Merge pull request #1987 from danielaskdd/llm-optimization

feat: Support turning off thinking on OpenRouter/vLLM

README-zh.md

@@ -142,6 +142,8 @@ LightRAG对大型语言模型（LLM）的能力要求远高于传统RAG，因为
 - **LLM选型**：
   - 推荐选用参数量至少为32B的LLM。
   - 上下文长度至少为32KB，推荐达到64KB。
+  - 在文档索引阶段不建议选择推理模型。
+  - 在查询阶段建议选择比索引阶段能力更强的模型，以达到更高的查询效果。
 - **Embedding模型**：
   - 高性能的Embedding模型对RAG至关重要。
   - 推荐使用主流的多语言Embedding模型，例如：BAAI/bge-m3 和 text-embedding-3-large。
@@ -265,7 +267,7 @@ if __name__ == "__main__":
 | **embedding_func_max_async** | `int` | 最大并发异步嵌入进程数 | `16` |
 | **llm_model_func** | `callable` | LLM生成的函数 | `gpt_4o_mini_complete` |
 | **llm_model_name** | `str` | 用于生成的LLM模型名称 | `meta-llama/Llama-3.2-1B-Instruct` |
-| **summary_max_tokens** | `int` | 生成实体关系摘要时送给LLM的最大令牌数 | `32000`（默认值由环境变量MAX_TOKENS更改） |
+| **summary_max_tokens** | `int` | 生成实体关系摘要时送给LLM的最大令牌数 | `32000`（由环境变量 SUMMARY_MAX_TOKENS 设置） |
 | **llm_model_max_async** | `int` | 最大并发异步LLM进程数 | `4`（默认值由环境变量MAX_ASYNC更改） |
 | **llm_model_kwargs** | `dict` | LLM生成的附加参数 | |
 | **vector_db_storage_cls_kwargs** | `dict` | 向量数据库的附加参数，如设置节点和关系检索的阈值 | cosine_better_than_threshold: 0.2（默认值由环境变量COSINE_THRESHOLD更改） |

README.md

@@ -141,6 +141,8 @@ LightRAG's demands on the capabilities of Large Language Models (LLMs) are signi
 - **LLM Selection**:
   - It is recommended to use an LLM with at least 32 billion parameters.
   - The context length should be at least 32KB, with 64KB being recommended.
+  - It is not recommended to choose reasoning models during the document indexing stage.
+  - During the query stage, it is recommended to choose models with stronger capabilities than those used in the indexing stage to achieve better query results.
 - **Embedding Model**:
   - A high-performance Embedding model is essential for RAG.
   - We recommend using mainstream multilingual Embedding models, such as: `BAAI/bge-m3` and `text-embedding-3-large`.
@@ -272,7 +274,7 @@ A full list of LightRAG init parameters:
 | **embedding_func_max_async** | `int` | Maximum number of concurrent asynchronous embedding processes | `16` |
 | **llm_model_func** | `callable` | Function for LLM generation | `gpt_4o_mini_complete` |
 | **llm_model_name** | `str` | LLM model name for generation | `meta-llama/Llama-3.2-1B-Instruct` |
-| **summary_max_tokens** | `int` | Maximum tokens send to LLM to generate entity relation summaries | `32000`（default value changed by env var MAX_TOKENS) |
+| **summary_max_tokens** | `int` | Maximum tokens send to LLM to generate entity relation summaries | `32000`（configured by env var SUMMARY_MAX_TOKENS) |
 | **llm_model_max_async** | `int` | Maximum number of concurrent asynchronous LLM processes | `4`（default value changed by env var MAX_ASYNC) |
 | **llm_model_kwargs** | `dict` | Additional parameters for LLM generation | |
 | **vector_db_storage_cls_kwargs** | `dict` | Additional parameters for vector database, like setting the threshold for nodes and relations retrieval | cosine_better_than_threshold: 0.2（default value changed by env var COSINE_THRESHOLD) |
@@ -1287,8 +1289,10 @@ LightRAG now seamlessly integrates with [RAG-Anything](https://github.com/HKUDS/
                     ),
                 )
             )
+
             # Initialize storage (this will load existing data if available)
             await lightrag_instance.initialize_storages()
+
             # Now initialize RAGAnything with the existing LightRAG instance
             rag = RAGAnything(
                 lightrag=lightrag_instance,  # Pass the existing LightRAG instance
@@ -1317,12 +1321,14 @@ LightRAG now seamlessly integrates with [RAG-Anything](https://github.com/HKUDS/
                 )
                 # Note: working_dir, llm_model_func, embedding_func, etc. are inherited from lightrag_instance
             )
+
             # Query the existing knowledge base
             result = await rag.query_with_multimodal(
                 "What data has been processed in this LightRAG instance?",
                 mode="hybrid"
             )
             print("Query result:", result)
+
             # Add new multimodal documents to the existing LightRAG instance
             await rag.process_document_complete(
                 file_path="path/to/new/multimodal_document.pdf",

