Merge 3a9257d3c6 into 252c71c0b2

README.md

@@ -366,6 +366,7 @@ For MCP OAuth, an upstream may advertise dynamic client registration but refuse
 | [Text Completion Codestral (`text-completion-codestral`)](https://docs.litellm.ai/docs/providers/codestral) | ✅ | ✅ | ✅ |  |  |  |  |  |  |  |
 | [Text Completion OpenAI (`text-completion-openai`)](https://docs.litellm.ai/docs/providers/text_completion_openai) | ✅ | ✅ | ✅ |  |  | ✅ | ✅ | ✅ | ✅ |  |
 | [Together AI (`together_ai`)](https://docs.litellm.ai/docs/providers/togetherai) | ✅ | ✅ | ✅ |  |  |  |  |  |  |  |
+| [TokenLab (`tokenlab`)](https://docs.tokenlab.sh/integrations/litellm) | ✅ | ✅ | ✅ | ✅ |  |  |  |  |  |  |
 | [Topaz (`topaz`)](https://docs.litellm.ai/docs/providers/topaz) | ✅ | ✅ | ✅ |  |  |  |  |  |  |  |
 | [Triton (`triton`)](https://docs.litellm.ai/docs/providers/triton-inference-server) | ✅ | ✅ | ✅ |  |  |  |  |  |  |  |
 | [V0 (`v0`)](https://docs.litellm.ai/docs/providers/v0) | ✅ | ✅ | ✅ |  |  |  |  |  |  |  |

litellm/__init__.py

@@ -680,6 +680,7 @@ qwen_ai_platform_models: Set = set()
 moonshot_models: Set = set()
 publicai_models: Set = set()
 darkbloom_models: Set = set()
+tokenlab_models: Set = set()  # mutable-ok: populated while loading the static model registry
 v0_models: Set = set()
 morph_models: Set = set()
 lambda_ai_models: Set = set()
@@ -939,6 +940,8 @@ def _populate_provider_model_sets(model_cost_map: Dict) -> None:
             publicai_models.add(key)
         elif value.get("litellm_provider") == "darkbloom":
             darkbloom_models.add(key)
+        elif value.get("litellm_provider") == "tokenlab":
+            tokenlab_models.add(key)
         elif value.get("litellm_provider") == "v0":
             v0_models.add(key)
         elif value.get("litellm_provider") == "morph":
@@ -1099,6 +1102,7 @@ model_list = list(
     | moonshot_models
     | publicai_models
     | darkbloom_models
+    | tokenlab_models
     | v0_models
     | morph_models
     | lambda_ai_models
@@ -1208,6 +1212,7 @@ def _build_models_by_provider() -> dict:
         "moonshot": moonshot_models,
         "publicai": publicai_models,
         "darkbloom": darkbloom_models,
+        "tokenlab": tokenlab_models,
         "v0": v0_models,
         "morph": morph_models,
         "lambda_ai": lambda_ai_models,

litellm/constants.py

@@ -893,6 +893,7 @@ openai_compatible_endpoints: Final[list] = [
     "https://api.inference.wandb.ai/v1",
     "https://api.clarifai.com/v2/ext/openai/v1",
     "https://api.libertai.io/v1",
+    "https://api.tokenlab.sh/v1",
     "https://pinstripes.io/v1",
     "https://api.meta.ai/v1",
     "https://api.cognition.ai/v1",
@@ -965,6 +966,7 @@ openai_compatible_providers: Final[list] = [
     "docker_model_runner",
     "ragflow",
     "pinstripes",  # Pinstripes - JSON-configured provider
+    "tokenlab",  # TokenLab - JSON-configured provider
     "darkbloom",
     "meta",  # Meta Model API (Muse Spark) - JSON-configured provider
     "cognition",

litellm/litellm_core_utils/get_llm_provider_logic.py

@@ -363,6 +363,10 @@ def get_llm_provider(
                     elif endpoint == "https://api.inference.wandb.ai/v1":
                         custom_llm_provider = "wandb"
                         dynamic_api_key = get_secret_str("WANDB_API_KEY")
+                    elif endpoint == "https://api.tokenlab.sh/v1":
+                        custom_llm_provider = "tokenlab"
+                        tokenlab_config: Final = JSONProviderRegistry.get("tokenlab")
+                        dynamic_api_key = get_secret_str(tokenlab_config.api_key_env) if tokenlab_config else None
                     elif endpoint == "https://pinstripes.io/v1":
                         custom_llm_provider = "pinstripes"
                         dynamic_api_key = get_secret_str("PINSTRIPES_API_KEY")

litellm/llms/openai_like/json_loader.py

@@ -78,6 +78,14 @@ class JSONProviderRegistry:
             return False
         return "/v1/responses" in provider.supported_endpoints
 
