diff --git a/litellm/constants.py b/litellm/constants.py index fd9812e2cf5..10c943656f7 100644 --- a/litellm/constants.py +++ b/litellm/constants.py @@ -949,6 +949,7 @@ openai_compatible_endpoints: Final[list] = [ "https://api.sailresearch.com/v1", "https://api.cognition.ai/v1", "https://api.scx.ai/v1", + "https://api.prisminference.com/v1", "https://gigachat.devices.sberbank.ru/api/v1", ] @@ -1022,6 +1023,7 @@ openai_compatible_providers: Final[list] = [ "meta", # Meta Model API (Muse Spark) - JSON-configured provider "cognition", "scx-ai", + "prism", "sail", ] diff --git a/litellm/llms/openai_like/providers.json b/litellm/llms/openai_like/providers.json index ae09b48bd1e..440e5490d71 100644 --- a/litellm/llms/openai_like/providers.json +++ b/litellm/llms/openai_like/providers.json @@ -201,6 +201,12 @@ }, "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"] + }, "sail": { "base_url": "https://api.sailresearch.com/v1", "api_key_env": "SAIL_API_KEY", diff --git a/litellm/model_prices_and_context_window_backup.json b/litellm/model_prices_and_context_window_backup.json index 0c694b6bcf6..d5490f5df9e 100644 --- a/litellm/model_prices_and_context_window_backup.json +++ b/litellm/model_prices_and_context_window_backup.json @@ -78485,5 +78485,55 @@ "thinking_always_on": true, "prompt_cache_min_tokens": 512, "source": "https://cloud.google.com/gemini-enterprise-agent-platform/generative-ai/pricing" + }, + "prism/deepseek-v4.1-flash": { + "cache_read_input_token_cost": 7e-08, + "input_cost_per_token": 1.7e-07, + "litellm_provider": "prism", + "max_input_tokens": 1000000, + "max_output_tokens": 384000, + "max_tokens": 384000, + "mode": "chat", + "output_cost_per_token": 6.3e-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, + "supports_vision": true + }, + "prism/deepseek-v4-flash": { + "cache_read_input_token_cost": 7e-08, + "input_cost_per_token": 1.7e-07, + "litellm_provider": "prism", + "max_input_tokens": 1000000, + "max_output_tokens": 384000, + "max_tokens": 384000, + "mode": "chat", + "output_cost_per_token": 2.1e-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, + "supports_vision": false } } diff --git a/litellm/provider_endpoints_support_backup.json b/litellm/provider_endpoints_support_backup.json index 1fcb7600a5e..ad6e5857218 100644 --- a/litellm/provider_endpoints_support_backup.json +++ b/litellm/provider_endpoints_support_backup.json @@ -1943,6 +1943,23 @@ "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 8bd7ed81583..67a8c356a4a 100644 --- a/litellm/proxy/public_endpoints/provider_create_fields.json +++ b/litellm/proxy/public_endpoints/provider_create_fields.json @@ -2855,6 +2855,34 @@ ], "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 8862df9dc22..b9be8b6858e 100644 --- a/litellm/types/utils.py +++ b/litellm/types/utils.py @@ -4139,6 +4139,7 @@ class LlmProviders(str, Enum): PINSTRIPES = "pinstripes" COGNITION = "cognition" SCX_AI = "scx-ai" + PRISM = "prism" DARKBLOOM = "darkbloom" META = "meta" SAIL = "sail" diff --git a/model_prices_and_context_window.json b/model_prices_and_context_window.json index 0c694b6bcf6..d5490f5df9e 100644 --- a/model_prices_and_context_window.json +++ b/model_prices_and_context_window.json @@ -78485,5 +78485,55 @@ "thinking_always_on": true, "prompt_cache_min_tokens": 512, "source": "https://cloud.google.com/gemini-enterprise-agent-platform/generative-ai/pricing" + }, + "prism/deepseek-v4.1-flash": { + "cache_read_input_token_cost": 7e-08, + "input_cost_per_token": 1.7e-07, + "litellm_provider": "prism", + "max_input_tokens": 1000000, + "max_output_tokens": 384000, + "max_tokens": 384000, + "mode": "chat", + "output_cost_per_token": 6.3e-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, + "supports_vision": true + }, + "prism/deepseek-v4-flash": { + "cache_read_input_token_cost": 