env.example

@@ -8,6 +8,8 @@ PORT=9621
 WEBUI_TITLE='My Graph KB'
 WEBUI_DESCRIPTION="Simple and Fast Graph Based RAG System"
 # WORKERS=2
+### gunicorn worker timeout(as default LLM request timeout if LLM_TIMEOUT is not set)
+# TIMEOUT=150
 # CORS_ORIGINS=http://localhost:3000,http://localhost:8080
 
 ### Optional SSL Configuration
@@ -105,7 +107,7 @@ ENABLE_LLM_CACHE_FOR_EXTRACT=true
 ### Entity and relation summarization configuration
 ### Number of duplicated entities/edges to trigger LLM re-summary on merge (at least 3 is recommented)， and max tokens send to LLM
 # FORCE_LLM_SUMMARY_ON_MERGE=4
-# MAX_TOKENS=10000
+# SUMMARY_MAX_TOKENS=10000
 ### Maximum number of entity extraction attempts for ambiguous content
 # MAX_GLEANING=1
 
@@ -125,8 +127,8 @@ MAX_PARALLEL_INSERT=2
 ### LLM Configuration
 ### LLM_BINDING type: openai, ollama, lollms, azure_openai, aws_bedrock
 ###########################################################
-### LLM temperature setting for all llm binding (openai, azure_openai, ollama)
+### LLM request timeout setting for all llm (set to TIMEOUT if not specified, 0 means no timeout for Ollma)
-# TEMPERATURE=1.0
+# LLM_TIMEOUT=150
 ### Some models like o1-mini require temperature to be set to 1, some LLM can fall into output loops with low temperature
 
 LLM_BINDING=openai
@@ -145,23 +147,33 @@ LLM_BINDING_API_KEY=your_api_key
 # LLM_BINDING=openai
 
 ### OpenAI Specific Parameters
-### Apply frequency penalty to prevent the LLM from generating repetitive or looping outputs
+# OPENAI_LLM_TEMPERATURE=1.0
-# OPENAI_LLM_FREQUENCY_PENALTY=1.1
+# OPENAI_LLM_REASONING_EFFORT=low
-### use the following command to see all support options for openai and azure_openai
+### For models like Qwen3 with fewer than 32B param, it is recommended to set the presence penalty to 1.5
+# OPENAI_LLM_PRESENCE_PENALTY=1.5
+### If the presence penalty still can not stop the model from generates repetitive or unconstrained output
+# OPENAI_LLM_MAX_COMPLETION_TOKENS=16384
+
+### OpenRouter Specific Parameters
+# OPENAI_LLM_EXTRA_BODY='{"reasoning": {"enabled": false}}'
+### Qwen3 Specific Parameters depoly by vLLM
+# OPENAI_LLM_EXTRA_BODY='{"chat_template_kwargs": {"enable_thinking": false}}'
+
+### use the following command to see all support options for OpenAI, azure_openai or OpenRouter
 ### lightrag-server --llm-binding openai --help
 
