Set the default LLM temperature to 1.0 and centralize constant management

env.example

@@ -111,19 +111,21 @@ MAX_PARALLEL_INSERT=2
 #######################
 ### LLM Configuration
 #######################
-### Time out in seconds for LLM, None for infinite timeout
+### Some models like o1-mini require temperature to be set to 1, some LLM can fall into output loops with low temperature
-TIMEOUT=240
+# TEMPERATURE=1.0
-### Some models like o1-mini require temperature to be set to 1
+
-TEMPERATURE=0
 ### LLM Binding type: openai, ollama, lollms, azure_openai
 LLM_BINDING=openai
 LLM_MODEL=gpt-4o
 LLM_BINDING_HOST=https://api.openai.com/v1
 LLM_BINDING_API_KEY=your_api_key
 
-### Set as num_ctx option for Ollama LLM (Must be larger than MAX_TOTAL_TOKENS+2000)
+### Most Commont Parameters for Ollama Server
 ### see also env.ollama-binding-options.example for fine tuning ollama
+### OLLAMA_LLM_NUM_CTX must be larger than MAX_TOTAL_TOKENS + 2000
 # OLLAMA_LLM_NUM_CTX=32768
+### Time out in seconds, None for infinite timeout
+TIMEOUT=240
 
 ### Optional for Azure
 # AZURE_OPENAI_API_VERSION=2024-08-01-preview

lightrag/api/README-zh.md

@@ -468,7 +468,6 @@ MAX_PARALLEL_INSERT=2
 
 ### LLM Configuration (Use valid host. For local services installed with docker, you can use host.docker.internal)
 TIMEOUT=200
-TEMPERATURE=0.0
 MAX_ASYNC=4
 
 LLM_BINDING=openai

lightrag/api/README.md

@@ -476,7 +476,6 @@ MAX_PARALLEL_INSERT=2
 
 ### LLM Configuration (Use valid host. For local services installed with docker, you can use host.docker.internal)
 TIMEOUT=200
-TEMPERATURE=0.0
 MAX_ASYNC=4
 
 LLM_BINDING=openai

lightrag/api/config.py

@@ -33,6 +33,7 @@ from lightrag.constants import (
     DEFAULT_OLLAMA_MODEL_SIZE,
     DEFAULT_OLLAMA_CREATED_AT,
     DEFAULT_OLLAMA_DIGEST,
+    DEFAULT_TEMPERATURE,
 )
 
 # use the .env that is inside the current folder
@@ -332,7 +333,7 @@ def parse_args() -> argparse.Namespace:
     args.enable_llm_cache = get_env_value("ENABLE_LLM_CACHE", True, bool)
 
     # Inject LLM temperature configuration
-    args.temperature = get_env_value("TEMPERATURE", 0.5, float)
+    args.temperature = get_env_value("TEMPERATURE", DEFAULT_TEMPERATURE, float)
 
     # Select Document loading tool (DOCLING, DEFAULT)
     args.document_loading_engine = get_env_value("DOCUMENT_LOADING_ENGINE", "DEFAULT")

lightrag/constants.py

@@ -8,7 +8,7 @@ consistency and makes maintenance easier.
 
 # Default values for server settings
 DEFAULT_WOKERS = 2
-DEFAULT_TIMEOUT = 150
+DEFAULT_MAX_GRAPH_NODES = 1000
 
 # Default values for extraction settings
 DEFAULT_SUMMARY_LANGUAGE = "English"  # Default language for summaries
@@ -34,16 +34,23 @@ DEFAULT_HISTORY_TURNS = 0
 DEFAULT_ENABLE_RERANK = True
 DEFAULT_MIN_RERANK_SCORE = 0.0
 
-# File path configuration for vector and graph database
+# File path configuration for vector and graph database(Should not be changed, used in Milvus Schema)
 DEFAULT_MAX_FILE_PATH_LENGTH = 4090
 
+# Default temperature for LLM
+DEFAULT_TEMPERATURE = 1.0
+
 # Async configuration defaults
 DEFAULT_MAX_ASYNC = 4  # Default maximum async operations
+DEFAULT_MAX_PARALLEL_INSERT = 2  # Default maximum parallel insert operations
 
 # Embedding configuration defaults
 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
+DEFAULT_TIMEOUT = 150
+
 # Logging configuration defaults
 DEFAULT_LOG_MAX_BYTES = 10485760  # Default 10MB
 DEFAULT_LOG_BACKUP_COUNT = 5  # Default 5 backups

lightrag/lightrag.py

@@ -33,6 +33,9 @@ from lightrag.constants import (
     DEFAULT_RELATED_CHUNK_NUMBER,
     DEFAULT_MIN_RERANK_SCORE,
     DEFAULT_SUMMARY_MAX_TOKENS,
+    DEFAULT_MAX_ASYNC,
+    DEFAULT_MAX_PARALLEL_INSERT,
+    DEFAULT_MAX_GRAPH_NODES,
 )
 from lightrag.utils import get_env_value
 
@@ -283,7 +286,9 @@ class LightRAG:
     )
     """Maximum number of tokens allowed per LLM response."""
 
-    llm_model_max_async: int = field(default=int(os.getenv("MAX_ASYNC", 4)))
+    llm_model_max_async: int = field(
+        default=int(os.getenv("MAX_ASYNC", DEFAULT_MAX_ASYNC))
+    )
     """Maximum number of concurrent LLM calls."""
 
     llm_model_kwargs: dict[str, Any] = field(default_factory=dict)
@@ -315,10 +320,14 @@ class LightRAG:
     # Extensions
     # ---
 
-    max_parallel_insert: int = field(default=int(os.getenv("MAX_PARALLEL_INSERT", 2)))
+    max_parallel_insert: int = field(
+        default=int(os.getenv("MAX_PARALLEL_INSERT", DEFAULT_MAX_PARALLEL_INSERT))
+    )
     """Maximum number of parallel insert operations."""
 
-    max_graph_nodes: int = field(default=get_env_value("MAX_GRAPH_NODES", 1000, int))
+    max_graph_nodes: int = field(
+        default=get_env_value("MAX_GRAPH_NODES", DEFAULT_MAX_GRAPH_NODES, int)
+    )
     """Maximum number of graph nodes to return in knowledge graph queries."""
 
     addon_params: dict[str, Any] = field(

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", 0.1),
+        "temperature": kwargs.get("temperature", 0.8),
         "top_k": kwargs.get("top_k", 50),
         "top_p": kwargs.get("top_p", 0.95),
         "repeat_penalty": kwargs.get("repeat_penalty", 0.8),