whosellm 0.2.1__tar.gz → 0.2.2__tar.gz

{whosellm-0.2.1 → whosellm-0.2.2}/.bumpversion.toml

@@ -2,7 +2,7 @@
 # 文档 / Documentation: https://callowayproject.github.io/bump-my-version/
 
 [tool.bumpversion]
-current_version = "0.2.1"
+current_version = "0.2.2"
 parse = """(?x)
     (?P<major>0|[1-9]\\d*)\\.
     (?P<minor>0|[1-9]\\d*)\\.

{whosellm-0.2.1 → whosellm-0.2.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: whosellm
-Version: 0.2.1
+Version: 0.2.2
 Summary: A unified LLM model version and capability management library
 Author-email: JQQ <jqq1716@gmail.com>
 License: MIT

{whosellm-0.2.1 → whosellm-0.2.2}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "whosellm"
-version = "0.2.1"
+version = "0.2.2"
 description = "A unified LLM model version and capability management library"
 authors = [
     {name = "JQQ", email = "jqq1716@gmail.com"}

{whosellm-0.2.1 → whosellm-0.2.2}/tests/models/families/test_deepseek.py

@@ -11,7 +11,7 @@ from whosellm.provider import Provider
 
 
 def test_deepseek_default_capabilities() -> None:
-    """验证 DeepSeek 官方家族默认能力 / Validate DeepSeek official family default capabilities"""
+    """验证 DeepSeek 官方家族默认能力（V4 基线） / Validate DeepSeek official family default capabilities (V4 baseline)"""
     from whosellm import LLMeta
 
     # 使用官方 Provider 前缀确保获取官方配置
@@ -20,8 +20,10 @@ def test_deepseek_default_capabilities() -> None:
 
     assert capabilities.supports_streaming is True
     assert capabilities.supports_function_calling is True
-    assert capabilities.max_tokens == 8000
-    assert capabilities.context_window == 128000
+    # V4 系列：1M 上下文，384K 最大输出
+    assert capabilities.max_tokens == 384_000
+    assert capabilities.context_window == 1_000_000
+    # deepseek-chat 是 v4-flash 非思考模式的别名
     assert capabilities.supports_thinking is False
 
 
@@ -31,13 +33,13 @@ def test_deepseek_chat_specific_model() -> None:
 
     assert config is not None
     version, variant, capabilities = config
-    assert version == "1.0"
+    assert version == "4.0"
     assert variant == "chat"
     assert capabilities is not None
     assert capabilities.supports_function_calling is True
     assert capabilities.supports_streaming is True
-    assert capabilities.max_tokens == 8000
-    assert capabilities.context_window == 128000
+    assert capabilities.max_tokens == 384_000
+    assert capabilities.context_window == 1_000_000
 
 
 def test_deepseek_chat_pattern_matching() -> None:
@@ -50,7 +52,7 @@ def test_deepseek_chat_pattern_matching() -> None:
 
         assert model.family == ModelFamily.DEEPSEEK
         assert model.variant == "chat"
-        assert model.version == "1.0"
+        assert model.version == "4.0"
         assert model.provider == Provider.DEEPSEEK
 
 
@@ -60,10 +62,12 @@ def test_deepseek_reasoner_specific_model() -> None:
 
     assert config is not None
     version, variant, capabilities = config
-    assert version == "1.0"
+    assert version == "4.0"
     assert variant == "reasoner"
     assert capabilities is not None
     assert capabilities.supports_thinking is True
+    assert capabilities.max_tokens == 384_000
+    assert capabilities.context_window == 1_000_000
 
 
 def test_deepseek_base_pattern_without_variant() -> None:
@@ -71,17 +75,16 @@ def test_deepseek_base_pattern_without_variant() -> None:
     from whosellm import LLMeta
 