 ### Ollama Server Specific Parameters
-### Time out in seconds, None for infinite timeout
+### OLLAMA_LLM_NUM_CTX must be provided, and should at least larger than MAX_TOTAL_TOKENS + 2000
-TIMEOUT=240
-### OLLAMA_LLM_NUM_CTX must be larger than MAX_TOTAL_TOKENS + 2000
 OLLAMA_LLM_NUM_CTX=32768
+# OLLAMA_LLM_TEMPERATURE=1.0
 ### Stop sequences for Ollama LLM
 # OLLAMA_LLM_STOP='["</s>", "Assistant:", "\n\n"]'
-### If OLLAMA_LLM_TEMPERATURE is not specified, the system will default to the value defined by TEMPERATURE
-# OLLAMA_LLM_TEMPERATURE=0.85
 ### use the following command to see all support options for Ollama LLM
 ### lightrag-server --llm-binding ollama --help
 
+### Bedrock Specific Parameters
+# BEDROCK_LLM_TEMPERATURE=1.0
+
 ####################################################################################
 ### Embedding Configuration (Should not be changed after the first file processed)
 ####################################################################################

lightrag/api/README-zh.md

@@ -357,7 +357,7 @@ API 服务器可以通过三种方式配置（优先级从高到低）：
 LightRAG 支持绑定到各种 LLM/嵌入后端：
 
 * ollama
-* openai 和 openai 兼容
+* openai (含openai 兼容)
 * azure_openai
 * lollms
 * aws_bedrock
@@ -372,7 +372,10 @@ lightrag-server --llm-binding ollama --help
 lightrag-server --embedding-binding ollama --help
 ```
 
+> 请使用openai兼容方式访问OpenRouter或vLLM部署的LLM。可以通过 `OPENAI_LLM_EXTRA_BODY` 环境变量给OpenRouter或vLLM传递额外的参数，实现推理模式的关闭或者其它个性化控制。
+
 ### 实体提取配置
+
 * ENABLE_LLM_CACHE_FOR_EXTRACT：为实体提取启用 LLM 缓存（默认：true）
 
 在测试环境中将 `ENABLE_LLM_CACHE_FOR_EXTRACT` 设置为 true 以减少 LLM 调用成本是很常见的做法。
@@ -478,7 +481,7 @@ SUMMARY_LANGUAGE=Chinese
 MAX_PARALLEL_INSERT=2
 
 ### LLM Configuration (Use valid host. For local services installed with docker, you can use host.docker.internal)
-TIMEOUT=200
+TIMEOUT=150
 MAX_ASYNC=4
 
 LLM_BINDING=openai

lightrag/api/README.md

@@ -360,7 +360,7 @@ Most of the configurations come with default settings; check out the details in
 LightRAG supports binding to various LLM/Embedding backends:
 
 * ollama
-* openai & openai compatible
+* openai (including openai compatible)
 * azure_openai
 * lollms
 * aws_bedrock
@@ -374,6 +374,8 @@ lightrag-server --llm-binding ollama --help
 lightrag-server --embedding-binding ollama --help
 ```
 
+> Please use OpenAI-compatible method to access LLMs deployed by OpenRouter or vLLM. You can pass additional parameters to OpenRouter or vLLM through the `OPENAI_LLM_EXTRA_BODY` environment variable to disable reasoning mode or achieve other personalized controls.
+
 ### Entity Extraction Configuration
 * ENABLE_LLM_CACHE_FOR_EXTRACT: Enable LLM cache for entity extraction (default: true)
 
@@ -485,7 +487,7 @@ SUMMARY_LANGUAGE=Chinese
 MAX_PARALLEL_INSERT=2
 
 ### LLM Configuration (Use valid host. For local services installed with docker, you can use host.docker.internal)
-TIMEOUT=200
+TIMEOUT=150
 MAX_ASYNC=4
 
 LLM_BINDING=openai

lightrag/api/config.py

@@ -35,7 +35,6 @@ from lightrag.constants import (
     DEFAULT_EMBEDDING_BATCH_NUM,
     DEFAULT_OLLAMA_MODEL_NAME,
     DEFAULT_OLLAMA_MODEL_TAG,
-    DEFAULT_TEMPERATURE,
 )
 
