diff --git a/README.md b/README.md index 4feb6e00ffa..5f83f2389d6 100644 --- a/README.md +++ b/README.md @@ -427,6 +427,7 @@ Set `LITELLM_PROXY_API_BASE` and `LITELLM_PROXY_API_KEY` and every model call th | [Watsonx Text (`watsonx_text`)](https://docs.litellm.ai/docs/providers/watsonx) | ✅ | ✅ | ✅ | | | | | | | | | [xAI (`xai`)](https://docs.litellm.ai/docs/providers/xai) | ✅ | ✅ | ✅ | | | | | | | | | [Xinference (`xinference`)](https://docs.litellm.ai/docs/providers/xinference) | | | | ✅ | | | | | | | +| [ZeroGPU (`zerogpu`)](https://docs.litellm.ai/docs/providers/zerogpu) | ✅ | ✅ | ✅ | | | | | | | | [**Read the Docs**](https://docs.litellm.ai/docs/) diff --git a/litellm-rust/crates/core-utils/src/exception_mapping_utils/original.rs b/litellm-rust/crates/core-utils/src/exception_mapping_utils/original.rs index 6517b093bbd..615d530af19 100644 --- a/litellm-rust/crates/core-utils/src/exception_mapping_utils/original.rs +++ b/litellm-rust/crates/core-utils/src/exception_mapping_utils/original.rs @@ -113,6 +113,7 @@ const OPENAI_COMPATIBLE_PROVIDERS: &[&str] = &[ "meta", "cognition", "scx-ai", + "zerogpu", ]; impl ExceptionFamily { diff --git a/litellm/__init__.py b/litellm/__init__.py index c2c13a12eea..e2b29ae9cb9 100644 --- a/litellm/__init__.py +++ b/litellm/__init__.py @@ -680,6 +680,7 @@ moonshot_models: Set = set() publicai_models: Set = set() darkbloom_models: Set = set() openinfer_models: Final[Set[str]] = set() +zerogpu_models: Final[Set[str]] = set() v0_models: Set = set() morph_models: Set = set() opencode_go_models: Final[set[str]] = set() # mutable-ok: populated when the shared model catalog loads @@ -967,6 +968,8 @@ def _populate_provider_model_sets(model_cost_map: Mapping[str, Mapping[str, obje darkbloom_models.add(key) elif value.get("litellm_provider") == "openinfer": openinfer_models.add(key) + elif value.get("litellm_provider") == "zerogpu": + zerogpu_models.add(key) elif value.get("litellm_provider") == "v0": v0_models.add(key) elif value.get("litellm_provider") == "morph": @@ -1138,6 +1141,7 @@ model_list = list( | publicai_models | darkbloom_models | openinfer_models + | zerogpu_models | v0_models | morph_models | opencode_go_models @@ -1256,6 +1260,7 @@ def _build_models_by_provider() -> dict: "publicai": publicai_models, "darkbloom": darkbloom_models, "openinfer": openinfer_models, + "zerogpu": zerogpu_models, "v0": v0_models, "morph": morph_models, "opencode_go": opencode_go_models, diff --git a/litellm/constants.py b/litellm/constants.py index 2901b7d8948..4757803763a 100644 --- a/litellm/constants.py +++ b/litellm/constants.py @@ -1025,6 +1025,7 @@ openai_compatible_endpoints: Final[list[str]] = [ "api.x.ai/v1", "ollama.com", "api.galadriel.ai/v1", + "https://api.zerogpu.ai/v1", "api.llama.com/compat/v1/", "api.featherless.ai/v1", "inference.api.nscale.com/v1", @@ -1094,6 +1095,7 @@ openai_compatible_providers: Final[list[str]] = [ "llamafile", "lm_studio", "galadriel", + "zerogpu", "github_copilot", # GitHub Copilot Chat API "chatgpt", # ChatGPT subscription API "novita", diff --git a/litellm/llms/openai_like/providers.json b/litellm/llms/openai_like/providers.json index 115d914d3ac..cdac0b7ba3c 100644 --- a/litellm/llms/openai_like/providers.json +++ b/litellm/llms/openai_like/providers.json @@ -58,6 +58,12 @@ "max_completion_tokens": "max_tokens" } }, + "zerogpu": { + "base_url": "https://api.zerogpu.ai/v1", + "api_key_env": "ZEROGPU_API_KEY", + "api_base_env": "ZEROGPU_API_BASE", + "supported_endpoints": ["/v1/chat/completions", "/v1/responses"] + }, "chutes": { "base_url": "https://llm.chutes.ai/v1/", "api_key_env": "CHUTES_API_KEY", diff --git a/litellm/model_prices_and_context_window_backup.json b/litellm/model_prices_and_context_window_backup.json index b820b058eb3..aa29c3ba36e 100644 --- a/litellm/model_prices_and_context_window_backup.json +++ b/litellm/model_prices_and_context_window_backup.json @@ -65151,6 +65151,139 @@ "supports_reasoning": true, "supports_tool_choice": true }, + "zerogpu/deepseek-v4.1-flash": { + "cache_read_input_token_cost": 3e-09, + "input_cost_per_token": 3e-07, + "litellm_provider": "zerogpu", + "max_input_tokens": 1048576, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "output_cost_per_token": 1.2e-06, + "source": "https://docs.zerogpu.ai/docs/model-catalog", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true + }, + "zerogpu/glm-5.3-flash": { + "cache_read_input_token_cost": 1.5e-08, + "input_cost_per_token": 1e-07, + "litellm_provider": "zerogpu", + "max_input_tokens": 1048576, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 3.5e-07, + "source": "https://docs.zerogpu.ai/docs/model-catalog", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true + }, + "zerogpu/gpt-4.1-mini": { + "cache_read_input_token_cost": 1e-07, + "input_cost_per_token": 4e-07, + "litellm_provider": "zerogpu", + "max_input_tokens": 1047576, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 1.6e-06, + "source": "https://docs.zerogpu.ai/docs/model-catalog", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_response_schema": true, + "supports_tool_choice": true + }, + "zerogpu/gpt-5.4-nano": { + "cache_read_input_token_cost": 2e-08, + "input_cost_per_token": 2e-07, + "litellm_provider": "zerogpu", + "max_input_tokens": 272000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 1.25e-06, + "source": "https://docs.zerogpu.ai/docs/model-catalog", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true + }, + "zerogpu/gpt-5.6-luna": { + "cache_read_input_token_cost": 2e-07, + "input_cost_per_token": 2e-07, + "litellm_provider": "zerogpu", + "max_input_tokens": 272000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 1.2e-06, + "source": "https://docs.zerogpu.ai/docs/model-catalog", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true + }, + "zerogpu/gpt-oss-120b": { + "cache_read_input_token_cost": 3e-08, + "input_cost_per_token": 1.5e-07, + "litellm_provider": "zerogpu", + "max_input_tokens": 128000, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 6e-07, + "source": "https://docs.zerogpu.ai/docs/model-catalog", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true + }, + "zerogpu/LFM2.5-1.2B-Instruct": { + "input_cost_per_token": 2e-08, + "litellm_provider": "zerogpu", + "max_input_tokens": 32768, + "max_output_tokens": 8192, + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 5e-08, + "source": "https://docs.zerogpu.ai/docs/model-catalog" + }, + "zerogpu/llama-3.1-8b-instruct-fast": { + "cache_read_input_token_cost": 2.5e-08, + "input_cost_per_token": 1.5e-07, + "litellm_provider": "zerogpu", + "max_input_tokens": 131072, + "max_output_tokens": 8192, + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 2.8e-07, + "source": "https://docs.zerogpu.ai/docs/model-catalog", + "supports_function_calling": true, + "supports_response_schema": true, + "supports_tool_choice": true + }, + "zerogpu/qwen3-30b-a3b-fp8": { + "cache_read_input_token_cost": 5e-08, + "input_cost_per_token": 1e-07, + "litellm_provider": "zerogpu", + "max_input_tokens": 32768, + "max_output_tokens": 16384, + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 4.5e-07, + "source": "https://docs.zerogpu.ai/docs/model-catalog", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true + }, "together_ai/Qwen/Qwen3.8-Flash": { "input_cost_per_token": 1.5e-07, "litellm_provider": "together_ai", diff --git a/litellm/provider_endpoints_support_backup.json b/litellm/provider_endpoints_support_backup.json index 49b667af819..3fd5a1cfcfd 100644 --- a/litellm/provider_endpoints_support_backup.json +++ b/litellm/provider_endpoints_support_backup.json @@ -2529,6 +2529,23 @@ "interactions": true } }, + "zerogpu": { + "display_name": "ZeroGPU (`zerogpu`)", + "url": "https://docs.litellm.ai/docs/providers/zerogpu", + "endpoints": { + "chat_completions": true, + "messages": true, + "responses": true, + "embeddings": false, + "image_generations": false, + "audio_transcriptions": false, + "audio_speech": false, + "moderations": false, + "batches": false, + "rerank": false, + "a2a": false + } + }, "ragflow": { "display_name": "RAGFlow (`ragflow`)", "url": "https://docs.litellm.ai/docs/providers/ragflow", diff --git a/litellm/proxy/public_endpoints/provider_create_fields.json b/litellm/proxy/public_endpoints/provider_create_fields.json index d96f2976508..adc39147c7d 100644 --- a/litellm/proxy/public_endpoints/provider_create_fields.json +++ b/litellm/proxy/public_endpoints/provider_create_fields.json @@ -4117,6 +4117,34 @@ ], "default_model_placeholder": "gpt-3.5-turbo" }, + { + "provider": "ZeroGPU", + "provider_display_name": "ZeroGPU", + "litellm_provider": "zerogpu", + "credential_fields": [ + { + "key": "api_base", + "label": "API Base", + "placeholder": "https://api.zerogpu.ai/v1", + "tooltip": null, + "required": false, + "field_type": "text", + "options": null, + "default_value": null + }, + { + "key": "api_key", + "label": "API Key", + "placeholder": null, + "tooltip": null, + "required": true, + "field_type": "password", + "options": null, + "default_value": null + } + ], + "default_model_placeholder": "zerogpu/gpt-oss-120b" + }, { "provider": "CURSOR", "provider_display_name": "Cursor", diff --git a/litellm/types/utils.py b/litellm/types/utils.py index 7e8645fd0dd..c677db6a5a6 100644 --- a/litellm/types/utils.py +++ b/litellm/types/utils.py @@ -4585,6 +4585,7 @@ class LlmProviders(str, Enum): PRISM = "prism" OPENINFER = "openinfer" INFERENCEAPIS = "inferenceapis" + ZEROGPU = "zerogpu" DARKBLOOM = "darkbloom" META = "meta" SAIL = "sail" diff --git a/model_prices_and_context_window.json b/model_prices_and_context_window.json index b820b058eb3..aa29c3ba36e 100644 --- a/model_prices_and_context_window.json +++ b/model_prices_and_context_window.json @@ -65151,6 +65151,139 @@ "supports_reasoning": true, "supports_tool_choice": true }, + "zerogpu/deepseek-v4.1-flash": { + "cache_read_input_token_cost": 3e-09, + "input_cost_per_token": 3e-07, + "litellm_provider": "zerogpu", + "max_input_tokens": 1048576, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "output_cost_per_token": 