     # 使用 Provider 前缀确保匹配官方配置
-    # DeepSeek 官方只支持 chat 和 reasoner，默认使用 chat
     model = LLMeta("deepseek::deepseek-chat")
 
     assert model.family == ModelFamily.DEEPSEEK
     assert model.variant == "chat"
-    assert model.version == "1.0"
+    assert model.version == "4.0"
     assert model.capabilities.supports_function_calling is True
 
 
 def test_deepseek_reasoner_does_not_use_chat_capabilities() -> None:
-    """验证 reasoner 不会意外继承 chat 的函数调用能力 / Ensure reasoner capabilities override family defaults"""
+    """验证 reasoner 不会意外继承 chat 的能力 / Ensure reasoner capabilities override family defaults"""
     from whosellm import LLMeta
 
     model = LLMeta("deepseek::deepseek-reasoner")
@@ -95,7 +98,7 @@ def test_deepseek_no_structured_outputs() -> None:
     """验证 DeepSeek 官方模型不支持 structured_outputs（仅支持 json_object）"""
     from whosellm import LLMeta
 
-    for model_id in ["deepseek-chat", "deepseek-reasoner"]:
+    for model_id in ["deepseek-chat", "deepseek-reasoner", "deepseek-v4-flash", "deepseek-v4-pro"]:
         model = LLMeta(f"deepseek::{model_id}")
         assert model.capabilities.supports_structured_outputs is False, (
             f"{model_id}: DeepSeek API 仅支持 response_format={{type:'json_object'}}，"
@@ -106,15 +109,68 @@ def test_deepseek_no_structured_outputs() -> None:
         )
 
 
-def test_deepseek_official_invalid_model_names() -> None:
-    """验证 DeepSeek 官方不支持的模型名称会被识别为 UNKNOWN / Validate unsupported official model names are recognized as UNKNOWN"""
+def test_deepseek_v4_flash_specific_model() -> None:
+    """验证 deepseek-v4-flash 特定模型配置 / Validate deepseek-v4-flash specific configuration"""
     from whosellm import LLMeta
 
-    # DeepSeek 官方不提供 deepseek-v3.2-exp 这样的命名方式
-    # 官方只支持 deepseek-chat-{suffix} 和 deepseek-reasoner-{suffix}
-    model = LLMeta("deepseek::deepseek-v3.2-exp")
+    model = LLMeta("deepseek::deepseek-v4-flash")
 
-    # 当使用 Provider 前缀时，Provider 会被保留，但 family 会是 UNKNOWN
-    assert model.family == ModelFamily.UNKNOWN
+    assert model.family == ModelFamily.DEEPSEEK
+    assert model.provider == Provider.DEEPSEEK
+    assert model.version == "4.0"
+    assert model.variant == "flash"
+
+    caps = model.capabilities
+    assert caps.supports_thinking is True
+    assert caps.supports_function_calling is True
+    assert caps.supports_streaming is True
+    assert caps.supports_json_outputs is True
+    assert caps.supports_structured_outputs is False
+    assert caps.max_tokens == 384_000
+    assert caps.context_window == 1_000_000
+
+
+def test_deepseek_v4_pro_specific_model() -> None:
+    """验证 deepseek-v4-pro 特定模型配置 / Validate deepseek-v4-pro specific configuration"""
+    from whosellm import LLMeta
+
+    model = LLMeta("deepseek::deepseek-v4-pro")
+
+    assert model.family == ModelFamily.DEEPSEEK
+    assert model.provider == Provider.DEEPSEEK
+    assert model.version == "4.0"
+    assert model.variant == "pro"
+
+    caps = model.capabilities
+    assert caps.supports_thinking is True
+    assert caps.supports_function_calling is True
+    assert caps.max_tokens == 384_000
+    assert caps.context_window == 1_000_000
+
+
+def test_deepseek_v4_variant_ordering() -> None:
+    """验证 v4-flash < v4-pro 的排序关系 / Validate v4-flash < v4-pro ordering"""
+    from whosellm import LLMeta
+
+    flash = LLMeta("deepseek::deepseek-v4-flash")
+    pro = LLMeta("deepseek::deepseek-v4-pro")
+
+    assert flash < pro
+
+
+def test_deepseek_versioned_pattern_matches_official_family() -> None:
+    """验证带版本号的 DS 模型名能被官方 family 识别 / Validate versioned DS names match the official family"""
+    from whosellm import LLMeta
+
+    # V4 是 DS 官方首次开放版本号命名的系列
+    model = LLMeta("deepseek::deepseek-v4-flash")
+    assert model.family == ModelFamily.DEEPSEEK
     assert model.provider == Provider.DEEPSEEK
-    assert model.variant == ""  # 无法匹配到任何 variant
+    assert model.version == "4.0"
+
+    # 带小数版本号也能匹配 family（即使当前没有 specific_model 条目）
+    model_v32 = LLMeta("deepseek::deepseek-v3.2-exp")
+    assert model_v32.family == ModelFamily.DEEPSEEK
+    assert model_v32.provider == Provider.DEEPSEEK
+    assert model_v32.version == "3.2"
+    assert model_v32.variant == "exp"