 # use the .env that is inside the current folder
@@ -123,7 +122,7 @@ def parse_args() -> argparse.Namespace:
     parser.add_argument(
         "--max-tokens",
         type=int,
-        default=get_env_value("MAX_TOKENS", DEFAULT_SUMMARY_MAX_TOKENS, int),
+        default=get_env_value("SUMMARY_MAX_TOKENS", DEFAULT_SUMMARY_MAX_TOKENS, int),
         help=f"Maximum token size (default: from env or {DEFAULT_SUMMARY_MAX_TOKENS})",
     )
 
@@ -264,14 +263,6 @@ def parse_args() -> argparse.Namespace:
     elif os.environ.get("LLM_BINDING") in ["openai", "azure_openai"]:
         OpenAILLMOptions.add_args(parser)
 
-    # Add global temperature command line argument
-    parser.add_argument(
-        "--temperature",
-        type=float,
-        default=get_env_value("TEMPERATURE", DEFAULT_TEMPERATURE, float),
-        help="Global temperature setting for LLM (default: from env TEMPERATURE or 0.1)",
-    )
-
     args = parser.parse_args()
 
     # convert relative path to absolute path
@@ -330,32 +321,6 @@ def parse_args() -> argparse.Namespace:
     )
     args.enable_llm_cache = get_env_value("ENABLE_LLM_CACHE", True, bool)
 
-    # Handle Ollama LLM temperature with priority cascade when llm-binding is ollama
-    if args.llm_binding == "ollama":
-        # Priority order (highest to lowest):
-        # 1. --ollama-llm-temperature command argument
-        # 2. OLLAMA_LLM_TEMPERATURE environment variable
-        # 3. --temperature command argument
-        # 4. TEMPERATURE environment variable
-
-        # Check if --ollama-llm-temperature was explicitly provided in command line
-        if "--ollama-llm-temperature" not in sys.argv:
-            # Use args.temperature which handles --temperature command arg and TEMPERATURE env var priority
-            args.ollama_llm_temperature = args.temperature
-
-    # Handle OpenAI LLM temperature with priority cascade when llm-binding is openai or azure_openai
-    if args.llm_binding in ["openai", "azure_openai"]:
-        # Priority order (highest to lowest):
-        # 1. --openai-llm-temperature command argument
-        # 2. OPENAI_LLM_TEMPERATURE environment variable
-        # 3. --temperature command argument
-        # 4. TEMPERATURE environment variable
-
-        # Check if --openai-llm-temperature was explicitly provided in command line
-        if "--openai-llm-temperature" not in sys.argv:
-            # Use args.temperature which handles --temperature command arg and TEMPERATURE env var priority
-            args.openai_llm_temperature = args.temperature
-
     # Select Document loading tool (DOCLING, DEFAULT)
     args.document_loading_engine = get_env_value("DOCUMENT_LOADING_ENGINE", "DEFAULT")
 

lightrag/api/lightrag_server.py

@@ -254,6 +254,8 @@ def create_app(args):
     if args.embedding_binding == "jina":
         from lightrag.llm.jina import jina_embed
 
+    llm_timeout = get_env_value("LLM_TIMEOUT", args.timeout, int)
+
     async def openai_alike_model_complete(
         prompt,
         system_prompt=None,
@@ -267,12 +269,10 @@ def create_app(args):
         if history_messages is None:
             history_messages = []
 
-        # Use OpenAI LLM options if available, otherwise fallback to global temperature
+        # Use OpenAI LLM options if available
-        if args.llm_binding == "openai":
         openai_options = OpenAILLMOptions.options_dict(args)
+        kwargs["timeout"] = llm_timeout
         kwargs.update(openai_options)
-        else:
-            kwargs["temperature"] = args.temperature
 
         return await openai_complete_if_cache(
             args.llm_model,
@@ -297,12 +297,10 @@ def create_app(args):
         if history_messages is None:
             history_messages = []
 
-        # Use OpenAI LLM options if available, otherwise fallback to global temperature
+        # Use OpenAI LLM options
-        if args.llm_binding == "azure_openai":
         openai_options = OpenAILLMOptions.options_dict(args)
+        kwargs["timeout"] = llm_timeout
         kwargs.update(openai_options)
-        else:
-            kwargs["temperature"] = args.temperature
 
         return await azure_openai_complete_if_cache(
             args.llm_model,
@@ -329,7 +327,7 @@ def create_app(args):
             history_messages = []
 