1.2e-06, + "source": "https://docs.zerogpu.ai/docs/model-catalog", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true + }, + "zerogpu/glm-5.3-flash": { + "cache_read_input_token_cost": 1.5e-08, + "input_cost_per_token": 1e-07, + "litellm_provider": "zerogpu", + "max_input_tokens": 1048576, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 3.5e-07, + "source": "https://docs.zerogpu.ai/docs/model-catalog", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true + }, + "zerogpu/gpt-4.1-mini": { + "cache_read_input_token_cost": 1e-07, + "input_cost_per_token": 4e-07, + "litellm_provider": "zerogpu", + "max_input_tokens": 1047576, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 1.6e-06, + "source": "https://docs.zerogpu.ai/docs/model-catalog", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_response_schema": true, + "supports_tool_choice": true + }, + "zerogpu/gpt-5.4-nano": { + "cache_read_input_token_cost": 2e-08, + "input_cost_per_token": 2e-07, + "litellm_provider": "zerogpu", + "max_input_tokens": 272000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 1.25e-06, + "source": "https://docs.zerogpu.ai/docs/model-catalog", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true + }, + "zerogpu/gpt-5.6-luna": { + "cache_read_input_token_cost": 2e-07, + "input_cost_per_token": 2e-07, + "litellm_provider": "zerogpu", + "max_input_tokens": 272000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 1.2e-06, + "source": "https://docs.zerogpu.ai/docs/model-catalog", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true + }, + "zerogpu/gpt-oss-120b": { + "cache_read_input_token_cost": 3e-08, + "input_cost_per_token": 1.5e-07, + "litellm_provider": "zerogpu", + "max_input_tokens": 128000, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 6e-07, + "source": "https://docs.zerogpu.ai/docs/model-catalog", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true + }, + "zerogpu/LFM2.5-1.2B-Instruct": { + "input_cost_per_token": 2e-08, + "litellm_provider": "zerogpu", + "max_input_tokens": 32768, + "max_output_tokens": 8192, + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 5e-08, + "source": "https://docs.zerogpu.ai/docs/model-catalog" + }, + "zerogpu/llama-3.1-8b-instruct-fast": { + "cache_read_input_token_cost": 2.5e-08, + "input_cost_per_token": 1.5e-07, + "litellm_provider": "zerogpu", + "max_input_tokens": 131072, + "max_output_tokens": 8192, + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 2.8e-07, + "source": "https://docs.zerogpu.ai/docs/model-catalog", + "supports_function_calling": true, + "supports_response_schema": true, + "supports_tool_choice": true + }, + "zerogpu/qwen3-30b-a3b-fp8": { + "cache_read_input_token_cost": 5e-08, + "input_cost_per_token": 1e-07, + "litellm_provider": "zerogpu", + "max_input_tokens": 32768, + "max_output_tokens": 16384, + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 4.5e-07, + "source": "https://docs.zerogpu.ai/docs/model-catalog", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true + }, "together_ai/Qwen/Qwen3.8-Flash": { "input_cost_per_token": 1.5e-07, "litellm_provider": "together_ai", diff --git a/provider_endpoints_support.json b/provider_endpoints_support.json index e62b74aa87d..721c626396f 100644 --- a/provider_endpoints_support.json +++ b/provider_endpoints_support.json @@ -2900,6 +2900,23 @@ "interactions": true } }, + "zerogpu": { + "display_name": "ZeroGPU (`zerogpu`)", + "url": "https://docs.litellm.ai/docs/providers/zerogpu", + "endpoints": { + "chat_completions": true, + "messages": true, + "responses": true, + "embeddings": false, + "image_generations": false, + "audio_transcriptions": false, + "audio_speech": false, + "moderations": false, + "batches": false, + "rerank": false, + "a2a": false + } + }, "ragflow": { "display_name": "RAGFlow (`ragflow`)", "url": "https://docs.litellm.ai/docs/providers/ragflow", diff --git a/tests/unit/llms/openai_like/test_zerogpu_provider.py b/tests/unit/llms/openai_like/test_zerogpu_provider.py new file mode 100644 index 00000000000..991796ae268 --- /dev/null +++ b/tests/unit/llms/openai_like/test_zerogpu_provider.py @@ -0,0 +1,241 @@ +import json +from collections.abc import Mapping +from typing import Final + +import httpx +import pytest +import respx +from openai import OpenAI + +import litellm +from litellm.litellm_core_utils.get_llm_provider_logic import get_llm_provider +from litellm.proxy.auth.model_checks import get_provider_models + +ZEROGPU_API_BASE: Final = "https://api.zerogpu.ai/v1" + +ZEROGPU_MODELS: Final = ( + "deepseek-v4.1-flash", + "glm-5.3-flash", + "gpt-4.1-mini", + "gpt-5.4-nano", + "gpt-5.6-luna", + "gpt-oss-120b", + "LFM2.5-1.2B-Instruct", + "llama-3.1-8b-instruct-fast", + "qwen3-30b-a3b-fp8", +) + + +WEATHER_TOOL: Final = { + "type": "function", + "function": { + "name": "get_weather", + "parameters": {"type": "object", "properties": {"city": {"type": "string"}}}, + }, +} + + +class _RecordingZeroGPU: + def __init__(self, usage: Mapping[str, object]) -> None: + self.usage: Final = usage + self.requests: tuple[httpx.Request, ...] = () + + def respond(self, request: httpx.Request) -> httpx.Response: + self.requests = (*self.requests, request) + return httpx.Response( + 200, + json={ + "id": "chatcmpl-zerogpu", + "object": "chat.completion", + "created": 1234567890, + "model": json.loads(request.content)["model"], + "choices": [{"index": 0, "message": {"role": "assistant", "content": "hi"}, "finish_reason": "stop"}], + "usage": dict(self.usage), + }, + ) + + def client(self) -> OpenAI: + return OpenAI( + api_key="zerogpu-test-key", + base_url=ZEROGPU_API_BASE, + http_client=httpx.Client(transport=httpx.MockTransport(self.respond)), + ) + + +def test_prefixed_model_reaches_zerogpu_chat_completions_without_the_prefix(): + zerogpu: Final = _RecordingZeroGPU({"prompt_tokens": 3, "completion_tokens": 1, "total_tokens": 4}) + + with zerogpu.client() as client: + litellm.completion( + model="zerogpu/gpt-oss-120b", + messages=[{"role": "user", "content": "hello"}], + client=client, + ) + + assert len(zerogpu.requests) == 1 + assert str(zerogpu.requests[0].url) == f"{ZEROGPU_API_BASE}/chat/completions" + assert json.loads(zerogpu.requests[0].content) == { + "model": "gpt-oss-120b", + "messages": [{"role": "user", "content": "hello"}], + } + + +def test_responses_request_reaches_zerogpu_responses_natively(): + with respx.mock() as upstream: + route: Final = upstream.post(f"{ZEROGPU_API_BASE}/responses").respond( + 200, + json={ + "id": "resp_zerogpu", + "object": "response", + "created_at": 1_792_000_000, + "model": "gpt-oss-120b", + "status": "completed", + "output": [ + { + "id": "msg_zerogpu", + "type": "message", + "role": "assistant", + "status": "completed", + "content": [{"type": "output_text", "text": "hi", "annotations": []}], + } + ], + "usage": {"input_tokens": 