refactor: improve JSON parsing reliability with json-repair library

Replace regex-based JSON extraction with json-repair for better handling of malformed LLM responses. Remove deprecated JSON parsing utilities and clean up keyword_extraction parameter across LLM providers.

- Remove locate_json_string_body_from_string() and convert_response_to_json()
- Use json-repair.loads() in extract_keywords_only() for robust parsing
- Clean up LLM interfaces and remove unused parameters
- Add json-repair dependency

README-zh.md

@@ -272,7 +272,6 @@ if __name__ == "__main__":
 | **enable_llm_cache** | `bool` | 如果为`TRUE`，将LLM结果存储在缓存中；重复的提示返回缓存的响应 | `TRUE` |
 | **enable_llm_cache_for_entity_extract** | `bool` | 如果为`TRUE`，将实体提取的LLM结果存储在缓存中；适合初学者调试应用程序 | `TRUE` |
 | **addon_params** | `dict` | 附加参数，例如`{"example_number": 1, "language": "Simplified Chinese", "entity_types": ["organization", "person", "geo", "event"]}`：设置示例限制、输出语言和文档处理的批量大小 | `example_number: 所有示例, language: English` |
-| **convert_response_to_json_func** | `callable` | 未使用 | `convert_response_to_json` |
 | **embedding_cache_config** | `dict` | 问答缓存的配置。包含三个参数：`enabled`：布尔值，启用/禁用缓存查找功能。启用时，系统将在生成新答案之前检查缓存的响应。`similarity_threshold`：浮点值（0-1），相似度阈值。当新问题与缓存问题的相似度超过此阈值时，将直接返回缓存的答案而不调用LLM。`use_llm_check`：布尔值，启用/禁用LLM相似度验证。启用时，在返回缓存答案之前，将使用LLM作为二次检查来验证问题之间的相似度。 | 默认：`{"enabled": False, "similarity_threshold": 0.95, "use_llm_check": False}` |
 
 </details>

README.md

@@ -279,7 +279,6 @@ A full list of LightRAG init parameters:
 | **enable_llm_cache** | `bool` | If `TRUE`, stores LLM results in cache; repeated prompts return cached responses | `TRUE` |
 | **enable_llm_cache_for_entity_extract** | `bool` | If `TRUE`, stores LLM results in cache for entity extraction; Good for beginners to debug your application | `TRUE` |
 | **addon_params** | `dict` | Additional parameters, e.g., `{"example_number": 1, "language": "Simplified Chinese", "entity_types": ["organization", "person", "geo", "event"]}`: sets example limit, entiy/relation extraction output language | `example_number: all examples, language: English` |
-| **convert_response_to_json_func** | `callable` | Not used | `convert_response_to_json` |
 | **embedding_cache_config** | `dict` | Configuration for question-answer caching. Contains three parameters: `enabled`: Boolean value to enable/disable cache lookup functionality. When enabled, the system will check cached responses before generating new answers. `similarity_threshold`: Float value (0-1), similarity threshold. When a new question's similarity with a cached question exceeds this threshold, the cached answer will be returned directly without calling the LLM. `use_llm_check`: Boolean value to enable/disable LLM similarity verification. When enabled, LLM will be used as a secondary check to verify the similarity between questions before returning cached answers. | Default: `{"enabled": False, "similarity_threshold": 0.95, "use_llm_check": False}` |
 
 </details>

lightrag/lightrag.py

@@ -81,7 +81,6 @@ from .utils import (
     EmbeddingFunc,
     always_get_an_event_loop,
     compute_mdhash_id,
-    convert_response_to_json,
     lazy_external_import,
     priority_limit_async_func_call,
     get_content_summary,
@@ -341,15 +340,6 @@ class LightRAG:
     # Storages Management
     # ---
 