         # Use global temperature for Bedrock
-        kwargs["temperature"] = args.temperature
+        kwargs["temperature"] = get_env_value("BEDROCK_LLM_TEMPERATURE", 1.0, float)
 
         return await bedrock_complete_if_cache(
             args.llm_model,
@@ -451,7 +449,7 @@ def create_app(args):
             llm_model_kwargs=(
                 {
                     "host": args.llm_binding_host,
-                    "timeout": args.timeout,
+                    "timeout": llm_timeout,
                     "options": OllamaLLMOptions.options_dict(args),
                     "api_key": args.llm_binding_api_key,
                 }
@@ -481,9 +479,6 @@ def create_app(args):
             llm_model_func=azure_openai_model_complete,
             chunk_token_size=int(args.chunk_size),
             chunk_overlap_token_size=int(args.chunk_overlap_size),
-            llm_model_kwargs={
-                "timeout": args.timeout,
-            },
             llm_model_name=args.llm_model,
             llm_model_max_async=args.max_async,
             summary_max_tokens=args.max_tokens,

lightrag/api/run_with_gunicorn.py

@@ -153,7 +153,7 @@ def main():
 
             # Timeout configuration prioritizes command line arguments
             gunicorn_config.timeout = (
-                global_args.timeout * 2
+                global_args.timeout + 30
                 if global_args.timeout is not None
                 else get_env_value(
                     "TIMEOUT", DEFAULT_TIMEOUT + 30, int, special_none=True

lightrag/api/utils_api.py

@@ -201,6 +201,8 @@ def display_splash_screen(args: argparse.Namespace) -> None:
     ASCIIColors.yellow(f"{args.port}")
     ASCIIColors.white("    ├─ Workers: ", end="")
     ASCIIColors.yellow(f"{args.workers}")
+    ASCIIColors.white("    ├─ Timeout: ", end="")
+    ASCIIColors.yellow(f"{args.timeout}")
     ASCIIColors.white("    ├─ CORS Origins: ", end="")
     ASCIIColors.yellow(f"{args.cors_origins}")
     ASCIIColors.white("    ├─ SSL Enabled: ", end="")
@@ -238,14 +240,10 @@ def display_splash_screen(args: argparse.Namespace) -> None:
     ASCIIColors.yellow(f"{args.llm_binding_host}")
     ASCIIColors.white("    ├─ Model: ", end="")
     ASCIIColors.yellow(f"{args.llm_model}")
-    ASCIIColors.white("    ├─ Temperature: ", end="")
-    ASCIIColors.yellow(f"{args.temperature}")
     ASCIIColors.white("    ├─ Max Async for LLM: ", end="")
     ASCIIColors.yellow(f"{args.max_async}")
     ASCIIColors.white("    ├─ Max Tokens: ", end="")
     ASCIIColors.yellow(f"{args.max_tokens}")
-    ASCIIColors.white("    ├─ Timeout: ", end="")
-    ASCIIColors.yellow(f"{args.timeout if args.timeout else 'None (infinite)'}")
     ASCIIColors.white("    ├─ LLM Cache Enabled: ", end="")
     ASCIIColors.yellow(f"{args.enable_llm_cache}")
     ASCIIColors.white("    └─ LLM Cache for Extraction Enabled: ", end="")

lightrag/constants.py

@@ -49,7 +49,7 @@ DEFAULT_MAX_PARALLEL_INSERT = 2  # Default maximum parallel insert operations
 DEFAULT_EMBEDDING_FUNC_MAX_ASYNC = 8  # Default max async for embedding functions
 DEFAULT_EMBEDDING_BATCH_NUM = 10  # Default batch size for embedding computations
 
-# Ollama Server Timetout in seconds
+# gunicorn worker timeout(as default LLM request timeout if LLM_TIMEOUT is not set)
 DEFAULT_TIMEOUT = 150
 
 # Logging configuration defaults

lightrag/lightrag.py

@@ -283,7 +283,7 @@ class LightRAG:
     """Name of the LLM model used for generating responses."""
 
