diff --git a/litellm/constants.py b/litellm/constants.py index 5f8fa203b37..6ddfe673202 100644 --- a/litellm/constants.py +++ b/litellm/constants.py @@ -897,6 +897,7 @@ openai_compatible_endpoints: Final[list] = [ "https://api.meta.ai/v1", "https://api.cognition.ai/v1", "https://api.scx.ai/v1", + "https://api.prisminference.com/v1", "https://gigachat.devices.sberbank.ru/api/v1", ] @@ -969,6 +970,7 @@ openai_compatible_providers: Final[list] = [ "meta", # Meta Model API (Muse Spark) - JSON-configured provider "cognition", "scx-ai", + "prism", ] openai_text_completion_compatible_providers: Final[list] = [ # providers that support `/v1/completions` "together_ai", diff --git a/litellm/llms/openai_like/providers.json b/litellm/llms/openai_like/providers.json index fe10293c420..70a690c72df 100644 --- a/litellm/llms/openai_like/providers.json +++ b/litellm/llms/openai_like/providers.json @@ -200,5 +200,11 @@ "temperature_max": 1.99 }, "supported_endpoints": ["/v1/chat/completions"] + }, + "prism": { + "base_url": "https://api.prisminference.com/v1", + "api_key_env": "PRISM_API_KEY", + "api_base_env": "PRISM_API_BASE", + "supported_endpoints": ["/v1/chat/completions", "/v1/responses", "/v1/messages"] } } diff --git a/litellm/model_prices_and_context_window_backup.json b/litellm/model_prices_and_context_window_backup.json index 2220d0e1fe5..ac14a018ad8 100644 --- a/litellm/model_prices_and_context_window_backup.json +++ b/litellm/model_prices_and_context_window_backup.json @@ -65270,5 +65270,53 @@ "supports_tool_choice": false, "supports_response_schema": true, "supports_vision": false + }, + "prism/deepseek-v4.1-flash": { + "cache_read_input_token_cost": 7e-08, + "input_cost_per_token": 3e-07, + "litellm_provider": "prism", + "max_input_tokens": 1000000, + "max_output_tokens": 384000, + "max_tokens": 384000, + "mode": "chat", + "output_cost_per_token": 1.2e-06, + "source": "https://prisminference.com/pricing", + "supported_endpoints": [ + "/v1/chat/completions", + "/v1/responses", + "/v1/messages" + ], + "supports_function_calling": true, + "supports_native_streaming": true, + "supports_parallel_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_system_messages": true, + "supports_tool_choice": true + }, + "prism/deepseek-v4-flash": { + "cache_read_input_token_cost": 7e-08, + "input_cost_per_token": 1.4e-07, + "litellm_provider": "prism", + "max_input_tokens": 1000000, + "max_output_tokens": 393216, + "max_tokens": 393216, + "mode": "chat", + "output_cost_per_token": 2.8e-07, + "source": "https://prisminference.com/pricing", + "supported_endpoints": [ + "/v1/chat/completions", + "/v1/responses", + "/v1/messages" + ], + "supports_function_calling": true, + "supports_native_streaming": true, + "supports_parallel_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_system_messages": true, + "supports_tool_choice": true } } diff --git a/litellm/provider_endpoints_support_backup.json b/litellm/provider_endpoints_support_backup.json index dbeaccdda2d..35941d357d8 100644 --- a/litellm/provider_endpoints_support_backup.json +++ b/litellm/provider_endpoints_support_backup.json @@ -1925,6 +1925,23 @@ "interactions": true } }, + "prism": { + "display_name": "Prism (`prism`)", + "url": "https://docs.litellm.ai/docs/providers/prism", + "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 + } + }, "recraft": { "display_name": "Recraft (`recraft`)", "url": "https://docs.litellm.ai/docs/providers/recraft", diff --git a/litellm/proxy/public_endpoints/provider_create_fields.json b/litellm/proxy/public_endpoints/provider_create_fields.json index cd781abee26..fd7fe759926 100644 --- a/litellm/proxy/public_endpoints/provider_create_fields.json +++ b/litellm/proxy/public_endpoints/provider_create_fields.json @@ -2703,6 +2703,34 @@ ], "default_model_placeholder": "gpt-3.5-turbo" }, + { + "provider": "PRISM", + "provider_display_name": "Prism", + "litellm_provider": "prism", + "credential_fields": [ + { + "key": "api_base", + "label": "API Base", + "placeholder": "https://api.prisminference.com/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": "prism/deepseek-v4.1-flash" + }, { "provider": "RECRAFT", "provider_display_name": "Recraft", diff --git a/litellm/types/utils.py b/litellm/types/utils.py index 00c55b35182..379ea918e3e 100644 --- a/litellm/types/utils.py +++ b/litellm/types/utils.py @@ -4046,6 +4046,7 @@ class LlmProviders(str, Enum): PINSTRIPES = "pinstripes" COGNITION = "cognition" SCX_AI = "scx-ai" + PRISM = "prism" DARKBLOOM = "darkbloom" META = "meta" LITELLM_AGENT = "litellm_agent" diff --git a/model_prices_and_context_window.json b/model_prices_and_context_window.json index 2220d0e1fe5..ac14a018ad8 100644 --- a/model_prices_and_context_window.json +++ b/model_prices_and_context_window.json @@ -65270,5 +65270,53 @@ "supports_tool_choice": false, "supports_response_schema": true, "supports_vision": false + }, + "prism/deepseek-v4.1-flash": { + "cache_read_input_token_cost": 7e-08, + "input_cost_per_token": 3e-07, + "litellm_provider": "prism", + "max_input_tokens": 1000000, + "max_output_tokens": 384000, + "max_tokens": 384000, + "mode": "chat", + "output_cost_per_token": 1.2e-06, + "source": "https://prisminference.com/pricing", + "supported_endpoints": [ + "/v1/chat/completions", + "/v1/responses", + "/v1/messages" + ], + "supports_function_calling": true, + "supports_native_streaming": true, + "supports_parallel_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_system_messages": true, + "supports_tool_choice": true + }, + "prism/deepseek-v4-flash": { + "cache_read_input_token_cost": 7e-08, + "input_cost_per_token": 1.4e-07, + "litellm_provider": "prism", + "max_input_tokens": 1000000, + "max_output_tokens": 393216, + "max_tokens": 393216, + "mode": "chat", + "output_cost_per_token": 2.8e-07, + "source": "https://prisminference.com/pricing", + "supported_endpoints": [ + "/v1/chat/completions", + "/v1/responses", + "/v1/messages" + ], + "supports_function_calling": true, + "supports_native_streaming": true, + "supports_parallel_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_system_messages": true, + "supports_tool_choice": true } } diff --git a/provider_endpoints_support.json b/provider_endpoints_support.json index c71f4a82a4a..f1cec19edbd 100644 --- a/provider_endpoints_support.json +++ b/provider_endpoints_support.json @@ -2142,6 +2142,23 @@ "interactions": true } }, + "prism": { + "display_name": "Prism (`prism`)", + "url": "https://docs.litellm.ai/docs/providers/prism", + "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 + } + }, "recraft": { "display_name": "Recraft (`recraft`)", "url": "https://docs.litellm.ai/docs/providers/recraft", diff --git a/tests/test_litellm/llms/openai_like/test_prism_provider.py b/tests/test_litellm/llms/openai_like/test_prism_provider.py new file mode 100644 index 00000000000..f6b56612fd1 --- /dev/null +++ b/tests/test_litellm/llms/openai_like/test_prism_provider.py @@ -0,0 +1,207 @@ +import json +from pathlib import Path + +import httpx +import pytest + +import litellm +from litellm.llms.custom_httpx.http_handler import AsyncHTTPHandler, HTTPHandler + + +def _prism_client(requests: list[httpx.Request], response_body: dict[str, object]) -> HTTPHandler: + def respond(request: httpx.Request) -> httpx.Response: + requests.append(request) + return httpx.Response(200, json=response_body) + + return HTTPHandler(client=httpx.Client(transport=httpx.MockTransport(respond))) + + +def test_prism_provider_resolution(monkeypatch: pytest.MonkeyPatch): + from litellm.litellm_core_utils.get_llm_provider_logic import get_llm_provider + + monkeypatch.setenv("PRISM_API_KEY", "prism-test-key") + + model, provider, api_key, api_base = get_llm_provider( + model="prism/deepseek-v4-flash", + custom_llm_provider=None, + api_base=None, + api_key=None, + ) + + assert model == "deepseek-v4-flash" + assert provider == "prism" + assert api_key == "prism-test-key" + assert api_base == "https://api.prisminference.com/v1" + + +def test_prism_provider_keeps_explicit_credentials(monkeypatch: pytest.MonkeyPatch): + from litellm.litellm_core_utils.get_llm_provider_logic import get_llm_provider + + monkeypatch.setenv("PRISM_API_KEY", "prism-env-key") + + _, provider, api_key, api_base = get_llm_provider( + model="prism/deepseek-v4-flash", + custom_llm_provider=None, + api_base="https://prism.internal.example/v1", + api_key="prism-explicit-key", + ) + + assert provider == "prism" + assert api_key == "prism-explicit-key" + assert api_base == "https://prism.internal.example/v1" + + +@pytest.mark.parametrize( + ("model", "input_cost", "output_cost", "max_output_tokens"), + [ + ("prism/deepseek-v4.1-flash", 0.30, 1.20, 384_000), + ("prism/deepseek-v4-flash", 0.14, 0.28, 393_216), + ], +) +def test_prism_model_cost_and_capabilities( + model: str, + input_cost: float, + output_cost: float, + max_output_tokens: int, +): + from litellm.cost_calculator import cost_per_token + + prompt_cost, completion_cost = cost_per_token( + model=model, + prompt_tokens=1_000_000, + completion_tokens=1_000_000, + custom_llm_provider="prism", + ) + model_info = litellm.get_model_info(model) + + assert prompt_cost == pytest.approx(input_cost) + assert completion_cost == pytest.approx(output_cost) + assert model_info["cache_read_input_token_cost"] == pytest.approx(7e-08) + assert model_info["max_input_tokens"] == 1_000_000 + assert model_info["max_output_tokens"] == max_output_tokens + assert model_info["supports_function_calling"] is True + assert model_info["supports_native_streaming"] is True + assert model_info["supports_reasoning"] is True + assert model_info["supports_response_schema"] is True + + +def test_prism_is_available_in_add_model_form(): + fields_path = Path(litellm.