{whosellm-0.2.1 → whosellm-0.2.2}/tests/models/families/test_deepseek_tencent.py

@@ -66,6 +66,8 @@ def test_tencent_deepseek_r1_0528() -> None:
     assert capabilities is not None
     assert capabilities.supports_thinking is True
     assert capabilities.context_window == 128000
+    # 腾讯云 LKE 官方文档明确列为不支持 Function Calling
+    assert capabilities.supports_function_calling is False
 
 
 def test_tencent_deepseek_v3_1() -> None:
@@ -109,6 +111,22 @@ def test_tencent_deepseek_v3_2_exp() -> None:
     assert capabilities.context_window == 128000
 
 
+def test_tencent_deepseek_v3_2() -> None:
+    """测试腾讯云 DeepSeek-V3.2（GA） / Test Tencent DeepSeek-V3.2 (GA)"""
+    config = get_specific_model_config("deepseek-v3.2")
+
+    assert config is not None
+    version, variant, capabilities = config
+    assert version == "v3.2"
+    assert variant == "v3.2"
+    assert capabilities is not None
+    assert capabilities.supports_thinking is True
+    assert capabilities.supports_function_calling is True
+    assert capabilities.supports_streaming is True
+    assert capabilities.max_tokens == 32000
+    assert capabilities.context_window == 128000
+
+
 def test_tencent_deepseek_pattern_matching() -> None:
     """测试腾讯云 DeepSeek 模式匹配 / Test Tencent DeepSeek pattern matching"""
     # 测试 V3 系列
@@ -158,8 +176,12 @@ def test_tencent_deepseek_no_collision_with_official() -> None:
     assert tencent_v3.variant == "base"
 
     # 不同的能力配置
-    assert official_chat.capabilities.max_tokens == 8000
+    # 官方 deepseek-chat 当前指向 V4-flash 非思考模式（1M 上下文，384K 输出）
+    assert official_chat.capabilities.max_tokens == 384_000
+    assert official_chat.capabilities.context_window == 1_000_000
+    # 腾讯云保持其自有的 DeepSeek 部署规格
     assert tencent_v3.capabilities.max_tokens == 16000
+    assert tencent_v3.capabilities.context_window == 64000
 
 
 def test_tencent_deepseek_invalid_model_names() -> None:

{whosellm-0.2.1 → whosellm-0.2.2}/tests/models/families/test_gpt5.py

@@ -193,7 +193,6 @@ def test_gpt5_4_model():
     assert m.capabilities.supports_computer_use is True
 
 
-
 def test_gpt5_4_with_date_suffix():
     """测试带日期的GPT-5.4模型 / Test GPT-5.4 with date suffix"""
     m = LLMeta("gpt-5.4-2026-03-05")
@@ -239,7 +238,6 @@ def test_gpt5_4_mini_model():
     assert m.capabilities.supports_computer_use is True
 
 
-
 def test_gpt5_4_mini_with_date_suffix():
     """测试带日期的GPT-5.4-mini模型 / Test GPT-5.4-mini with date suffix"""
     m = LLMeta("gpt-5.4-mini-2026-03-17")
@@ -267,7 +265,6 @@ def test_gpt5_4_nano_model():
     assert m.capabilities.supports_computer_use is False
 
 
-
 def test_gpt5_4_nano_with_date_suffix():
     """测试带日期的GPT-5.4-nano模型 / Test GPT-5.4-nano with date suffix"""
     m = LLMeta("gpt-5.4-nano-2026-03-17")