     summary_max_tokens: int = field(
-        default=int(os.getenv("MAX_TOKENS", DEFAULT_SUMMARY_MAX_TOKENS))
+        default=int(os.getenv("SUMMARY_MAX_TOKENS", DEFAULT_SUMMARY_MAX_TOKENS))
     )
     """Maximum number of tokens allowed per LLM response."""
 

lightrag/llm/Readme.md

@@ -36,7 +36,6 @@ async def llm_model_func(prompt, system_prompt=None, history_messages=[], **kwar
             llm_instance = OpenAI(
                 model="gpt-4",
                 api_key="your-openai-key",
-                temperature=0.7,
             )
             kwargs['llm_instance'] = llm_instance
 
@@ -91,7 +90,6 @@ async def llm_model_func(prompt, system_prompt=None, history_messages=[], **kwar
                 model=f"openai/{settings.LLM_MODEL}",  # Format: "provider/model_name"
                 api_base=settings.LITELLM_URL,
                 api_key=settings.LITELLM_KEY,
-                temperature=0.7,
             )
             kwargs['llm_instance'] = llm_instance
 

lightrag/llm/anthropic.py

@@ -77,14 +77,23 @@ async def anthropic_complete_if_cache(
     if not VERBOSE_DEBUG and logger.level == logging.DEBUG:
         logging.getLogger("anthropic").setLevel(logging.INFO)
 
+    kwargs.pop("hashing_kv", None)
+    kwargs.pop("keyword_extraction", None)
+    timeout = kwargs.pop("timeout", None)
+
     anthropic_async_client = (
-        AsyncAnthropic(default_headers=default_headers, api_key=api_key)
+        AsyncAnthropic(
+            default_headers=default_headers, api_key=api_key, timeout=timeout
+        )
         if base_url is None
         else AsyncAnthropic(
-            base_url=base_url, default_headers=default_headers, api_key=api_key
+            base_url=base_url,
+            default_headers=default_headers,
+            api_key=api_key,
+            timeout=timeout,
         )
     )
-    kwargs.pop("hashing_kv", None)
+
     messages: list[dict[str, Any]] = []
     if system_prompt:
         messages.append({"role": "system", "content": system_prompt})

lightrag/llm/azure_openai.py

@@ -59,13 +59,17 @@ async def azure_openai_complete_if_cache(
         or os.getenv("OPENAI_API_VERSION")
     )
 
+    kwargs.pop("hashing_kv", None)
+    kwargs.pop("keyword_extraction", None)
+    timeout = kwargs.pop("timeout", None)
+
     openai_async_client = AsyncAzureOpenAI(
         azure_endpoint=base_url,
         azure_deployment=deployment,
         api_key=api_key,
         api_version=api_version,
+        timeout=timeout,
     )
-    kwargs.pop("hashing_kv", None)
     messages = []
     if system_prompt:
         messages.append({"role": "system", "content": system_prompt})

lightrag/llm/binding_options.py

@@ -99,7 +99,7 @@ class BindingOptions:
         group = parser.add_argument_group(f"{cls._binding_name} binding options")
         for arg_item in cls.args_env_name_type_value():
             # Handle JSON parsing for list types
-            if arg_item["type"] == List[str]:
+            if arg_item["type"] is List[str]:
 
                 def json_list_parser(value):
                     try:
@@ -126,6 +126,34 @@ class BindingOptions:
                     default=env_value,
                     help=arg_item["help"],
                 )
+            # Handle JSON parsing for dict types
+            elif arg_item["type"] is dict:
+
+                def json_dict_parser(value):
+                    try:
+                        parsed = json.loads(value)
+                        if not isinstance(parsed, dict):
+                            raise argparse.ArgumentTypeError(
+                                f"Expected JSON object, got {type(parsed).__name__}"
+                            )
+                        return parsed
+                    except json.JSONDecodeError as e:
+                        raise argparse.ArgumentTypeError(f"Invalid JSON: {e}")
+
+                # Get environment variable with JSON parsing
+                env_value = get_env_value(f"{arg_item['env_name']}", argparse.SUPPRESS)
+                if env_value is not argparse.SUPPRESS:
+                    try:
+                        env_value = json_dict_parser(env_value)
+                    except argparse.ArgumentTypeError:
+                        env_value = argparse.SUPPRESS
+
+                group.add_argument(
+                    f"--{arg_item['argname']}",
+                    type=json_dict_parser,
+                    default=env_value,
+                    help=arg_item["help"],
+                )
             else:
                 group.add_argument(
                     f"--{arg_item['argname']}",
@@ -234,8 +262,8 @@ class BindingOptions:
                 if arg_item["help"]:
                     sample_stream.write(f"# {arg_item['help']}\n")
 
