
+    },
+    "statusIndicator": {
+      "connected": "Connected",
+      "disconnected": "Disconnected"
+    },
+    "statusCard": {
+      "unavailable": "Status information unavailable",
+      "storageInfo": "Storage Info",
+      "workingDirectory": "Working Directory",
+      "inputDirectory": "Input Directory",
+      "llmConfig": "LLM Configuration",
+      "llmBinding": "LLM Binding",
+      "llmBindingHost": "LLM Binding Host",
+      "llmModel": "LLM Model",
+      "maxTokens": "Max Tokens",
+      "embeddingConfig": "Embedding Configuration",
+      "embeddingBinding": "Embedding Binding",
+      "embeddingBindingHost": "Embedding Binding Host",
+      "embeddingModel": "Embedding Model",
+      "storageConfig": "Storage Configuration",
+      "kvStorage": "KV Storage",
+      "docStatusStorage": "Doc Status Storage",
+      "graphStorage": "Graph Storage",
+      "vectorStorage": "Vector Storage"
+    },
+    "propertiesView": {
+      "node": {
+        "title": "Node",
+        "id": "ID",
+        "labels": "Labels",
+        "degree": "Degree",
+        "properties": "Properties",
+        "relationships": "Relationships"
+      },
+      "edge": {
+        "title": "Relationship",
+        "id": "ID",
+        "type": "Type",
+        "source": "Source",
+        "target": "Target",
+        "properties": "Properties"
+      }
+    },
+    "search": {
+      "placeholder": "Search nodes...",
+      "message": "And {count} others"
+    },
+    "graphLabels": {
+      "selectTooltip": "Select query label",
+      "noLabels": "No labels found",
+      "label": "Label",
+      "placeholder": "Search labels...",
+      "andOthers": "And {count} others"
+    }
+  },
+  "retrievePanel": {
+    "chatMessage": {
+      "copyTooltip": "Copy to clipboard",
+      "copyError": "Failed to copy text to clipboard"
+    },
+    "retrieval": {
+      "startPrompt": "Start a retrieval by typing your query below",
+      "clear": "Clear",
+      "send": "Send",
+      "placeholder": "Type your query...",
+      "error": "Error: Failed to get response"
+    },
+    "querySettings": {
+      "parametersTitle": "Parameters",
+      "parametersDescription": "Configure your query parameters",
+
+      "queryMode": "Query Mode",
+      "queryModeTooltip": "Select the retrieval strategy:\n• Naive: Basic search without advanced techniques\n• Local: Context-dependent information retrieval\n• Global: Utilizes global knowledge base\n• Hybrid: Combines local and global retrieval\n• Mix: Integrates knowledge graph with vector retrieval",
+      "queryModeOptions": {
+        "naive": "Naive",
+        "local": "Local",
+        "global": "Global",
+        "hybrid": "Hybrid",
+        "mix": "Mix"
+      },
+
+      "responseFormat": "Response Format",
+      "responseFormatTooltip": "Defines the response format. Examples:\n• Multiple Paragraphs\n• Single Paragraph\n• Bullet Points",
+      "responseFormatOptions": {
+        "multipleParagraphs": "Multiple Paragraphs",
+        "singleParagraph": "Single Paragraph",
+        "bulletPoints": "Bullet Points"
+      },
+
+      "topK": "Top K Results",
+      "topKTooltip": "Number of top items to retrieve. Represents entities in 'local' mode and relationships in 'global' mode",
+      "topKPlaceholder": "Number of results",
+
+      "maxTokensTextUnit": "Max Tokens for Text Unit",
+      "maxTokensTextUnitTooltip": "Maximum number of tokens allowed for each retrieved text chunk",
+
+      "maxTokensGlobalContext": "Max Tokens for Global Context",
+      "maxTokensGlobalContextTooltip": "Maximum number of tokens allocated for relationship descriptions in global retrieval",
+
+      "maxTokensLocalContext": "Max Tokens for Local Context",
+      "maxTokensLocalContextTooltip": "Maximum number of tokens allocated for entity descriptions in local retrieval",
+
+      "historyTurns": "History Turns",
+      "historyTurnsTooltip": "Number of complete conversation turns (user-assistant pairs) to consider in the response context",
+      "historyTurnsPlaceholder": "Number of history turns",
+
+      "hlKeywords": "High-Level Keywords",
+      "hlKeywordsTooltip": "List of high-level keywords to prioritize in retrieval. Separate with commas",
+      "hlkeywordsPlaceHolder": "Enter keywords",
+
+      "llKeywords": "Low-Level Keywords",
+      "llKeywordsTooltip": "List of low-level keywords to refine retrieval focus. Separate with commas",
+
+      "onlyNeedContext": "Only Need Context",
+      "onlyNeedContextTooltip": "If True, only returns the retrieved context without generating a response",
+
+      "onlyNeedPrompt": "Only Need Prompt",
+      "onlyNeedPromptTooltip": "If True, only returns the generated prompt without producing a response",
+
+      "streamResponse": "Stream Response",
+      "streamResponseTooltip": "If True, enables streaming output for real-time responses"
+    }
+  }
+}