-    convert_response_to_json_func: Callable[[str], dict[str, Any]] = field(
-        default_factory=lambda: convert_response_to_json
-    )
-    """
-    Custom function for converting LLM responses to JSON format.
-
-    The default function is :func:`.utils.convert_response_to_json`.
-    """
-
     cosine_better_than_threshold: float = field(
         default=float(os.getenv("COSINE_THRESHOLD", 0.2))
     )

lightrag/llm/azure_openai.py

@@ -23,7 +23,6 @@ from tenacity import (
 
 from lightrag.utils import (
     wrap_embedding_func_with_attrs,
-    locate_json_string_body_from_string,
     safe_unicode_decode,
 )
 
@@ -108,7 +107,7 @@ async def azure_openai_complete_if_cache(
 async def azure_openai_complete(
     prompt, system_prompt=None, history_messages=[], keyword_extraction=False, **kwargs
 ) -> str:
-    keyword_extraction = kwargs.pop("keyword_extraction", None)
+    kwargs.pop("keyword_extraction", None)
     result = await azure_openai_complete_if_cache(
         os.getenv("LLM_MODEL", "gpt-4o-mini"),
         prompt,
@@ -116,8 +115,6 @@ async def azure_openai_complete(
         history_messages=history_messages,
         **kwargs,
     )
-    if keyword_extraction:  # TODO: use JSON API
-        return locate_json_string_body_from_string(result)
     return result
 
 

lightrag/llm/bedrock.py

@@ -15,10 +15,6 @@ from tenacity import (
     retry_if_exception_type,
 )
 
-from lightrag.utils import (
-    locate_json_string_body_from_string,
-)
-
 
 class BedrockError(Exception):
     """Generic error for issues related to Amazon Bedrock"""
@@ -96,7 +92,7 @@ async def bedrock_complete_if_cache(
 async def bedrock_complete(
     prompt, system_prompt=None, history_messages=[], keyword_extraction=False, **kwargs
 ) -> str:
-    keyword_extraction = kwargs.pop("keyword_extraction", None)
+    kwargs.pop("keyword_extraction", None)
     model_name = kwargs["hashing_kv"].global_config["llm_model_name"]
     result = await bedrock_complete_if_cache(
         model_name,
@@ -105,8 +101,6 @@ async def bedrock_complete(
         history_messages=history_messages,
         **kwargs,
     )
-    if keyword_extraction:  # TODO: use JSON API
-        return locate_json_string_body_from_string(result)
     return result
 
 

lightrag/llm/hf.py

@@ -24,9 +24,6 @@ from lightrag.exceptions import (
     RateLimitError,
     APITimeoutError,
 )
-from lightrag.utils import (
-    locate_json_string_body_from_string,
-)
 import torch
 import numpy as np
 
@@ -119,7 +116,7 @@ async def hf_model_if_cache(
 async def hf_model_complete(
     prompt, system_prompt=None, history_messages=[], keyword_extraction=False, **kwargs
 ) -> str:
-    keyword_extraction = kwargs.pop("keyword_extraction", None)
+    kwargs.pop("keyword_extraction", None)
     model_name = kwargs["hashing_kv"].global_config["llm_model_name"]
     result = await hf_model_if_cache(
         model_name,
@@ -128,8 +125,6 @@ async def hf_model_complete(
         history_messages=history_messages,
         **kwargs,
     )
-    if keyword_extraction:  # TODO: use JSON API
-        return locate_json_string_body_from_string(result)
     return result
 
 

lightrag/llm/llama_index_impl.py

@@ -21,7 +21,6 @@ from tenacity import (
 )
 from lightrag.utils import (
     wrap_embedding_func_with_attrs,
-    locate_json_string_body_from_string,
 )
 from lightrag.exceptions import (
     APIConnectionError,
@@ -157,7 +156,7 @@ async def llama_index_complete(
     if history_messages is None:
         history_messages = []
 
-    keyword_extraction = kwargs.pop("keyword_extraction", None)
+    kwargs.pop("keyword_extraction", None)
     result = await llama_index_complete_if_cache(
         kwargs.get("llm_instance"),
         prompt,
@@ -165,8 +164,6 @@ async def llama_index_complete(
         history_messages=history_messages,
         **kwargs,
     )
-    if keyword_extraction:
-        return locate_json_string_body_from_string(result)
     return result
 