{whosellm-0.2.1 → whosellm-0.2.2}/tests/test_auto_register.py

@@ -102,7 +102,8 @@ class TestAutoRegister(unittest.TestCase):
         assert model.family == ModelFamily.DEEPSEEK
         assert model.provider == Provider.DEEPSEEK
         assert model.capabilities.supports_function_calling is True
-        assert model.capabilities.context_window == 128000
+        # V4 起 deepseek-chat 别名为 v4-flash 非思考模式，上下文扩展为 1M
+        assert model.capabilities.context_window == 1_000_000
 
     def test_auto_register_qwen_variant(self) -> None:
         """测试自动注册 Qwen 新型号 / Test auto-register Qwen new variant"""

{whosellm-0.2.1 → whosellm-0.2.2}/uv.lock

@@ -968,7 +968,7 @@ wheels = [
 
 [[package]]
 name = "whosellm"
-version = "0.2.1"
+version = "0.2.2"
 source = { editable = "." }
 dependencies = [
     { name = "parse" },

{whosellm-0.2.1 → whosellm-0.2.2}/whosellm/__init__.py

@@ -7,7 +7,7 @@
 LLMeta - 统一的大语言模型版本和能力管理库 / A unified LLM model version and capability management library
 """
 
-__version__ = "0.2.1"
+__version__ = "0.2.2"
 
 from whosellm.capabilities import ModelCapabilities
 from whosellm.model_version import LLMeta

whosellm-0.2.2/whosellm/models/families/deepseek/deepseek_official.py

@@ -0,0 +1,117 @@
+# filename: deepseek_official.py
+# @Time    : 2025/11/9 15:57
+# @Author  : Cascade AI
+"""
+DeepSeek 官方模型家族配置 / DeepSeek official model family configuration
+
+当前主力：deepseek-v4-flash / deepseek-v4-pro（1M 上下文，384K 最大输出，
+支持思考与非思考双模式）。
+Current primary models: deepseek-v4-flash / deepseek-v4-pro (1M context, 384K max output,
+both thinking and non-thinking modes supported).
+
+兼容别名：deepseek-chat / deepseek-reasoner（官方宣布未来将废弃，当前分别对应
+deepseek-v4-flash 的非思考与思考模式）。
+Legacy aliases: deepseek-chat / deepseek-reasoner (announced to be deprecated;
+currently map to deepseek-v4-flash non-thinking and thinking modes respectively).
+"""
+
+from whosellm.capabilities import ModelCapabilities
+from whosellm.models.base import ModelFamily
+from whosellm.models.config import ModelFamilyConfig, SpecificModelConfig
+from whosellm.provider import Provider
+
+# V4 系列共用能力基线 / Shared capability baseline for V4 series
+_V4_CAPABILITIES = ModelCapabilities(
+    supports_thinking=True,
+    supports_function_calling=True,
+    supports_streaming=True,
+    supports_json_outputs=True,
+    # DeepSeek 仅提供 response_format={type:"json_object"}，不支持 json_schema
+    # DeepSeek only supports response_format={type:"json_object"}, not json_schema
+    supports_structured_outputs=False,
+    max_tokens=384_000,
+    context_window=1_000_000,
+)
+
+DEEPSEEK = ModelFamilyConfig(
+    family=ModelFamily.DEEPSEEK,
+    provider=Provider.DEEPSEEK,
+    version_default="4.0",
+    variant_default="flash",
+    variant_priority_default=(0,),
+    patterns=[
+        # 版本号命名（V4 起开放给 API 调用） / Version-numbered naming (open to API since V4)
+        "deepseek-v{major:d}.{minor:d}-{variant:variant}",
+        "deepseek-v{major:d}.{minor:d}",
+        "deepseek-v{major:d}-{variant:variant}",
+        "deepseek-v{major:d}",
+        # 兼容别名 / Legacy aliases
+        "deepseek-chat-{suffix}",
+        "deepseek-chat",
+        "deepseek-reasoner-{suffix}",
+        "deepseek-reasoner",
+    ],
+    capabilities=_V4_CAPABILITIES,
+    specific_models={
+        # 当前旗舰：deepseek-v4-flash / Current flagship: deepseek-v4-flash
+        "deepseek-v4-flash": SpecificModelConfig(
+            version_default="4.0",
+            variant_default="flash",
+            variant_priority=(0,),
+            capabilities=_V4_CAPABILITIES,
+            patterns=[
+                "deepseek-v4-flash",
+            ],
+        ),
+        # 高阶版：deepseek-v4-pro / Professional tier: deepseek-v4-pro
+        "deepseek-v4-pro": SpecificModelConfig(
+            version_default="4.0",
+            variant_default="pro",
+            variant_priority=(4,),
+            capabilities=_V4_CAPABILITIES,
+            patterns=[
+                "deepseek-v4-pro",
+            ],
+        ),
+        # 兼容别名：deepseek-chat → v4-flash 非思考模式
+        # Legacy alias: deepseek-chat → v4-flash non-thinking mode
+        "deepseek-chat": SpecificModelConfig(
+            version_default="4.0",
+            variant_default="chat",
+            variant_priority=(1,),
+            capabilities=ModelCapabilities(
+                supports_thinking=False,
+                supports_function_calling=True,
+                supports_streaming=True,
+                supports_json_outputs=True,
+                supports_structured_outputs=False,
+                max_tokens=384_000,
+                context_window=1_000_000,
+            ),
+            patterns=[
+                "deepseek-chat-{suffix}",
+                "deepseek-chat",
+            ],
+        ),
+        # 兼容别名：deepseek-reasoner → v4-flash 思考模式
+        # Legacy alias: deepseek-reasoner → v4-flash thinking mode
+        "deepseek-reasoner": SpecificModelConfig(
+            version_default="4.0",
+            variant_default="reasoner",
+            variant_priority=(2,),
+            capabilities=ModelCapabilities(
+                supports_thinking=True,
+                supports_function_calling=True,
+                supports_streaming=True,
+                supports_json_outputs=True,
+                supports_structured_outputs=False,
+                max_tokens=384_000,
+                context_window=1_000_000,
+            ),
+            patterns=[
+                "deepseek-reasoner-{suffix}",
+                "deepseek-reasoner",
+            ],
+        ),
+    },
+)