3, "output_tokens": 1, "total_tokens": 4}, + }, + ) + response: Final = litellm.responses(model="zerogpu/gpt-oss-120b", input="hello", api_key="zerogpu-test-key") + + body: Final = json.loads(route.calls.last.request.content) + assert body["model"] == "gpt-oss-120b" + assert body["input"] == "hello" + assert response.output[0].content[0].text == "hi" + + +def test_anthropic_messages_request_is_bridged_to_zerogpu_chat_completions(): + zerogpu: Final = _RecordingZeroGPU({"prompt_tokens": 3, "completion_tokens": 1, "total_tokens": 4}) + + with zerogpu.client() as client: + response: Final = litellm.anthropic.messages.create( + model="zerogpu/gpt-oss-120b", + messages=[{"role": "user", "content": "hello"}], + max_tokens=10, + client=client, + ) + + assert [str(request.url) for request in zerogpu.requests] == [f"{ZEROGPU_API_BASE}/chat/completions"] + assert json.loads(zerogpu.requests[0].content) == { + "model": "gpt-oss-120b", + "messages": [{"role": "user", "content": "hello"}], + "max_tokens": 10, + } + assert response["content"] == [{"type": "text", "text": "hi"}] + + +@pytest.mark.parametrize("model", ZEROGPU_MODELS) +def test_zerogpu_completion_is_charged_at_the_cost_map_rates(model: str): + usage: Final = { + "prompt_tokens": 1000, + "completion_tokens": 200, + "total_tokens": 1200, + "prompt_tokens_details": {"cached_tokens": 400}, + } + zerogpu: Final = _RecordingZeroGPU(usage) + + with zerogpu.client() as client: + response: Final = litellm.completion( + model=f"zerogpu/{model}", + messages=[{"role": "user", "content": "hello"}], + client=client, + ) + + rates: Final = litellm.model_cost[f"zerogpu/{model}"] + cached_rate: Final = rates.get("cache_read_input_token_cost", rates["input_cost_per_token"]) + expected_cost: Final = ( + 600 * rates["input_cost_per_token"] + 400 * cached_rate + 200 * rates["output_cost_per_token"] + ) + assert response._hidden_params["response_cost"] == pytest.approx(expected_cost) + assert expected_cost > 0 + + +@pytest.mark.parametrize("model", ZEROGPU_MODELS) +def test_trim_messages_fits_zerogpu_history_into_the_model_context_window(model: str): + budget: Final = 20 + trim_ratio: Final = budget / litellm.model_cost[f"zerogpu/{model}"]["max_input_tokens"] + messages: Final = [{"role": "user", "content": "lorem ipsum " * 300}] + + trimmed: Final = litellm.utils.trim_messages(messages, model=f"zerogpu/{model}", trim_ratio=trim_ratio) + + trimmed_tokens: Final = litellm.token_counter(model=f"zerogpu/{model}", messages=trimmed) + assert trimmed_tokens <= budget, trimmed + + +@pytest.mark.parametrize( + ("model", "forwards_tools", "forwards_reasoning_effort"), + [("gpt-5.6-luna", True, True), ("gpt-oss-120b", True, True), ("LFM2.5-1.2B-Instruct", False, False)], +) +def test_tools_and_reasoning_effort_reach_zerogpu_only_for_models_that_support_them( + model: str, forwards_tools: bool, forwards_reasoning_effort: bool +): + zerogpu: Final = _RecordingZeroGPU({"prompt_tokens": 3, "completion_tokens": 1, "total_tokens": 4}) + + with zerogpu.client() as client: + litellm.completion( + model=f"zerogpu/{model}", + messages=[{"role": "user", "content": "weather in Paris?"}], + tools=[WEATHER_TOOL], + tool_choice="auto", + reasoning_effort="high", + drop_params=True, + client=client, + ) + + tools: Final = {"tools": [WEATHER_TOOL], "tool_choice": "auto"} if forwards_tools else {} + reasoning: Final = {"reasoning_effort": "high"} if forwards_reasoning_effort else {} + assert json.loads(zerogpu.requests[0].content) == { + "model": model, + "messages": [{"role": "user", "content": "weather in Paris?"