+    @classmethod
+    def supports_anthropic_messages_api(cls, slug: str) -> bool:
+        """Check if a JSON provider supports the Anthropic Messages API"""
+        provider = cls._providers.get(slug)
+        if provider is None:
+            return False
+        return "/v1/messages" in provider.supported_endpoints
+
     @classmethod
     def list_providers(cls) -> list:
         """List all registered provider slugs"""

litellm/llms/openai_like/providers.json

@@ -180,6 +180,15 @@
     "api_key_env": "COGNITION_API_KEY",
     "api_base_env": "COGNITION_API_BASE"
   },
+  "tokenlab": {
+    "base_url": "https://api.tokenlab.sh/v1",
+    "api_key_env": "TOKENLAB_API_KEY",
+    "api_base_env": "TOKENLAB_API_BASE",
+    "param_mappings": {
+      "max_completion_tokens": "max_tokens"
+    },
+    "supported_endpoints": ["/v1/chat/completions", "/v1/messages", "/v1/responses", "/v1/embeddings"]
+  },
   "pinstripes": {
     "base_url": "https://pinstripes.io/v1",
     "api_key_env": "PINSTRIPES_API_KEY",

litellm/provider_endpoints_support_backup.json

@@ -1628,6 +1628,23 @@
         "a2a": false
       }
     },
+    "tokenlab": {
+      "display_name": "TokenLab (`tokenlab`)",
+      "url": "https://docs.tokenlab.sh/integrations/litellm",
+      "endpoints": {
+        "chat_completions": true,
+        "messages": true,
+        "responses": true,
+        "embeddings": true,
+        "image_generations": false,
+        "audio_transcriptions": false,
+        "audio_speech": false,
+        "moderations": false,
+        "batches": false,
+        "rerank": false,
+        "a2a": false
+      }
+    },
     "nvidia_nim": {
       "display_name": "Nvidia NIM (`nvidia_nim`)",
       "url": "https://docs.litellm.ai/docs/providers/nvidia_nim",

litellm/types/utils.py

@@ -4042,6 +4042,7 @@ class LlmProviders(str, Enum):
     PARASAIL = "parasail"
     XIAOMI_MIMO = "xiaomi_mimo"
     TENSORMESH = "tensormesh"
+    TOKENLAB = "tokenlab"
     LIBERTAI = "libertai"
     PINSTRIPES = "pinstripes"
     COGNITION = "cognition"

provider_endpoints_support.json

@@ -2366,6 +2366,23 @@
         "a2a": false
       }
     },
+    "tokenlab": {
+      "display_name": "TokenLab (`tokenlab`)",
+      "url": "https://docs.tokenlab.sh/integrations/litellm",
+      "endpoints": {
+        "chat_completions": true,
+        "messages": true,
+        "responses": true,
+        "embeddings": true,
+        "image_generations": false,
+        "audio_transcriptions": false,
+        "audio_speech": false,
+        "moderations": false,
+        "batches": false,
+        "rerank": false,
+        "a2a": false
+      }
+    },
     "tensormesh": {
       "display_name": "Tensormesh (`tensormesh`)",
       "url": "https://docs.litellm.ai/docs/providers/tensormesh",