7e-08, + "input_cost_per_token": 1.7e-07, + "litellm_provider": "prism", + "max_input_tokens": 1000000, + "max_output_tokens": 384000, + "max_tokens": 384000, + "mode": "chat", + "output_cost_per_token": 2.1e-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, + "supports_vision": false } } diff --git a/provider_endpoints_support.json b/provider_endpoints_support.json index 790a050a878..44ef9363b64 100644 --- a/provider_endpoints_support.json +++ b/provider_endpoints_support.json @@ -2195,6 +2195,23 @@ "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/unit/llms/openai_like/test_prism_provider.py b/tests/unit/llms/openai_like/test_prism_provider.py new file mode 100644 index 00000000000..c1775c63dc9 --- /dev/null +++ b/tests/unit/llms/openai_like/test_prism_provider.py @@ -0,0 +1,192 @@ +import json +from pathlib import Path +from typing import Final + +import pytest +import respx + +import litellm +from litellm.caching.llm_caching_handler import LLMClientCache + + +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" + + +PRISM_MODELS = tuple(sorted(name for name in litellm.model_cost if name.startswith("prism/"))) + + +@pytest.mark.parametrize("model", PRISM_MODELS) +def test_prism_model_cost_and_capabilities(model: str): + 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(model_info["input_cost_per_token"] * 1_000_000) + assert completion_cost == pytest.approx(model_info["output_cost_per_token"] * 1_000_000) + assert 0 < model_info["cache_read_input_token_cost"] < model_info["input_cost_per_token"] + assert model_info["output_cost_per_token"] > 0 + assert model_info["max_tokens"] == model_info["max_output_tokens"] <= model_info["max_input_tokens"] + assert model_info["litellm_provider"] == "prism" + assert model_info["mode"] == "chat" + 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 + assert litellm.supports_vision(model) is model_info["supports_vision"] + + +def test_prism_backup_registry_mirrors_cost_map(): + package_root = Path(litellm.__file__).parent + cost_map = json.loads((package_root.parent / "model_prices_and_context_window.json").read_text()) + backup = json.loads((package_root / "model_prices_and_context_window_backup.json").read_text()) + prism_entries = {name: entry for name, entry in cost_map.items() if name.startswith("prism/")} + + assert tuple(sorted(prism_entries)) == PRISM_MODELS + assert prism_entries + assert all("supports_vision" in entry for entry in prism_entries.values()) + assert prism_entries == {name: backup[name] for name in prism_entries} + + +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(): + with respx.mock() as upstream: + route: Final = upstream.post("https://api.prisminference.com/v1/responses").respond( + 200, + json={ + "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: Final = litellm.responses( + model="prism/deepseek-v4-flash", + input="Say hello", + api_key="prism-test-key", + ) + + request: Final = route.calls.last.request + body: Final = json.loads(request.content) + assert route.call_count == 1 + 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(monkeypatch: pytest.MonkeyPatch): + monkeypatch.setattr(litellm, "disable_aiohttp_transport", True) + monkeypatch.setattr(litellm, "in_memory_llm_clients_cache", LLMClientCache()) + with respx.mock() as upstream: + route: Final = upstream.post("https://api.prisminference.com/v1/messages").respond( + 200, + json={ + "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}, + }, + ) + response: Final = await litellm.anthropic.messages.acreate( + model="prism/deepseek-v4-flash", + messages=[{"role": "user", "content": "Say hello"}], + max_tokens=32, + api_key="prism-test-key", + ) + + request: Final = route.calls.last.request + body: Final = json.loads(request.content) + assert route.call_count == 1 + 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"