diff --git a/README-zh.md b/README-zh.md
index 45335489..e9599099 100644
--- a/README-zh.md
+++ b/README-zh.md
@@ -30,7 +30,7 @@
- 
+
diff --git a/lightrag/operate.py b/lightrag/operate.py
index 88837435..4e219cf8 100644
--- a/lightrag/operate.py
+++ b/lightrag/operate.py
@@ -26,6 +26,7 @@ from .utils import (
get_conversation_turns,
use_llm_func_with_cache,
update_chunk_cache_list,
+ remove_think_tags,
)
from .base import (
BaseGraphStorage,
@@ -1703,7 +1704,8 @@ async def extract_keywords_only(
result = await use_model_func(kw_prompt, keyword_extraction=True)
# 6. Parse out JSON from the LLM response
- match = re.search(r"\{.*\}", result, re.DOTALL)
+ result = remove_think_tags(result)
+ match = re.search(r"\{.*?\}", result, re.DOTALL)
if not match:
logger.error("No JSON-like structure found in the LLM respond.")
return [], []
diff --git a/lightrag/utils.py b/lightrag/utils.py
index c6e2def9..386de3ab 100644
--- a/lightrag/utils.py
+++ b/lightrag/utils.py
@@ -1465,6 +1465,11 @@ async def update_chunk_cache_list(
)
+def remove_think_tags(text: str) -> str:
+ """Remove tags from the text"""
+ return re.sub(r"^(.*?|)", "", text, flags=re.DOTALL).strip()
+
+
async def use_llm_func_with_cache(
input_text: str,
use_llm_func: callable,
@@ -1531,6 +1536,7 @@ async def use_llm_func_with_cache(
kwargs["max_tokens"] = max_tokens
res: str = await use_llm_func(input_text, **kwargs)
+ res = remove_think_tags(res)
if llm_response_cache.global_config.get("enable_llm_cache_for_entity_extract"):
await save_to_cache(
@@ -1557,8 +1563,9 @@ async def use_llm_func_with_cache(
if max_tokens is not None:
kwargs["max_tokens"] = max_tokens
- logger.info(f"Call LLM function with query text lenght: {len(input_text)}")
- return await use_llm_func(input_text, **kwargs)
+ logger.info(f"Call LLM function with query text length: {len(input_text)}")
+ res = await use_llm_func(input_text, **kwargs)
+ return remove_think_tags(res)
def get_content_summary(content: str, max_length: int = 250) -> str: