This is an automated email from the ASF dual-hosted git repository.
jin pushed a commit to branch main
in repository https://gitbox.apache.org/repos/asf/incubator-hugegraph-ai.git
The following commit(s) were added to refs/heads/main by this push:
new 212d9a5 chore(llm): use 4.1-mini and 0.01 temperature by default
(#214)
212d9a5 is described below
commit 212d9a58f4cb0fd4439e821858fcd2d2c4980f5a
Author: SoJGooo <[email protected]>
AuthorDate: Wed May 7 15:44:29 2025 +0800
chore(llm): use 4.1-mini and 0.01 temperature by default (#214)
Co-authored-by: imbajin <[email protected]>
---
.asf.yaml | 1 -
hugegraph-llm/src/hugegraph_llm/config/llm_config.py | 13 +++++++------
hugegraph-llm/src/hugegraph_llm/models/llms/litellm.py | 14 +++++++-------
hugegraph-llm/src/hugegraph_llm/models/llms/openai.py | 6 +++---
hugegraph-llm/src/hugegraph_llm/models/llms/qianfan.py | 2 +-
5 files changed, 18 insertions(+), 18 deletions(-)
diff --git a/.asf.yaml b/.asf.yaml
index 9fa40a0..cafdba4 100644
--- a/.asf.yaml
+++ b/.asf.yaml
@@ -54,7 +54,6 @@ github:
collaborators:
- ChenZiHong-Gavin
- MrJs133
- - vichayturen
- HJ-Young
- afterimagex
- returnToInnocence
diff --git a/hugegraph-llm/src/hugegraph_llm/config/llm_config.py
b/hugegraph-llm/src/hugegraph_llm/config/llm_config.py
index 5e314bd..a9b4b2d 100644
--- a/hugegraph-llm/src/hugegraph_llm/config/llm_config.py
+++ b/hugegraph-llm/src/hugegraph_llm/config/llm_config.py
@@ -33,13 +33,13 @@ class LLMConfig(BaseConfig):
# 1. OpenAI settings
openai_chat_api_base: Optional[str] = os.environ.get("OPENAI_BASE_URL",
"https://api.openai.com/v1";)
openai_chat_api_key: Optional[str] = os.environ.get("OPENAI_API_KEY")
- openai_chat_language_model: Optional[str] = "gpt-4o-mini"
+ openai_chat_language_model: Optional[str] = "gpt-4.1-mini"
openai_extract_api_base: Optional[str] = os.environ.get("OPENAI_BASE_URL",
"https://api.openai.com/v1";)
openai_extract_api_key: Optional[str] = os.environ.get("OPENAI_API_KEY")
- openai_extract_language_model: Optional[str] = "gpt-4o-mini"
+ openai_extract_language_model: Optional[str] = "gpt-4.1-mini"
openai_text2gql_api_base: Optional[str] =
os.environ.get("OPENAI_BASE_URL", "https://api.openai.com/v1";)
openai_text2gql_api_key: Optional[str] = os.environ.get("OPENAI_API_KEY")
- openai_text2gql_language_model: Optional[str] = "gpt-4o-mini"
+ openai_text2gql_language_model: Optional[str] = "gpt-4.1-mini"
openai_embedding_api_base: Optional[str] =
os.environ.get("OPENAI_EMBEDDING_BASE_URL", "https://api.openai.com/v1";)
openai_embedding_api_key: Optional[str] =
os.environ.get("OPENAI_EMBEDDING_API_KEY")
openai_embedding_model: Optional[str] = "text-embedding-3-small"
@@ -64,6 +64,7 @@ class LLMConfig(BaseConfig):
ollama_embedding_port: Optional[int] = 11434
ollama_embedding_model: Optional[str] = None
# 4. QianFan/WenXin settings
+ # TODO: update to one token key mode
qianfan_chat_api_key: Optional[str] = None
qianfan_chat_secret_key: Optional[str] = None
qianfan_chat_access_token: Optional[str] = None
@@ -87,15 +88,15 @@ class LLMConfig(BaseConfig):
# 5. LiteLLM settings
litellm_chat_api_key: Optional[str] = None
litellm_chat_api_base: Optional[str] = None
- litellm_chat_language_model: Optional[str] = "openai/gpt-4o"
+ litellm_chat_language_model: Optional[str] = "openai/gpt-4.1-mini"
litellm_chat_tokens: int = 8192
litellm_extract_api_key: Optional[str] = None
