openai · lucasalencarxisto-stack · Aug 31, 2025 · Aug 31, 2025
@@ -1,3 +1,12 @@
 {
     "python.analysis.importFormat": "relative",
+    "python.testing.unittestArgs": [
+        "-v",
+        "-s",
+        "./tests",
+        "-p",
+        "*test.py"
+    ],
+    "python.testing.pytestEnabled": false,
+    "python.testing.unittestEnabled": true,
 }
@@ -0,0 +1,8 @@
+import pytest
+
+def test_dummy():
+    assert True
+
+@pytest.mark.anyio
+async def test_images_generate_includes_content_filter_results_async():
+    assert True
@@ -0,0 +1,251 @@
+import time
+import anyio
+import httpx
+import pytest
+
+def test_respects_retry_after_seconds(monkeypatch):
+    sleeps = []
+
+    # Evita dormir de verdade
+    monkeypatch.setattr(time, "sleep", lambda s: sleeps.append(s))
+
+    attempts = {"n": 0}
+
+    def handler(request: httpx.Request):
+        attempts["n"] += 1
+        if attempts["n"] == 1:
+            # 1ª tentativa falha com 429 e Retry-After: 2
+            return httpx.Response(429, headers={"Retry-After": "2"}, json={"err": "rate"})
+        # 2ª tentativa sucesso
+        return httpx.Response(200, json={"ok": True})
+
+    transport = httpx.MockTransport(handler)
+
+    from openai import OpenAI, DefaultHttpxClient
+
+    client = OpenAI(
+        max_retries=1,
+        http_client=DefaultHttpxClient(transport=transport),
+    )
+
+    client.responses.create(model="gpt-4o-mini", input="hi")
+
+    # Deve ter "dormido" ~2s antes de retry
+    assert sleeps, "expected a sleep before retry"
+    assert sleeps[0] >= 2.0  # pode ser >2.0 se houver jitter adicional
+
+
+@pytest.mark.asyncio
+async def test_respects_retry_after_seconds_async(monkeypatch):
+    sleeps = []
+
+    async def fake_sleep(s):
+        sleeps.append(s)
+
+    monkeypatch.setattr(anyio, "sleep", fake_sleep)
+
+    attempts = {"n": 0}
+
+    async def handler(request: httpx.Request):
+        attempts["n"] += 1
+        if attempts["n"] == 1:
+            return httpx.Response(429, headers={"Retry-After": "2"}, json={"err": "rate"})
+        return httpx.Response(200, json={"ok": True})
+        import time
+        import anyio
+        import httpx
+        import pytest
+
+        def test_respects_retry_after_seconds(monkeypatch):
+            sleeps = []
+            monkeypatch.setattr(time, "sleep", lambda s: sleeps.append(s))
+            attempts = {"n": 0}
+
+            def handler(request: httpx.Request):
+                attempts["n"] += 1
+                if attempts["n"] == 1:
+                    return httpx.Response(429, headers={"Retry-After": "2"}, json={"err": "rate"})
+                return httpx.Response(200, json={"ok": True})
+
+            transport = httpx.MockTransport(handler)
+            from openai import OpenAI, DefaultHttpxClient
+
+            client = OpenAI(
+                max_retries=1,
+                http_client=DefaultHttpxClient(transport=transport),
+            )
+
+            client.responses.create(model="gpt-4o-mini", input="hi")
+            assert sleeps, "expected a sleep before retry"
+            assert sleeps[0] >= 2.0
+
+        def test_no_retry_on_success(monkeypatch):
+            sleeps = []
+            monkeypatch.setattr(time, "sleep", lambda s: sleeps.append(s))
+            attempts = {"n": 0}
+
+            def handler(request: httpx.Request):
+                attempts["n"] += 1
+                return httpx.Response(200, json={"ok": True})
+
+            transport = httpx.MockTransport(handler)
+            from openai import OpenAI, DefaultHttpxClient
+
+            client = OpenAI(
+                max_retries=3,
+                http_client=DefaultHttpxClient(transport=transport),
+            )
+
+            client.responses.create(model="gpt-4o-mini", input="hi")
+            assert attempts["n"] == 1
+            assert not sleeps
+
+        def test_max_retries_exceeded(monkeypatch):
+            sleeps = []
+            monkeypatch.setattr(time, "sleep", lambda s: sleeps.append(s))
+            attempts = {"n": 0}
+
+            def handler(request: httpx.Request):
+                attempts["n"] += 1
+                return httpx.Response(429, headers={"Retry-After": "1"}, json={"err": "rate"})
+
+            transport = httpx.MockTransport(handler)
+            from openai import OpenAI, DefaultHttpxClient
+
+            client = OpenAI(
+                max_retries=2,
+                http_client=DefaultHttpxClient(transport=transport),
+            )
+
+            with pytest.raises(httpx.HTTPStatusError):
+                client.responses.create(model="gpt-4o-mini", input="hi")
+            assert attempts["n"] == 3  # 1 original + 2 retries
+            assert len(sleeps) == 2
+            assert all(s >= 1.0 for s in sleeps)
+
+        @pytest.mark.asyncio
+        async def test_respects_retry_after_seconds_async(monkeypatch):
+            sleeps = []
+            async def fake_sleep(s):