lightrag_webui/src/locales/zh.json

@@ -0,0 +1,235 @@
+{
+  "header": {
+    "documents": "文档",
+    "knowledgeGraph": "知识图谱",
+    "retrieval": "检索",
+    "api": "API",
+    "projectRepository": "项目仓库",
+    "themeToggle": {
+      "switchToLight": "切换到亮色主题",
+      "switchToDark": "切换到暗色主题"
+    }
+  },
+  "documentPanel": {
+    "clearDocuments": {
+      "button": "清除",
+      "tooltip": "清除文档",
+      "title": "清除文档",
+      "confirm": "您确定要清除所有文档吗？",
+      "confirmButton": "确定",
+      "success": "文档已成功清除",
+      "failed": "清除文档失败:\n{{message}}",
+      "error": "清除文档失败:\n{{error}}"
+    },
+    "uploadDocuments": {
+      "button": "上传",
+      "tooltip": "上传文档",
+      "title": "上传文档",
+      "description": "拖放文档到此处或点击浏览。",
+      "uploading": "正在上传 {{name}}: {{percent}}%",
+      "success": "上传成功:\n{{name}} 上传成功",
+      "failed": "上传失败:\n{{name}}\n{{message}}",
+      "error": "上传失败:\n{{name}}\n{{error}}",
+      "generalError": "上传失败\n{{error}}",
+      "fileTypes": "支持的文件类型: TXT, MD, DOCX, PDF, PPTX, RTF, ODT, EPUB, HTML, HTM, TEX, JSON, XML, YAML, YML, CSV, LOG, CONF, INI, PROPERTIES, SQL, BAT, SH, C, CPP, PY, JAVA, JS, TS, SWIFT, GO, RB, PHP, CSS, SCSS, LESS"
+    },
+    "documentManager": {
+      "title": "文档管理",
+      "scanButton": "扫描",
+      "scanTooltip": "扫描文档",
+      "uploadedTitle": "已上传文档",
+      "uploadedDescription": "已上传文档及其状态列表。",
+      "emptyTitle": "暂无文档",
+      "emptyDescription": "尚未上传任何文档。",
+      "columns": {
+        "id": "ID",
+        "summary": "摘要",
+        "status": "状态",
+        "length": "长度",
+        "chunks": "分块",
+        "created": "创建时间",
+        "updated": "更新时间",
+        "metadata": "元数据"
+      },
+      "status": {
+        "completed": "已完成",
+        "processing": "处理中",
+        "pending": "待处理",
+        "failed": "失败"
+      },
+      "errors": {
+        "loadFailed": "加载文档失败\n{{error}}",
+        "scanFailed": "扫描文档失败\n{{error}}",
+        "scanProgressFailed": "获取扫描进度失败\n{{error}}"
+      }
+    }
+  },
+  "graphPanel": {
+    "sideBar": {
+      "settings": {
+        "settings": "设置",
+        "healthCheck": "健康检查",
+        "showPropertyPanel": "显示属性面板",
+        "showSearchBar": "显示搜索栏",
+        "showNodeLabel": "显示节点标签",
+        "nodeDraggable": "节点可拖动",
+        "showEdgeLabel": "显示边标签",
+        "hideUnselectedEdges": "隐藏未选中边",
+        "edgeEvents": "边事件",
+        "maxQueryDepth": "最大查询深度",
+        "minDegree": "最小度数",
+        "maxLayoutIterations": "最大布局迭代次数",
+        "apiKey": "API 密钥",
+        "enterYourAPIkey": "输入您的 API 密钥",
+        "save": "保存"
+      },
+
+      "zoomControl": {
+        "zoomIn": "放大",
+        "zoomOut": "缩小",
+        "resetZoom": "重置缩放"
+      },
+
+      "layoutsControl": {
+        "startAnimation": "开始布局动画",
+        "stopAnimation": "停止布局动画",
+        "layoutGraph": "布局图",
+        "layouts": {
+          "Circular": "环形布局",
+          "Circlepack": "圆形打包布局",
+          "Random": "随机布局",
+          "Noverlaps": "无重叠布局",
+          "Force Directed": "力导向布局",
+          "Force Atlas": "力导向图谱布局"
+        }
+      },
+
+      "fullScreenControl": {
+        "fullScreen": "全屏",
+        "windowed": "窗口模式"
+      }
+    },
+    "statusIndicator": {
+      "connected": "已连接",
+      "disconnected": "未连接"
+    },
+    "statusCard": {
+      "unavailable": "状态信息不可用",
+      "storageInfo": "存储信息",
+      "workingDirectory": "工作目录",
+      "inputDirectory": "输入目录",
+      "llmConfig": "LLM 配置",