{whosellm-0.2.1 → whosellm-0.2.2}/whosellm/models/families/deepseek/tencent.py

@@ -78,12 +78,14 @@ DEEPSEEK_TENCENT = ModelFamilyConfig(
             ],
         ),
         # DeepSeek-R1-0528
+        # 腾讯云 LKE 文档明确列为 Function Calling 不支持（位于"不支持设置的功能"）
+        # Tencent LKE docs explicitly list Function Calling as unsupported for this model
         "deepseek-r1-0528": SpecificModelConfig(
             version_default="r1-0528",
             variant_default="r1-0528",
             capabilities=ModelCapabilities(
                 supports_thinking=True,
-                supports_function_calling=True,
+                supports_function_calling=False,
                 supports_streaming=True,
                 max_tokens=16000,
                 context_window=128000,
@@ -125,7 +127,8 @@ DEEPSEEK_TENCENT = ModelFamilyConfig(
                 "deepseek-v3.1-terminus",
             ],
         ),
-        # DeepSeek-V3.2-Exp
+        # DeepSeek-V3.2-Exp（已被 deepseek-v3.2 GA 版替代，保留以兼容历史调用）
+        # DeepSeek-V3.2-Exp (superseded by deepseek-v3.2 GA; kept for backward compatibility)
         "deepseek-v3.2-exp": SpecificModelConfig(
             version_default="v3.2-exp",
             variant_default="v3.2-exp",
@@ -141,5 +144,22 @@ DEEPSEEK_TENCENT = ModelFamilyConfig(
                 "deepseek-v3.2-exp",
             ],
         ),
+        # DeepSeek-V3.2（GA 版，685B MoE，稀疏注意力，可选思考模式）
+        # DeepSeek-V3.2 (GA, 685B MoE, sparse attention, optional thinking mode)
+        "deepseek-v3.2": SpecificModelConfig(
+            version_default="v3.2",
+            variant_default="v3.2",
+            capabilities=ModelCapabilities(
+                supports_thinking=True,
+                supports_function_calling=True,
+                supports_streaming=True,
+                max_tokens=32000,
+                context_window=128000,
+            ),
+            variant_priority=(1,),
+            patterns=[
+                "deepseek-v3.2",
+            ],
+        ),
     },
 )