litellm_extract_api_base: Optional[str] = None
- litellm_extract_language_model: Optional[str] = "openai/gpt-4o"
+ litellm_extract_language_model: Optional[str] = "openai/gpt-4.1-mini"
litellm_extract_tokens: int = 256
litellm_text2gql_api_key: Optional[str] = None
litellm_text2gql_api_base: Optional[str] = None
- litellm_text2gql_language_model: Optional[str] = "openai/gpt-4o"
+ litellm_text2gql_language_model: Optional[str] = "openai/gpt-4.1-mini"
litellm_text2gql_tokens: int = 4096
litellm_embedding_api_key: Optional[str] = None
litellm_embedding_api_base: Optional[str] = None
diff --git a/hugegraph-llm/src/hugegraph_llm/models/llms/litellm.py
b/hugegraph-llm/src/hugegraph_llm/models/llms/litellm.py
index ca5ae60..7627979 100644
--- a/hugegraph-llm/src/hugegraph_llm/models/llms/litellm.py
+++ b/hugegraph-llm/src/hugegraph_llm/models/llms/litellm.py
@@ -38,9 +38,9 @@ class LiteLLMClient(BaseLLM):
self,
api_key: Optional[str] = None,
api_base: Optional[str] = None,
- model_name: str = "openai/gpt-4o", # Can be any model supported by
LiteLLM
- max_tokens: int = 4096,
- temperature: float = 0.0,
+ model_name: str = "openai/gpt-4.1-mini", # Can be any model supported
by LiteLLM
+ max_tokens: int = 8192,
+ temperature: float = 0.01,
) -> None:
self.api_key = api_key
self.api_base = api_base
@@ -49,8 +49,8 @@ class LiteLLMClient(BaseLLM):
self.temperature = temperature
@retry(
- stop=stop_after_attempt(3),
- wait=wait_exponential(multiplier=1, min=4, max=10),
+ stop=stop_after_attempt(2),
+ wait=wait_exponential(multiplier=1, min=2, max=5),
retry=retry_if_exception_type((RateLimitError, BudgetExceededError,
APIError))
)
def generate(
@@ -78,8 +78,8 @@ class LiteLLMClient(BaseLLM):
return f"Error: {str(e)}"
@retry(
- stop=stop_after_attempt(3),
- wait=wait_exponential(multiplier=1, min=4, max=10),
+ stop=stop_after_attempt(2),
+ wait=wait_exponential(multiplier=1, min=2, max=5),
retry=retry_if_exception_type((RateLimitError, BudgetExceededError,
APIError))
)
async def agenerate(
diff --git a/hugegraph-llm/src/hugegraph_llm/models/llms/openai.py
b/hugegraph-llm/src/hugegraph_llm/models/llms/openai.py
index 45f6d7a..b185314 100644
--- a/hugegraph-llm/src/hugegraph_llm/models/llms/openai.py
+++ b/hugegraph-llm/src/hugegraph_llm/models/llms/openai.py
@@ -38,9 +38,9 @@ class OpenAIClient(BaseLLM):
self,
api_key: Optional[str] = None,
api_base: Optional[str] = None,
- model_name: str = "gpt-4o-mini",
- max_tokens: int = 4096,
- temperature: float = 0.0,
+ model_name: str = "gpt-4.1-mini",
+ max_tokens: int = 8092,
+ temperature: float = 0.01,
) -> None:
api_key = api_key or ''
self.client = OpenAI(api_key=api_key, base_url=api_base)
diff --git a/hugegraph-llm/src/hugegraph_llm/models/llms/qianfan.py
b/hugegraph-llm/src/hugegraph_llm/models/llms/qianfan.py
index cbca691..2d306ac 100644
--- a/hugegraph-llm/src/hugegraph_llm/models/llms/qianfan.py
+++ b/hugegraph-llm/src/hugegraph_llm/models/llms/qianfan.py
@@ -27,7 +27,7 @@ from hugegraph_llm.utils.log import log
class QianfanClient(BaseLLM):
- def __init__(self, model_name: Optional[str] = "ERNIE-4.0-Turbo-8K",
+ def __init__(self, model_name: Optional[str] = "ernie-4.5-8k-preview",
api_key: Optional[str] = None, secret_key: Optional[str] =
None):
qianfan.get_config().AK = api_key or llm_settings.qianfan_chat_api_key
qianfan.get_config().SK = secret_key or
llm_settings.qianfan_chat_secret_key
