diff --git a/conf/llm_factories.json b/conf/llm_factories.json index 402577a3d..2341f3d10 100644 --- a/conf/llm_factories.json +++ b/conf/llm_factories.json @@ -1895,7 +1895,7 @@ "llm": [ { "llm_name": "step-1-8k", - "tags": "LLM,CHAT,15k", + "tags": "LLM,CHAT,8k", "max_tokens": 8192, "model_type": "chat" }, @@ -1916,6 +1916,12 @@ "tags": "LLM,CHAT,256k", "max_tokens": 262144, "model_type": "chat" + }, + { + "llm_name": "step-1v-8k", + "tags": "LLM,CHAT,IMAGE2TEXT", + "max_tokens": 8000, + "model_type": "image2text" } ] }, diff --git a/rag/llm/chat_model.py b/rag/llm/chat_model.py index 351548719..dc8879bdf 100644 --- a/rag/llm/chat_model.py +++ b/rag/llm/chat_model.py @@ -85,7 +85,6 @@ class MoonshotChat(Base): if not base_url: base_url="https://api.moonshot.cn/v1" super().__init__(key, model_name, base_url) - class XinferenceChat(Base): def __init__(self, key=None, model_name="", base_url=""): key = "xxx" @@ -898,9 +897,9 @@ class OpenRouterChat(Base): self.model_name = model_name class StepFunChat(Base): - def __init__(self, key, model_name, base_url="https://api.stepfun.com/v1/chat/completions"): + def __init__(self, key, model_name, base_url="https://api.stepfun.com/v1"): if not base_url: - base_url = "https://api.stepfun.com/v1/chat/completions" + base_url = "https://api.stepfun.com/v1" super().__init__(key, model_name, base_url)