{whosellm-0.2.1 → whosellm-0.2.2}/whosellm/models/families/openai/openai_gpt_4_1.py

@@ -31,7 +31,6 @@ GPT_4_1 = ModelFamilyConfig(
         supports_file_search=True,
         supports_image_generation=True,
         supports_code_interpreter=True,
-
         max_tokens=32768,
         context_window=1_047_576,
     ),

{whosellm-0.2.1 → whosellm-0.2.2}/whosellm/models/families/openai/openai_gpt_5.py

@@ -33,7 +33,6 @@ GPT_5 = ModelFamilyConfig(
         supports_file_search=True,
         supports_image_generation=True,
         supports_code_interpreter=True,
-
         max_tokens=128_000,
         context_window=400_000,
     ),
@@ -66,7 +65,6 @@ GPT_5 = ModelFamilyConfig(
                 supports_file_search=True,
                 supports_image_generation=False,
                 supports_code_interpreter=True,
-        
                 max_tokens=128_000,
                 context_window=400_000,
             ),
@@ -93,7 +91,6 @@ GPT_5 = ModelFamilyConfig(
                 supports_file_search=True,
                 supports_image_generation=False,
                 supports_code_interpreter=False,
-        
                 max_tokens=16_384,
                 context_window=128_000,
             ),
@@ -121,7 +118,6 @@ GPT_5 = ModelFamilyConfig(
                 supports_image_generation=True,
                 supports_code_interpreter=True,
                 supports_computer_use=True,
-        
                 max_tokens=128_000,
                 context_window=400_000,
             ),

{whosellm-0.2.1 → whosellm-0.2.2}/whosellm/models/families/openai/openai_gpt_5_1.py

@@ -25,7 +25,6 @@ GPT_5_1 = ModelFamilyConfig(
         supports_file_search=True,
         supports_image_generation=True,
         supports_code_interpreter=True,
-
         max_tokens=128_000,
         context_window=1_050_000,
     ),

{whosellm-0.2.1 → whosellm-0.2.2}/whosellm/models/families/openai/openai_gpt_5_2.py

@@ -25,7 +25,6 @@ GPT_5_2 = ModelFamilyConfig(
         supports_file_search=True,
         supports_image_generation=True,
         supports_code_interpreter=True,
-
         max_tokens=128_000,
         context_window=1_050_000,
     ),
@@ -57,7 +56,6 @@ GPT_5_2 = ModelFamilyConfig(
                 supports_image_generation=True,
                 supports_code_interpreter=True,
                 supports_computer_use=True,
-        
                 max_tokens=128_000,
                 context_window=1_050_000,
             ),

{whosellm-0.2.1 → whosellm-0.2.2}/whosellm/models/families/openai/openai_gpt_5_4.py

@@ -40,7 +40,6 @@ GPT_5_4 = ModelFamilyConfig(
         supports_image_generation=True,
         supports_code_interpreter=True,
         supports_computer_use=True,
-
         max_tokens=128_000,
         context_window=1_050_000,
     ),
@@ -72,7 +71,6 @@ GPT_5_4 = ModelFamilyConfig(
                 supports_image_generation=True,
                 supports_code_interpreter=False,
                 supports_computer_use=True,
-        
                 max_tokens=128_000,
                 context_window=1_050_000,
             ),
@@ -98,7 +96,6 @@ GPT_5_4 = ModelFamilyConfig(
                 supports_image_generation=True,
                 supports_code_interpreter=True,
                 supports_computer_use=True,
-        
                 max_tokens=128_000,
                 context_window=400_000,
             ),
@@ -124,7 +121,6 @@ GPT_5_4 = ModelFamilyConfig(
                 supports_image_generation=True,
                 supports_code_interpreter=True,
                 supports_computer_use=False,
-        
                 max_tokens=128_000,
                 context_window=400_000,
             ),