 

lightrag/llm/openai.py

@@ -27,7 +27,6 @@ from tenacity import (
 )
 from lightrag.utils import (
     wrap_embedding_func_with_attrs,
-    locate_json_string_body_from_string,
     safe_unicode_decode,
     logger,
 )
@@ -418,7 +417,7 @@ async def nvidia_openai_complete(
 ) -> str:
     if history_messages is None:
         history_messages = []
-    keyword_extraction = kwargs.pop("keyword_extraction", None)
+    kwargs.pop("keyword_extraction", None)
     result = await openai_complete_if_cache(
         "nvidia/llama-3.1-nemotron-70b-instruct",  # context length 128k
         prompt,
@@ -427,8 +426,6 @@ async def nvidia_openai_complete(
         base_url="https://integrate.api.nvidia.com/v1",
         **kwargs,
     )
-    if keyword_extraction:  # TODO: use JSON API
-        return locate_json_string_body_from_string(result)
     return result
 
 

lightrag/operate.py

@@ -5,6 +5,7 @@ import asyncio
 import json
 import re
 import os
+import json_repair
 from typing import Any, AsyncIterator
 from collections import Counter, defaultdict
 
@@ -1781,10 +1782,10 @@ async def extract_keywords_only(
     )
     if cached_response is not None:
         try:
-            keywords_data = json.loads(cached_response)
+            keywords_data = json_repair.loads(cached_response)
-            return keywords_data["high_level_keywords"], keywords_data[
+            return keywords_data.get("high_level_keywords", []), keywords_data.get(
-                "low_level_keywords"
+                "low_level_keywords", []
-            ]
+            )
         except (json.JSONDecodeError, KeyError):
             logger.warning(
                 "Invalid cache format for keywords, proceeding with extraction"
@@ -1832,12 +1833,11 @@ async def extract_keywords_only(
 
     # 6. Parse out JSON from the LLM response
     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 [], []
     try:
-        keywords_data = json.loads(match.group(0))
+        keywords_data = json_repair.loads(result)
+        if not keywords_data:
+            logger.error("No JSON-like structure found in the LLM respond.")
+            return [], []
     except json.JSONDecodeError as e:
         logger.error(f"JSON parsing error: {e}")
         logger.error(f"LLM respond: {result}")

lightrag/utils.py

@@ -248,43 +248,6 @@ class EmbeddingFunc:
         return await self.func(*args, **kwargs)
 
 
-def locate_json_string_body_from_string(content: str) -> str | None:
-    """Locate the JSON string body from a string"""
-    try:
-        maybe_json_str = re.search(r"{.*}", content, re.DOTALL)
-        if maybe_json_str is not None:
-            maybe_json_str = maybe_json_str.group(0)
-            maybe_json_str = maybe_json_str.replace("\\n", "")
-            maybe_json_str = maybe_json_str.replace("\n", "")
-            maybe_json_str = maybe_json_str.replace("'", '"')
-            # json.loads(maybe_json_str) # don't check here, cannot validate schema after all
-            return maybe_json_str
-    except Exception:
-        pass
-        # try:
-        #     content = (
-        #         content.replace(kw_prompt[:-1], "")
-        #         .replace("user", "")
-        #         .replace("model", "")
-        #         .strip()
-        #     )
-        #     maybe_json_str = "{" + content.split("{")[1].split("}")[0] + "}"
-        #     json.loads(maybe_json_str)
-
-        return None
-
-
-def convert_response_to_json(response: str) -> dict[str, Any]:
-    json_str = locate_json_string_body_from_string(response)
-    assert json_str is not None, f"Unable to parse JSON from response: {response}"
-    try:
-        data = json.loads(json_str)
-        return data
-    except json.JSONDecodeError as e:
-        logger.error(f"Failed to parse JSON: {json_str}")
-        raise e from None
-
-
 def compute_args_hash(*args: Any) -> str:
     """Compute a hash for the given arguments.
     Args:

pyproject.toml

@@ -66,6 +66,7 @@ api = [
     "httpcore",
     "httpx",
     "jiter",
+    "json-repair",
     "passlib[bcrypt]",
     "psutil",
     "PyJWT",