Upload folder using huggingface_hub

src/agentscope/credential/_openai_compat.py

@@ -23,6 +23,11 @@ class GroqCredential(OpenAICredential):
     )
     """Custom base URL for the Groq API."""
 
+    @classmethod
+    def get_chat_model_class(cls):
+        from ..model import GroqChatModel
+        return GroqChatModel
+
 
 class OpenRouterCredential(OpenAICredential):
     """The OpenRouter credential model."""
@@ -40,6 +45,11 @@ class OpenRouterCredential(OpenAICredential):
     )
     """Custom base URL for the OpenRouter API."""
 
+    @classmethod
+    def get_chat_model_class(cls):
+        from ..model import OpenRouterChatModel
+        return OpenRouterChatModel
+
 
 class NvidiaNimCredential(OpenAICredential):
     """The NVIDIA NIM credential model."""
@@ -56,3 +66,8 @@ class NvidiaNimCredential(OpenAICredential):
         description="The base URL for the NVIDIA NIM API.",
     )
     """Custom base URL for the NVIDIA NIM API."""
+
+    @classmethod
+    def get_chat_model_class(cls):
+        from ..model import NvidiaNimChatModel
+        return NvidiaNimChatModel

src/agentscope/model/__init__.py

@@ -14,6 +14,9 @@ from ._openai_chat import OpenAIChatModel
 from ._xai import XAIChatModel
 from ._moonshot import MoonshotChatModel
 from ._openai_response import OpenAIResponseModel
+from ._groq import GroqChatModel
+from ._openrouter import OpenRouterChatModel
+from ._nvidia_nim import NvidiaNimChatModel
 
 __all__ = [
     "ChatUsage",
@@ -30,4 +33,7 @@ __all__ = [
     "XAIChatModel",
     "MoonshotChatModel",
     "OpenAIResponseModel",
+    "GroqChatModel",
+    "OpenRouterChatModel",
+    "NvidiaNimChatModel",
 ]

src/agentscope/model/_groq/__init__.py

@@ -0,0 +1,3 @@
+from ._model import GroqChatModel
+
+__all__ = ["GroqChatModel"]

src/agentscope/model/_groq/_model.py

@@ -0,0 +1,26 @@
+# -*- coding: utf-8 -*-
+"""The Groq chat model implementation."""
+from typing import Literal
+
+from .._openai_chat import OpenAIChatModel
+from ...credential._openai_compat import GroqCredential
+
+
+class GroqChatModel(OpenAIChatModel):
+    """The Groq chat model."""
+
+    type: Literal["groq_chat"] = "groq_chat"
+    """The type of the chat model."""
+
+    def __init__(
+        self,
+        credential: GroqCredential,
+        model: str,
+        **kwargs,
+    ) -> None:
+        """Initialize the Groq chat model."""
+        super().__init__(
+            credential=credential,
+            model=model,
+            **kwargs,
+        )

src/agentscope/model/_groq/_models/llama3-70b.yaml

@@ -0,0 +1,7 @@
+name: llama3-70b-8192
+api_type: groq_chat
+description: The Meta Llama 3 70B model via Groq.
+max_context: 8192
+support_vision: false
+support_function_calling: true
+support_structured_output: true

src/agentscope/model/_nvidia_nim/__init__.py

@@ -0,0 +1,3 @@
+from ._model import NvidiaNimChatModel
+
+__all__ = ["NvidiaNimChatModel"]

src/agentscope/model/_nvidia_nim/_model.py

@@ -0,0 +1,26 @@
+# -*- coding: utf-8 -*-
+"""The NVIDIA NIM chat model implementation."""
+from typing import Literal
+
+from .._openai_chat import OpenAIChatModel
+from ...credential._openai_compat import NvidiaNimCredential
+
+
+class NvidiaNimChatModel(OpenAIChatModel):
+    """The NVIDIA NIM chat model."""
+
+    type: Literal["nvidia_nim_chat"] = "nvidia_nim_chat"
+    """The type of the chat model."""
+
+    def __init__(
+        self,
+        credential: NvidiaNimCredential,
+        model: str,
+        **kwargs,
+    ) -> None:
+        """Initialize the NVIDIA NIM chat model."""
+        super().__init__(
+            credential=credential,
+            model=model,
+            **kwargs,
+        )

src/agentscope/model/_nvidia_nim/_models/deepseek-ai-deepseek-v4-flash.yaml

@@ -0,0 +1,7 @@
+name: deepseek-ai/deepseek-v4-flash
+api_type: nvidia_nim_chat
+description: deepseek-v4-flash model via NVIDIA NIM.
+max_context: 1000000
+support_vision: false
+support_function_calling: true
+support_structured_output: true

src/agentscope/model/_nvidia_nim/_models/deepseek-ai-deepseek-v4-pro.yaml

@@ -0,0 +1,7 @@
+name: deepseek-ai/deepseek-v4-pro
+api_type: nvidia_nim_chat
+description: deepseek-v4-pro model via NVIDIA NIM.
+max_context: 1000000
+support_vision: false
+support_function_calling: true
+support_structured_output: true

src/agentscope/model/_nvidia_nim/_models/meta-llama3-70b-instruct.yaml

@@ -0,0 +1,7 @@
+name: meta/llama3-70b-instruct
+api_type: nvidia_nim_chat
+description: The Meta Llama 3 70B Instruct model via NVIDIA NIM.
+max_context: 8192
+support_vision: false
+support_function_calling: true
+support_structured_output: true

src/agentscope/model/_nvidia_nim/_models/minimaxai-minimax-m2.7.yaml

@@ -0,0 +1,7 @@
+name: minimaxai/minimax-m2.7
+api_type: nvidia_nim_chat
+description: minimax-m2.7 model via NVIDIA NIM.
+max_context: 204800
+support_vision: false
+support_function_calling: true
+support_structured_output: true

src/agentscope/model/_nvidia_nim/_models/minimaxai-minimax-m3.yaml

@@ -0,0 +1,7 @@
+name: minimaxai/minimax-m3
+api_type: nvidia_nim_chat
+description: minimax-m3 model via NVIDIA NIM.
+max_context: 1000000
+support_vision: false
+support_function_calling: true
+support_structured_output: true

src/agentscope/model/_nvidia_nim/_models/nvidia-nemotron-3-ultra-550b-a55b.yaml

@@ -0,0 +1,7 @@
+name: nvidia/nemotron-3-ultra-550b-a55b
+api_type: nvidia_nim_chat
+description: nemotron-3-ultra-550b-a55b model via NVIDIA NIM.
+max_context: 1000000
+support_vision: false
+support_function_calling: true
+support_structured_output: true

src/agentscope/model/_nvidia_nim/_models/qwen-qwen3.5-397b-a17b.yaml

@@ -0,0 +1,7 @@
+name: qwen/qwen3.5-397b-a17b
+api_type: nvidia_nim_chat
+description: qwen3.5-397b-a17b model via NVIDIA NIM.
+max_context: 262144
+support_vision: false
+support_function_calling: true
+support_structured_output: true

src/agentscope/model/_nvidia_nim/_models/stepfun-ai-step-3.7-flash.yaml

@@ -0,0 +1,7 @@
+name: stepfun-ai/step-3.7-flash
+api_type: nvidia_nim_chat
+description: step-3.7-flash model via NVIDIA NIM.
+max_context: 256000
+support_vision: false
+support_function_calling: true
+support_structured_output: true