{whosellm-0.2.1 → whosellm-0.2.2}/whosellm/models/families/zhipu.py

@@ -260,7 +260,6 @@ GLM_TEXT = ModelFamilyConfig(
                 supports_function_calling=True,
                 supports_structured_outputs=False,
                 supports_streaming=True,
-
                 max_tokens=128000,
                 context_window=200000,
             ),
@@ -275,7 +274,6 @@ GLM_TEXT = ModelFamilyConfig(
                 supports_function_calling=True,
                 supports_structured_outputs=False,
                 supports_streaming=True,
-
                 max_tokens=128000,
                 context_window=200000,
             ),
@@ -289,7 +287,6 @@ GLM_TEXT = ModelFamilyConfig(
                 supports_function_calling=True,
                 supports_structured_outputs=False,
                 supports_streaming=True,
-
                 max_tokens=128000,
                 context_window=200000,
             ),
@@ -308,7 +305,6 @@ GLM_TEXT = ModelFamilyConfig(
                 supports_function_calling=True,
                 supports_structured_outputs=False,
                 supports_streaming=True,
-
                 max_tokens=128000,
                 context_window=200000,
             ),
@@ -329,7 +325,6 @@ GLM_TEXT = ModelFamilyConfig(
                 supports_streaming=True,
                 supports_structured_outputs=False,
                 supports_web_search=True,
-
                 max_tokens=128000,
                 context_window=200000,
             ),

whosellm-0.2.1/whosellm/models/families/deepseek/deepseek_official.py

@@ -1,72 +0,0 @@
-# filename: deepseek_official.py
-# @Time    : 2025/11/9 15:57
-# @Author  : Cascade AI
-"""
-DeepSeek 官方模型家族配置 / DeepSeek official model family configuration
-
-包含 deepseek-chat 与 deepseek-reasoner 模型
-Includes deepseek-chat and deepseek-reasoner models
-"""
-
-from whosellm.capabilities import ModelCapabilities
-from whosellm.models.base import ModelFamily
-from whosellm.models.config import ModelFamilyConfig, SpecificModelConfig
-from whosellm.provider import Provider
-
-DEEPSEEK = ModelFamilyConfig(
-    family=ModelFamily.DEEPSEEK,
-    provider=Provider.DEEPSEEK,
-    version_default="1.0",
-    variant_default="base",
-    variant_priority_default=(1,),
-    patterns=[
-        "deepseek-chat-{suffix}",
-        "deepseek-chat",
-        "deepseek-reasoner-{suffix}",
-        "deepseek-reasoner",
-    ],
-    capabilities=ModelCapabilities(
-        supports_function_calling=True,
-        supports_streaming=True,
-        supports_structured_outputs=False,
-        max_tokens=8000,
-        context_window=128000,
-    ),
-    specific_models={
-        "deepseek-chat": SpecificModelConfig(
-            version_default="1.0",
-            variant_default="chat",
-            capabilities=ModelCapabilities(
-                supports_function_calling=True,
-                supports_streaming=True,
-                supports_structured_outputs=False,
-                max_tokens=8000,
-                context_window=128000,
-            ),
-            variant_priority=(1,),
-            patterns=[
-                "deepseek-chat-{suffix}",
-                "deepseek-chat",
-            ],
-        ),
-        "deepseek-reasoner": SpecificModelConfig(
-            version_default="1.0",
-            variant_default="reasoner",
-            capabilities=ModelCapabilities(
-                supports_thinking=True,
-                supports_streaming=True,
-                # DS Reasoner模型原生是不支持工具调用的，但是官方做了优化，如果向R1模型发起带有Tools调用的请求，会自动路由至Chat模型。
-                # 因此在这里标记为支持FunctionCall
-                supports_function_calling=True,
-                supports_structured_outputs=False,
-                max_tokens=64000,
-                context_window=128000,
-            ),
-            variant_priority=(2,),
-            patterns=[
-                "deepseek-reasoner-{suffix}",
-                "deepseek-reasoner",
-            ],
-        ),
-    },
-)