tests/test_litellm/llms/openai_like/test_tokenlab_provider.py

@@ -0,0 +1,176 @@
+import pytest
+
+import litellm
+
+
+TOKENLAB_API_BASE = "https://api.tokenlab.sh/v1"
+TOKENLAB_SAMPLE_MODELS = [
+    "tokenlab/gpt-5.5",
+    "tokenlab/claude-opus-4-8",
+    "tokenlab/gemini-3.5-flash",
+    "tokenlab/deepseek-v4-pro",
+    "tokenlab/qwen3.7-max",
+    "tokenlab/text-embedding-3-small",
+]
+
+
+def test_tokenlab_json_registry():
+    from litellm.llms.openai_like.json_loader import JSONProviderRegistry
+
+    assert litellm.LlmProviders.TOKENLAB.value == "tokenlab"
+    assert litellm.LlmProviders("tokenlab") == litellm.LlmProviders.TOKENLAB
+    assert JSONProviderRegistry.exists("tokenlab")
+    config = JSONProviderRegistry.get("tokenlab")
+    assert config is not None
+    assert config.base_url == TOKENLAB_API_BASE
+    assert config.api_key_env == "TOKENLAB_API_KEY"
+    assert config.api_base_env == "TOKENLAB_API_BASE"
+    assert config.param_mappings["max_completion_tokens"] == "max_tokens"
+    assert config.supported_endpoints == [
+        "/v1/chat/completions",
+        "/v1/messages",
+        "/v1/responses",
+        "/v1/embeddings",
+    ]
+    assert JSONProviderRegistry.supports_responses_api("tokenlab") is True
+    assert JSONProviderRegistry.supports_anthropic_messages_api("tokenlab") is True
+
+
+def test_tokenlab_provider_detection_by_prefix():
+    from litellm.litellm_core_utils.get_llm_provider_logic import get_llm_provider
+
+    model, provider, _, api_base = get_llm_provider("tokenlab/gpt-5.5")
+
+    assert model == "gpt-5.5"
+    assert provider == "tokenlab"
+    assert api_base == TOKENLAB_API_BASE
+
+
+def test_tokenlab_api_base_override():
+    from litellm.litellm_core_utils.get_llm_provider_logic import get_llm_provider
+
+    _, provider, api_key, api_base = get_llm_provider(
+        model="tokenlab/gpt-5.5",
+        api_base="https://custom.tokenlab.example/v1",
+        api_key="sk-test",
+    )
+
+    assert provider == "tokenlab"
+    assert api_base == "https://custom.tokenlab.example/v1"
+    assert api_key == "sk-test"
+
+
+def test_tokenlab_url_autodetection():
+    from litellm.litellm_core_utils.get_llm_provider_logic import get_llm_provider
+
+    _, provider, _, api_base = get_llm_provider(
+        model="gpt-5.5",
+        api_base=TOKENLAB_API_BASE,
+    )
+
+    assert provider == "tokenlab"
+    assert api_base == TOKENLAB_API_BASE
+
+
+def test_tokenlab_chat_complete_url():
+    from litellm.llms.openai_like.dynamic_config import create_config_class
+    from litellm.llms.openai_like.json_loader import JSONProviderRegistry
+
+    config = create_config_class(JSONProviderRegistry.get("tokenlab"))()
+
+    assert (
+        config.get_complete_url(
+            api_base=None,
+            api_key=None,
+            model="gpt-5.5",
+            optional_params={},
+            litellm_params={},
+        )
+        == "https://api.tokenlab.sh/v1/chat/completions"
+    )
+
+
+def test_tokenlab_responses_api_config():
+    from litellm.llms.openai.responses.transformation import OpenAIResponsesAPIConfig
+    from litellm.utils import ProviderConfigManager
+
+    config = ProviderConfigManager.get_provider_responses_api_config(
+        provider="tokenlab",
+        model="tokenlab/gpt-5.5",
+    )
+
+    assert isinstance(config, OpenAIResponsesAPIConfig)
+    assert config.custom_llm_provider == "tokenlab"
+    assert config.get_complete_url(api_base=None, litellm_params={}) == "https://api.tokenlab.sh/v1/responses"
+
+
+def test_tokenlab_anthropic_messages_config():
+    from litellm.llms.openai_like.messages.transformation import (
+        OpenAILikeAnthropicMessagesConfig,
+    )
+    from litellm.utils import ProviderConfigManager
+
+    config = ProviderConfigManager.get_provider_anthropic_messages_config(
+        provider=litellm.LlmProviders.TOKENLAB,
+        model="claude-opus-4-8",
+    )
+
+    assert isinstance(config, OpenAILikeAnthropicMessagesConfig)
+    assert (
+        config.get_complete_url(
+            api_base=TOKENLAB_API_BASE,
+            api_key="sk-test",
+            model="claude-opus-4-8",
+            optional_params={},
+            litellm_params={},
+        )
+        == "https://api.tokenlab.sh/v1/messages"
+    )
+
+
+class TestTokenLabCostMap:
+    @pytest.fixture(autouse=True)
+    def _use_local_model_cost_map(self, monkeypatch):
+        original_model_cost = litellm.model_cost
+        monkeypatch.setenv("LITELLM_LOCAL_MODEL_COST_MAP", "True")
+        litellm.model_cost = litellm.get_model_cost_map(url="")
+        litellm.get_model_info.cache_clear()
+        try:
+            yield
+        finally:
+            litellm.model_cost = original_model_cost
+            litellm.get_model_info.cache_clear()
+
+    def test_sample_models_registered(self):
+        for model in TOKENLAB_SAMPLE_MODELS:
+            info = litellm.get_model_info(model)
+            assert info["litellm_provider"] == "tokenlab"
+            assert info["mode"] in {"chat", "embedding"}
+
+    def test_claude_model_cost_map_keeps_schema_supported_endpoints(self):
+        from litellm.llms.openai_like.json_loader import JSONProviderRegistry
+
+        endpoints = litellm.model_cost["tokenlab/claude-opus-4-8"]["supported_endpoints"]
+        assert endpoints == ["/v1/chat/completions", "/v1/responses"]
+
+        provider_config = JSONProviderRegistry.get("tokenlab")
+        assert provider_config is not None
+        assert "/v1/messages" in provider_config.supported_endpoints
+
+    def test_chat_cost_is_wired(self):
+        prompt_cost, completion_cost = litellm.cost_per_token(
+            model="tokenlab/gpt-5.5",
+            prompt_tokens=1_000_000,
+            completion_tokens=1_000_000,
+        )
+        assert prompt_cost == pytest.approx(1.5)
+        assert completion_cost == pytest.approx(9.0)
+
+    def test_embedding_cost_is_wired(self):
+        prompt_cost, completion_cost = litellm.cost_per_token(
+            model="tokenlab/text-embedding-3-small",
+            prompt_tokens=1_000_000,
+            completion_tokens=0,
+        )
+        assert prompt_cost == pytest.approx(0.02)
+        assert completion_cost == 0