__file__).parent / "proxy" / "public_endpoints" / "provider_create_fields.json" + providers = json.loads(fields_path.read_text()) + prism = next(provider for provider in providers if provider["litellm_provider"] == "prism") + + assert prism["provider"] == "PRISM" + assert prism["provider_display_name"] == "Prism" + assert prism["default_model_placeholder"] == "prism/deepseek-v4.1-flash" + assert {field["key"]: field["required"] for field in prism["credential_fields"]} == { + "api_base": False, + "api_key": True, + } + + +def test_prism_supported_endpoints(): + matrix_path = Path(litellm.__file__).parent / "provider_endpoints_support_backup.json" + providers = json.loads(matrix_path.read_text())["providers"] + + assert providers["prism"]["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, + } + + +def test_prism_responses_request(): + requests: list[httpx.Request] = [] + client = _prism_client( + requests, + { + "id": "resp_prism", + "object": "response", + "created_at": 1_789_550_000, + "model": "deepseek-v4-flash", + "status": "completed", + "output": [ + { + "id": "msg_prism", + "type": "message", + "role": "assistant", + "status": "completed", + "content": [ + { + "type": "output_text", + "text": "Hello from Prism", + "annotations": [], + } + ], + } + ], + "usage": { + "input_tokens": 4, + "output_tokens": 3, + "total_tokens": 7, + }, + }, + ) + + response = litellm.responses( + model="prism/deepseek-v4-flash", + input="Say hello", + api_key="prism-test-key", + client=client, + ) + + request = requests[0] + body = json.loads(request.content) + assert str(request.url) == "https://api.prisminference.com/v1/responses" + assert request.headers["authorization"] == "Bearer prism-test-key" + assert body["model"] == "deepseek-v4-flash" + assert body["input"] == "Say hello" + assert response.output[0].content[0].text == "Hello from Prism" + + +@pytest.mark.asyncio +async def test_prism_anthropic_messages_request(): + requests: list[httpx.Request] = [] + response_body: dict[str, object] = { + "id": "msg_prism", + "type": "message", + "role": "assistant", + "model": "deepseek-v4-flash", + "content": [{"type": "text", "text": "Hello from Prism"}], + "stop_reason": "end_turn", + "stop_sequence": None, + "usage": {"input_tokens": 4, "output_tokens": 3}, + } + + def respond(request: httpx.Request) -> httpx.Response: + requests.append(request) + return httpx.Response(200, json=response_body) + + client = AsyncHTTPHandler() + await client.close() + client.client = httpx.AsyncClient(transport=httpx.MockTransport(respond)) + + response = await litellm.anthropic.messages.acreate( + model="prism/deepseek-v4-flash", + messages=[{"role": "user", "content": "Say hello"}], + max_tokens=32, + api_key="prism-test-key", + client=client, + ) + + request = requests[0] + body = json.loads(request.content) + assert str(request.url) == "https://api.prisminference.com/v1/messages" + assert request.headers["authorization"] == "Bearer prism-test-key" + assert request.headers["anthropic-version"] == "2023-06-01" + assert body["model"] == "deepseek-v4-flash" + assert body["messages"] == [{"role": "user", "content": "Say hello"}] + assert response["content"][0]["text"] == "Hello from Prism" diff --git a/tests/test_litellm/llms/vertex_ai/gemini_embeddings/test_batch_embed_content_transformation.py b/tests/test_litellm/llms/vertex_ai/gemini_embeddings/test_batch_embed_content_transformation.py index 86b3f0976ab..8c58ebbff7b 100644 --- a/tests/test_litellm/llms/vertex_ai/gemini_embeddings/test_batch_embed_content_transformation.py +++ b/tests/test_litellm/llms/vertex_ai/gemini_embeddings/test_batch_embed_content_transformation.py @@ -299,10 +299,17 @@ class TestProcessResponse: ) +@pytest.mark.usefixtures("local_model_cost_map") class TestProcessEmbedContentResponseUsage: """Gemini Embedding 2 embedContent usageMetadata must drive spend. Regression for multimodal calls recording prompt_tokens=0 / spend=$0. + + The spend assertions read the ``gemini-embedding-2`` rates, so they must be + pinned to the bundled in-repo cost map: the default network-fetched ``main`` + copy is a different branch and already re-prices this model per modality + token (no ``input_cost_per_image`` / ``input_cost_per_*_per_second``), which + makes these cases fail for reasons unrelated to the transformation. """ MODEL = "gemini-embedding-2"