+      "llmBinding": "LLM 绑定",
+      "llmBindingHost": "LLM 绑定主机",
+      "llmModel": "LLM 模型",
+      "maxTokens": "最大 Token 数",
+      "embeddingConfig": "嵌入配置",
+      "embeddingBinding": "嵌入绑定",
+      "embeddingBindingHost": "嵌入绑定主机",
+      "embeddingModel": "嵌入模型",
+      "storageConfig": "存储配置",
+      "kvStorage": "KV 存储",
+      "docStatusStorage": "文档状态存储",
+      "graphStorage": "图存储",
+      "vectorStorage": "向量存储"
+    },
+    "propertiesView": {
+      "node": {
+        "title": "节点",
+        "id": "ID",
+        "labels": "标签",
+        "degree": "度数",
+        "properties": "属性",
+        "relationships": "关系"
+      },
+      "edge": {
+        "title": "关系",
+        "id": "ID",
+        "type": "类型",
+        "source": "源",
+        "target": "目标",
+        "properties": "属性"
+      }
+    },
+    "search": {
+      "placeholder": "搜索节点...",
+      "message": "以及其它 {count} 项"
+    },
+    "graphLabels": {
+      "selectTooltip": "选择查询标签",
+      "noLabels": "未找到标签",
+      "label": "标签",
+      "placeholder": "搜索标签...",
+      "andOthers": "以及其它 {count} 个"
+    }
+  },
+  "retrievePanel": {
+    "chatMessage": {
+      "copyTooltip": "复制到剪贴板",
+      "copyError": "无法复制文本到剪贴板"
+    },
+
+    "retrieval": {
+      "startPrompt": "在下面输入您的查询以开始检索",
+      "clear": "清除",
+      "send": "发送",
+      "placeholder": "输入您的查询...",
+      "error": "错误：无法获取响应"
+    },
+    "querySettings": {
+      "parametersTitle": "参数设置",
+      "parametersDescription": "配置查询参数",
+
+      "queryMode": "查询模式",
+      "queryModeTooltip": "选择检索策略：\n• 朴素：不使用高级技术的基本搜索\n• 本地：基于上下文的信息检索\n• 全局：利用全局知识库\n• 混合：结合本地和全局检索\n• 综合：集成知识图谱与向量检索",
+      "queryModeOptions": {
+        "naive": "朴素",
+        "local": "本地",
+        "global": "全局",
+        "hybrid": "混合",
+        "mix": "综合"
+      },
+
+      "responseFormat": "响应格式",
+      "responseFormatTooltip": "定义响应格式。例如：\n• 多个段落\n• 单个段落\n• 项目符号",
+      "responseFormatOptions": {
+        "multipleParagraphs": "多个段落",
+        "singleParagraph": "单个段落",
+        "bulletPoints": "项目符号"
+      },
+
+      "topK": "Top K 结果数",
+      "topKTooltip": "要检索的前 K 个项目数量。在“本地”模式下表示实体，在“全局”模式下表示关系",
+      "topKPlaceholder": "结果数",
+
+      "maxTokensTextUnit": "文本单元最大 Token 数",
+      "maxTokensTextUnitTooltip": "每个检索到的文本块允许的最大 Token 数",
+
+      "maxTokensGlobalContext": "全局上下文最大 Token 数",
+      "maxTokensGlobalContextTooltip": "在全局检索中为关系描述分配的最大 Token 数",
+
+      "maxTokensLocalContext": "本地上下文最大 Token 数",
+      "maxTokensLocalContextTooltip": "在本地检索中为实体描述分配的最大 Token 数",
+
+      "historyTurns": "历史轮次",
+      "historyTurnsTooltip": "在响应上下文中考虑的完整对话轮次（用户-助手对）",
+      "historyTurnsPlaceholder": "历史轮次的数量",
+
+      "hlKeywords": "高级关键词",
+      "hlKeywordsTooltip": "检索时优先考虑的高级关键词。请用逗号分隔",
+      "hlkeywordsPlaceHolder": "输入关键词",
+
+      "llKeywords": "低级关键词",
+      "llKeywordsTooltip": "用于优化检索焦点的低级关键词。请用逗号分隔",
+
+      "onlyNeedContext": "仅需要上下文",
+      "onlyNeedContextTooltip": "如果为 True，则仅返回检索到的上下文，而不会生成回复",
+
+      "onlyNeedPrompt": "仅需要提示",
+      "onlyNeedPromptTooltip": "如果为 True，则仅返回生成的提示，而不会生成回复",
+
+      "streamResponse": "流式响应",
+      "streamResponseTooltip": "如果为 True，则启用流式输出以获得实时响应"
+    }
+  }
+}

lightrag_webui/src/main.tsx

@@ -2,6 +2,8 @@ import { StrictMode } from 'react'
 import { createRoot } from 'react-dom/client'
 import './index.css'
 import App from './App.tsx'
+import "./i18n";
+
 
 createRoot(document.getElementById('root')!).render(
   <StrictMode>