Add DeepSeek V4 Flash 0731 to OpenRouter

inference_gateway/providers/openrouter.py

@@ -44,6 +44,9 @@ def __init__(self, **data):
 
 if config.USE_OPENROUTER:
     WHITELISTED_OPENROUTER_INFERENCE_MODELS = [
+        WhitelistedOpenRouterModel(
+            name="deepseek-ai/DeepSeek-V4-Flash-0731", openrouter_name="deepseek/deepseek-v4-flash-0731"
+        ),
         WhitelistedOpenRouterModel(name="deepseek-ai/DeepSeek-R1-0528", openrouter_name="deepseek/deepseek-r1-0528"),
         WhitelistedOpenRouterModel(name="zai-org/GLM-4.6", openrouter_name="z-ai/glm-4.6"),
         WhitelistedOpenRouterModel(name="zai-org/GLM-4.6-FP8", openrouter_name="z-ai/glm-4.6"),

tests/inference_gateway/test_openrouter_models.py

@@ -0,0 +1,108 @@
+from __future__ import annotations
+
+import asyncio
+import importlib
+import sys
+from types import ModuleType
+from typing import Self
+
+import pytest
+
+
+@pytest.fixture
+def openrouter_module(monkeypatch: pytest.MonkeyPatch) -> ModuleType:
+    environment = {
+        "HOST": "127.0.0.1",
+        "PORT": "8000",
+        "USE_DATABASE": "false",
+        "MAX_COST_PER_EVALUATION_RUN_USD": "1",
+        "USE_CHUTES": "false",
+        "USE_TARGON": "false",
+        "USE_OPENROUTER": "true",
+        "OPENROUTER_BASE_URL": "https://openrouter.example/v1",
+        "OPENROUTER_API_KEY": "test-key",
+        "OPENROUTER_WEIGHT": "1",
+        "TEST_INFERENCE_MODELS": "false",
+        "TEST_EMBEDDING_MODELS": "false",
+    }
+    for name, value in environment.items():
+        monkeypatch.setenv(name, value)
+
+    sys.modules.pop("inference_gateway.providers.openrouter", None)
+    sys.modules.pop("inference_gateway.config", None)
+    return importlib.import_module("inference_gateway.providers.openrouter")
+
+
+def test_deepseek_v4_flash_uses_exact_openrouter_slug(openrouter_module: ModuleType) -> None:
+    model = next(
+        model
+        for model in openrouter_module.WHITELISTED_OPENROUTER_INFERENCE_MODELS
+        if model.name == "deepseek-ai/DeepSeek-V4-Flash-0731"
+    )
+
+    assert model.openrouter_name == "deepseek/deepseek-v4-flash-0731"
+
+
+def test_deepseek_v4_flash_metadata_comes_from_openrouter_catalog(
+    monkeypatch: pytest.MonkeyPatch, openrouter_module: ModuleType
+) -> None:
+    inference_catalog = []
+    for model in openrouter_module.WHITELISTED_OPENROUTER_INFERENCE_MODELS:
+        is_deepseek_v4 = model.name == "deepseek-ai/DeepSeek-V4-Flash-0731"
+        inference_catalog.append(
+            {
+                "id": model.openrouter_name,
+                "architecture": {"input_modalities": ["text"], "output_modalities": ["text"]},
+                "context_length": 1_048_576 if is_deepseek_v4 else 128_000,
+                "pricing": {
+                    "prompt": "0.00000014" if is_deepseek_v4 else "0.000001",
+                    "completion": "0.00000028" if is_deepseek_v4 else "0.000002",
+                },
+            }
+        )
+    embedding_catalog = [
+        {
+            "id": model.openrouter_name,
+            "architecture": {"input_modalities": ["text"], "output_modalities": ["embeddings"]},
+            "context_length": 32_768,
+            "pricing": {"prompt": "0.00000001"},
+        }
+        for model in openrouter_module.WHITELISTED_OPENROUTER_EMBEDDING_MODELS
+    ]
+
+    class FakeResponse:
+        def __init__(self, data: list[dict]) -> None:
+            self.data = data
+
+        def raise_for_status(self) -> None:
+            return None
+
+        def json(self) -> dict[str, list[dict]]:
+            return {"data": self.data}
+
+    class FakeAsyncClient:
+        async def __aenter__(self) -> Self:
+            return self
+
+        async def __aexit__(self, *args: object) -> None:
+            return None
+
+        async def get(self, url: str) -> FakeResponse:
+            catalog = embedding_catalog if "/embeddings/" in url else inference_catalog
+            return FakeResponse(catalog)
+
+    class FakeOpenAI:
+        def __init__(self, **kwargs: object) -> None:
+            self.kwargs = kwargs
+
+    monkeypatch.setattr(openrouter_module.httpx, "AsyncClient", FakeAsyncClient)
+    monkeypatch.setattr(openrouter_module, "AsyncOpenAI", FakeOpenAI)
+
+    provider = asyncio.run(openrouter_module.OpenRouterProvider().init())
+    model_info = provider.get_inference_model_info_by_name("deepseek-ai/DeepSeek-V4-Flash-0731")
+
+    assert model_info is not None
+    assert model_info.external_name == "deepseek/deepseek-v4-flash-0731"
+    assert model_info.max_input_tokens == 1_048_576
+    assert model_info.cost_usd_per_million_input_tokens == pytest.approx(0.14)
+    assert model_info.cost_usd_per_million_output_tokens == pytest.approx(0.28)