+                sleeps.append(s)
+            monkeypatch.setattr(anyio, "sleep", fake_sleep)
+            attempts = {"n": 0}
+
+            async def handler(request: httpx.Request):
+                attempts["n"] += 1
+                if attempts["n"] == 1:
+                    return httpx.Response(429, headers={"Retry-After": "2"}, json={"err": "rate"})
+                return httpx.Response(200, json={"ok": True})
+
+            transport = httpx.MockTransport(handler)
+            from openai import AsyncOpenAI, DefaultAsyncHttpxClient
+
+            client = AsyncOpenAI(
+                max_retries=1,
+                http_client=DefaultAsyncHttpxClient(transport=transport),
+            )
+
+            await client.responses.create(model="gpt-4o-mini", input="hi")
+            assert sleeps
+            assert sleeps[0] >= 2.0
+
+        @pytest.mark.asyncio
+        async def test_no_retry_on_success_async(monkeypatch):
+            sleeps = []
+            async def fake_sleep(s):
+                sleeps.append(s)
+            monkeypatch.setattr(anyio, "sleep", fake_sleep)
+            attempts = {"n": 0}
+
+            async def handler(request: httpx.Request):
+                attempts["n"] += 1
+                return httpx.Response(200, json={"ok": True})
+
+            transport = httpx.MockTransport(handler)
+            from openai import AsyncOpenAI, DefaultAsyncHttpxClient
+
+            client = AsyncOpenAI(
+                max_retries=3,
+                http_client=DefaultAsyncHttpxClient(transport=transport),
+            )
+
+            await client.responses.create(model="gpt-4o-mini", input="hi")
+            assert attempts["n"] == 1
+            assert not sleeps
+
+        @pytest.mark.asyncio
+        async def test_max_retries_exceeded_async(monkeypatch):
+            sleeps = []
+            async def fake_sleep(s):
+                sleeps.append(s)
+            monkeypatch.setattr(anyio, "sleep", fake_sleep)
+            attempts = {"n": 0}
+
+            async def handler(request: httpx.Request):
+                attempts["n"] += 1
+                return httpx.Response(429, headers={"Retry-After": "1"}, json={"err": "rate"})
+
+            transport = httpx.MockTransport(handler)
+            from openai import AsyncOpenAI, DefaultAsyncHttpxClient
+
+            client = AsyncOpenAI(
+                max_retries=2,
+                http_client=DefaultAsyncHttpxClient(transport=transport),
+            )
+
+            with pytest.raises(httpx.HTTPStatusError):
+                await client.responses.create(model="gpt-4o-mini", input="hi")
+            assert attempts["n"] == 3
+            assert len(sleeps) == 2
+            assert all(s >= 1.0 for s in sleeps)
+
+        def test_retry_with_jitter(monkeypatch):
+            # Testa se o retry respeita jitter (se implementado)
+            sleeps = []
+            monkeypatch.setattr(time, "sleep", lambda s: sleeps.append(s))
+            attempts = {"n": 0}
+
+            def handler(request: httpx.Request):
+                attempts["n"] += 1
+                if attempts["n"] < 3:
+                    return httpx.Response(429, headers={"Retry-After": "1"}, json={"err": "rate"})
+                return httpx.Response(200, json={"ok": True})
+
+            transport = httpx.MockTransport(handler)
+            from openai import OpenAI, DefaultHttpxClient
+
+            client = OpenAI(
+                max_retries=2,
+                http_client=DefaultHttpxClient(transport=transport),
+            )
+
+            with pytest.raises(httpx.HTTPStatusError):
+                client.responses.create(model="gpt-4o-mini", input="hi")
+            assert attempts["n"] == 3
+            assert len(sleeps) == 2
+
+        def test_retry_on_different_status(monkeypatch):
+            # Testa se o retry ocorre para outros status além de 429, se suportado
+            sleeps = []
+            monkeypatch.setattr(time, "sleep", lambda s: sleeps.append(s))
+            attempts = {"n": 0}
+
+            def handler(request: httpx.Request):
+                attempts["n"] += 1
+                if attempts["n"] == 1:
+                    return httpx.Response(500, json={"err": "server"})
+                return httpx.Response(200, json={"ok": True})
+
+            transport = httpx.MockTransport(handler)
+            from openai import OpenAI, DefaultHttpxClient
+
+            client = OpenAI(
+                max_retries=1,
+                http_client=DefaultHttpxClient(transport=transport),
+            )
+
+            client.responses.create(model="gpt-4o-mini", input="hi")
+            assert attempts["n"] == 2
+
+        # Estrutura assim está ótima: cada teste cobre um cenário e está bem separado.
+        # Se quiser, pode criar subpastas em tests/ para agrupar por tema (ex: tests/retries/, tests/timeouts/), mas para projetos pequenos/médios, manter arquivos separados já é suficiente e claro.