-                # Handle JSON formatting for list types
+                # Handle JSON formatting for list and dict types
-                if arg_item["type"] == List[str]:
+                if arg_item["type"] is List[str] or arg_item["type"] is dict:
                     default_value = json.dumps(arg_item["default"])
                 else:
                     default_value = arg_item["default"]
@@ -431,6 +459,8 @@ class OpenAILLMOptions(BindingOptions):
     stop: List[str] = field(default_factory=list)  # Stop sequences
     temperature: float = DEFAULT_TEMPERATURE  # Controls randomness (0.0 to 2.0)
     top_p: float = 1.0  # Nucleus sampling parameter (0.0 to 1.0)
+    max_tokens: int = None  # Maximum number of tokens to generate(deprecated, use max_completion_tokens instead)
+    extra_body: dict = None  # Extra body parameters for OpenRouter of vLLM
 
     # Help descriptions
     _help: ClassVar[dict[str, str]] = {
@@ -443,6 +473,8 @@ class OpenAILLMOptions(BindingOptions):
         "stop": 'Stop sequences (JSON array of strings, e.g., \'["</s>", "\\n\\n"]\')',
         "temperature": "Controls randomness (0.0-2.0, higher = more creative)",
         "top_p": "Nucleus sampling parameter (0.0-1.0, lower = more focused)",
+        "max_tokens": "Maximum number of tokens to generate (deprecated, use max_completion_tokens instead)",
+        "extra_body": 'Extra body parameters for OpenRouter of vLLM (JSON dict, e.g., \'"reasoning": {"reasoning": {"enabled": false}}\')',
     }
 
 
@@ -493,6 +525,8 @@ if __name__ == "__main__":
                 "1000",
                 "--openai-llm-stop",
                 '["</s>", "\\n\\n"]',
+                "--openai-llm-reasoning",
+                '{"effort": "high", "max_tokens": 2000, "exclude": false, "enabled": true}',
             ]
         )
         print("Final args for LLM and Embedding:")
@@ -518,5 +552,100 @@ if __name__ == "__main__":
         print("\nOpenAI LLM options instance:")
         print(openai_options.asdict())
 