}], + **tools, + **reasoning, + } + + +@pytest.mark.parametrize("model", ZEROGPU_MODELS) +def test_model_info_reports_the_reasoning_support_litellm_routes_with(model: str): + model_info: Final = litellm.get_model_info(f"zerogpu/{model}") + + assert bool(model_info.get("supports_reasoning")) == litellm.utils.supports_reasoning(f"zerogpu/{model}") + + +def test_zerogpu_key_and_base_come_from_zerogpu_env_vars(monkeypatch: pytest.MonkeyPatch): + monkeypatch.setenv("ZEROGPU_API_KEY", "zerogpu-env-key") + monkeypatch.delenv("ZEROGPU_API_BASE", raising=False) + + assert get_llm_provider(model="zerogpu/gpt-oss-120b") == ( + "gpt-oss-120b", + "zerogpu", + "zerogpu-env-key", + ZEROGPU_API_BASE, + ) + + monkeypatch.setenv("ZEROGPU_API_BASE", "https://zerogpu.internal.example/v1") + + assert get_llm_provider(model="zerogpu/gpt-oss-120b")[3] == "https://zerogpu.internal.example/v1" + + +def test_zerogpu_api_base_alone_selects_the_zerogpu_provider(monkeypatch: pytest.MonkeyPatch): + monkeypatch.setenv("ZEROGPU_API_KEY", "zerogpu-env-key") + + assert get_llm_provider(model="gpt-oss-120b", api_base=ZEROGPU_API_BASE) == ( + "gpt-oss-120b", + "zerogpu", + "zerogpu-env-key", + ZEROGPU_API_BASE, + ) + + +def test_zerogpu_wildcard_lists_every_zerogpu_catalog_model(): + catalog_models: Final = sorted( + key for key, row in litellm.model_cost.items() if row.get("litellm_provider") == "zerogpu" + ) + + assert catalog_models + assert sorted(get_provider_models("zerogpu") or []) == catalog_models diff --git a/tests/unit/proxy/public_endpoints/test_public_endpoints.py b/tests/unit/proxy/public_endpoints/test_public_endpoints.py index cc61b6f2fcf..f90fe60173b 100644 --- a/tests/unit/proxy/public_endpoints/test_public_endpoints.py +++ b/tests/unit/proxy/public_endpoints/test_public_endpoints.py @@ -363,6 +363,26 @@ def test_cognition_provider_fields(): assert fields_by_key["api_base"]["required"] is False +def test_zerogpu_add_model_form_defaults_resolve_to_a_priced_zerogpu_model(): + import litellm + from litellm.litellm_core_utils.get_llm_provider_logic import get_llm_provider + + app_instance: Final = FastAPI() + app_instance.include_router(router) + providers: Final = TestClient(app_instance).get("/public/providers/fields").json() + + zerogpu: Final = next(p for p in providers if p["litellm_provider"] == "zerogpu") + fields_by_key: Final = {f["key"]: f for f in zerogpu["credential_fields"]} + placeholder: Final = zerogpu["default_model_placeholder"] + _, provider, _, api_base = get_llm_provider(model=placeholder) + + assert provider == "zerogpu" + assert api_base == fields_by_key["api_base"]["placeholder"] + assert litellm.model_cost[placeholder]["litellm_provider"] == "zerogpu" + assert fields_by_key["api_key"]["required"] is True + assert fields_by_key["api_key"]["field_type"] == "password" + + def test_qwen_mainland_provider_fields_carry_the_qianwen_brand(): app_instance = FastAPI() app_instance.include_router(router) diff --git a/tests/unit/test_model_prices_and_context_window.py b/tests/unit/test_model_prices_and_context_window.py index 1f84be7710f..a01400646c4 100644 --- a/tests/unit/test_model_prices_and_context_window.py +++ b/tests/unit/test_model_prices_and_context_window.py @@ -115,6 +115,27 @@ def test_github_copilot_backup_rows_match_the_main_catalog() -> None: assert _github_copilot_catalog_rows(PRICES_PATH) == _github_copilot_catalog_rows(BACKUP_PRICES_PATH) +class _TokenLimits(BaseModel): + model_config = ConfigDict(frozen=True, extra="ignore") + + max_input_tokens: int + max_output_tokens: int + max_tokens: int + + +@pytest.mark.parametrize("path", (PRICES_PATH, BACKUP_PRICES_PATH), ids=("main", "backup")) +def test_zerogpu_rows_cap_max_tokens_at_an_output_limit_that_fits_the_context_window(path: Path) -> None: + prices: Final[dict[str, JsonValue]] = TypeAdapter(dict[str, JsonValue]).validate_json(path.read_bytes()) + rows: Final = MappingProxyType( + {key: _TokenLimits.model_validate(value) for key, value in prices.items() if key.startswith("zerogpu/")} + ) + assert rows, f"No ZeroGPU rows found in {path}" + assert {key: row.max_tokens for key, row in rows.items()} == { + key: row.max_output_tokens for key, row in rows.items() + } + assert tuple(key for key, row in rows.items() if row.max_output_tokens > row.max_input_tokens) == () + + def test_github_copilot_rows_resolve_through_get_model_info() -> None: rows: Final[Mapping[str, _CopilotEndpointRow]] = _copilot_endpoint_rows(PRICES_PATH) assert { diff --git a/ui/litellm-dashboard/src/components/provider_info_helpers.test.tsx b/ui/litellm-dashboard/src/components/provider_info_helpers.test.tsx index 0ebd2cdff17..83e80911cff 100644 --- a/ui/litellm-dashboard/src/components/provider_info_helpers.test.tsx +++ b/ui/litellm-dashboard/src/components/provider_info_helpers.test.tsx @@ -224,6 +224,7 @@ describe("provider_info_helpers", () => { Providers.Sail, Providers.WANDB, Providers.ZAI, + Providers.ZeroGPU, ]; const logolessProviders = Object.values(Providers).filter((provider) => !providerLogoMap[provider]); expect([...logolessProviders].sort()).toEqual([...knownLogolessProviders].sort()); @@ -357,6 +358,10 @@ describe("provider_info_helpers", () => { expect(getPlaceholder(Providers.Cognition)).toBe("cognition/swe-1.7"); }); + it("should return zerogpu/gpt-oss-120b placeholder for ZeroGPU provider", () => { + expect(getPlaceholder("ZeroGPU")).toBe("zerogpu/gpt-oss-120b"); + }); + it("should return a chatgpt/ placeholder for the CHATGPT dropdown key", () => { expect(getPlaceholder("CHATGPT")).toBe("chatgpt/gpt-5.4"); }); diff --git a/ui/litellm-dashboard/src/components/provider_info_helpers.tsx b/ui/litellm-dashboard/src/components/provider_info_helpers.tsx index 2541aa45050..bfb578b218a 100644 --- a/ui/litellm-dashboard/src/components/provider_info_helpers.tsx +++ b/ui/litellm-dashboard/src/components/provider_info_helpers.tsx @@ -197,6 +197,7 @@ export enum Providers { xAI = "xAI", XINFERENCE = "Xinference", ZAI = "Z.AI (Zhipu AI)", + ZeroGPU = "ZeroGPU", } export const provider_map: Record = { @@ -325,6 +326,7 @@ export const provider_map: Record = { xAI: "xai", XINFERENCE: "xinference", ZAI: "zai", + ZeroGPU: "zerogpu", }; const standaloneSubproviderSlugs = new Set(["bedrock_mantle"]); @@ -497,6 +499,7 @@ const providerPlaceholderMap: Partial> = { [Providers.Voyage]: "voyage/", [Providers.WATSONX]: "watsonx/ibm/granite-3-3-8b-instruct", [Providers.ZAI]: "zai/glm-4.5", + [Providers.ZeroGPU]: "zerogpu/gpt-oss-120b", }; export const getPlaceholder = (selectedProvider: string): string => {