+        # Test creating OpenAI options instance with reasoning parameter
+        openai_options_with_reasoning = OpenAILLMOptions(
+            temperature=0.9,
+            max_completion_tokens=2000,
+            reasoning={
+                "effort": "medium",
+                "max_tokens": 1500,
+                "exclude": True,
+                "enabled": True,
+            },
+        )
+        print("\nOpenAI LLM options instance with reasoning:")
+        print(openai_options_with_reasoning.asdict())
+
+        # Test dict parsing functionality
+        print("\n" + "=" * 50)
+        print("TESTING DICT PARSING FUNCTIONALITY")
+        print("=" * 50)
+
+        # Test valid JSON dict parsing
+        test_parser = ArgumentParser(description="Test dict parsing")
+        OpenAILLMOptions.add_args(test_parser)
+
+        try:
+            test_args = test_parser.parse_args(
+                ["--openai-llm-reasoning", '{"effort": "low", "max_tokens": 1000}']
+            )
+            print("✓ Valid JSON dict parsing successful:")
+            print(
+                f"  Parsed reasoning: {OpenAILLMOptions.options_dict(test_args)['reasoning']}"
+            )
+        except Exception as e:
+            print(f"✗ Valid JSON dict parsing failed: {e}")
+
+        # Test invalid JSON dict parsing
+        try:
+            test_args = test_parser.parse_args(
+                [
+                    "--openai-llm-reasoning",
+                    '{"effort": "low", "max_tokens": 1000',  # Missing closing brace
+                ]
+            )
+            print("✗ Invalid JSON should have failed but didn't")
+        except SystemExit:
+            print("✓ Invalid JSON dict parsing correctly rejected")
+        except Exception as e:
+            print(f"✓ Invalid JSON dict parsing correctly rejected: {e}")
+
+        # Test non-dict JSON parsing
+        try:
+            test_args = test_parser.parse_args(
+                [
+                    "--openai-llm-reasoning",
+                    '["not", "a", "dict"]',  # Array instead of dict
+                ]
+            )
+            print("✗ Non-dict JSON should have failed but didn't")
+        except SystemExit:
+            print("✓ Non-dict JSON parsing correctly rejected")
+        except Exception as e:
+            print(f"✓ Non-dict JSON parsing correctly rejected: {e}")
+
+        print("\n" + "=" * 50)
+        print("TESTING ENVIRONMENT VARIABLE SUPPORT")
+        print("=" * 50)
+
+        # Test environment variable support for dict
+        import os
+
+        os.environ["OPENAI_LLM_REASONING"] = (
+            '{"effort": "high", "max_tokens": 3000, "exclude": false}'
+        )
+
+        env_parser = ArgumentParser(description="Test env var dict parsing")
+        OpenAILLMOptions.add_args(env_parser)
+
+        try:
+            env_args = env_parser.parse_args(
+                []
+            )  # No command line args, should use env var
+            reasoning_from_env = OpenAILLMOptions.options_dict(env_args).get(
+                "reasoning"
+            )
+            if reasoning_from_env:
+                print("✓ Environment variable dict parsing successful:")
+                print(f"  Parsed reasoning from env: {reasoning_from_env}")
+            else:
+                print("✗ Environment variable dict parsing failed: No reasoning found")
+        except Exception as e:
+            print(f"✗ Environment variable dict parsing failed: {e}")
+        finally:
+            # Clean up environment variable
+            if "OPENAI_LLM_REASONING" in os.environ:
+                del os.environ["OPENAI_LLM_REASONING"]
+
     else:
         print(BindingOptions.generate_dot_env_sample())

lightrag/llm/lollms.py

@@ -59,7 +59,7 @@ async def lollms_model_if_cache(
         "personality": kwargs.get("personality", -1),
         "n_predict": kwargs.get("n_predict", None),
         "stream": stream,
-        "temperature": kwargs.get("temperature", 0.8),
+        "temperature": kwargs.get("temperature", 1.0),
         "top_k": kwargs.get("top_k", 50),
         "top_p": kwargs.get("top_p", 0.95),
         "repeat_penalty": kwargs.get("repeat_penalty", 0.8),

lightrag/llm/ollama.py

@@ -51,6 +51,8 @@ async def _ollama_model_if_cache(
     # kwargs.pop("response_format", None) # allow json
     host = kwargs.pop("host", None)
     timeout = kwargs.pop("timeout", None)
+    if timeout == 0:
+        timeout = None
     kwargs.pop("hashing_kv", None)
     api_key = kwargs.pop("api_key", None)
     headers = {

lightrag/llm/openai.py

@@ -149,18 +149,20 @@ async def openai_complete_if_cache(
     if not VERBOSE_DEBUG and logger.level == logging.DEBUG:
         logging.getLogger("openai").setLevel(logging.INFO)
 
+    # Remove special kwargs that shouldn't be passed to OpenAI
+    kwargs.pop("hashing_kv", None)
+    kwargs.pop("keyword_extraction", None)
+
     # Extract client configuration options
     client_configs = kwargs.pop("openai_client_configs", {})
 
     # Create the OpenAI client
     openai_async_client = create_openai_async_client(
-        api_key=api_key, base_url=base_url, client_configs=client_configs
+        api_key=api_key,
+        base_url=base_url,
+        client_configs=client_configs,
     )
 
-    # Remove special kwargs that shouldn't be passed to OpenAI
-    kwargs.pop("hashing_kv", None)
-    kwargs.pop("keyword_extraction", None)
-
     # Prepare messages
     messages: list[dict[str, Any]] = []
     if system_prompt: