fix: use structured error metadata for app-layer retry

tests/providers/test_provider_error_metadata.py

@@ -0,0 +1,77 @@
+from types import SimpleNamespace
+
+from nanobot.providers.anthropic_provider import AnthropicProvider
+from nanobot.providers.openai_compat_provider import OpenAICompatProvider
+
+
+def _fake_response(
+    *,
+    status_code: int,
+    headers: dict[str, str] | None = None,
+    text: str = "",
+) -> SimpleNamespace:
+    return SimpleNamespace(
+        status_code=status_code,
+        headers=headers or {},
+        text=text,
+    )
+
+
+def test_openai_handle_error_extracts_structured_metadata() -> None:
+    class FakeStatusError(Exception):
+        pass
+
+    err = FakeStatusError("boom")
+    err.status_code = 409
+    err.response = _fake_response(
+        status_code=409,
+        headers={"retry-after-ms": "250", "x-should-retry": "false"},
+        text='{"error":"conflict"}',
+    )
+    err.body = {"error": "conflict"}
+
+    response = OpenAICompatProvider._handle_error(err)
+
+    assert response.finish_reason == "error"
+    assert response.error_status_code == 409
+    assert response.error_retry_after_s == 0.25
+    assert response.error_should_retry is False
+
+
+def test_openai_handle_error_marks_timeout_kind() -> None:
+    class FakeTimeoutError(Exception):
+        pass
+
+    response = OpenAICompatProvider._handle_error(FakeTimeoutError("timeout"))
+
+    assert response.finish_reason == "error"
+    assert response.error_kind == "timeout"
+
+
+def test_anthropic_error_response_extracts_structured_metadata() -> None:
+    class FakeStatusError(Exception):
+        pass
+
+    err = FakeStatusError("boom")
+    err.status_code = 408
+    err.response = _fake_response(
+        status_code=408,
+        headers={"retry-after": "1.5", "x-should-retry": "true"},
+    )
+
+    response = AnthropicProvider._error_response(err)
+
+    assert response.finish_reason == "error"
+    assert response.error_status_code == 408
+    assert response.error_retry_after_s == 1.5
+    assert response.error_should_retry is True
+
+
+def test_anthropic_error_response_marks_connection_kind() -> None:
+    class FakeConnectionError(Exception):
+        pass
+
+    response = AnthropicProvider._error_response(FakeConnectionError("connection"))
+
+    assert response.finish_reason == "error"
+    assert response.error_kind == "connection"

tests/providers/test_provider_retry.py

@@ -240,6 +240,100 @@ async def test_chat_with_retry_uses_retry_after_and_emits_wait_progress(monkeypa
     assert progress and "7s" in progress[0]
 
 
+@pytest.mark.asyncio
+async def test_chat_with_retry_retries_structured_status_code_without_keyword(monkeypatch) -> None:
+    provider = ScriptedProvider([
+        LLMResponse(
+            content="request failed",
+            finish_reason="error",
+            error_status_code=409,
+        ),
+        LLMResponse(content="ok"),
+    ])
+    delays: list[float] = []
+
+    async def _fake_sleep(delay: float) -> None:
+        delays.append(delay)
+
+    monkeypatch.setattr("nanobot.providers.base.asyncio.sleep", _fake_sleep)
+
+    response = await provider.chat_with_retry(messages=[{"role": "user", "content": "hello"}])
+
+    assert response.content == "ok"
+    assert provider.calls == 2
+    assert delays == [1]
+
+
+@pytest.mark.asyncio
+async def test_chat_with_retry_retries_structured_timeout_kind(monkeypatch) -> None:
+    provider = ScriptedProvider([
+        LLMResponse(
+            content="request failed",
+            finish_reason="error",
+            error_kind="timeout",
+        ),
+        LLMResponse(content="ok"),
+    ])
+    delays: list[float] = []
+
+    async def _fake_sleep(delay: float) -> None:
+        delays.append(delay)
+
+    monkeypatch.setattr("nanobot.providers.base.asyncio.sleep", _fake_sleep)
+
+    response = await provider.chat_with_retry(messages=[{"role": "user", "content": "hello"}])
+
+    assert response.content == "ok"
+    assert provider.calls == 2
+    assert delays == [1]
+
+
+@pytest.mark.asyncio
+async def test_chat_with_retry_structured_should_retry_false_disables_retry(monkeypatch) -> None:
+    provider = ScriptedProvider([
+        LLMResponse(
+            content="429 rate limit",
+            finish_reason="error",
+            error_should_retry=False,
+        ),
+    ])
+    delays: list[float] = []
+
+    async def _fake_sleep(delay: float) -> None:
+        delays.append(delay)
+
+    monkeypatch.setattr("nanobot.providers.base.asyncio.sleep", _fake_sleep)
+
+    response = await provider.chat_with_retry(messages=[{"role": "user", "content": "hello"}])
+
+    assert response.finish_reason == "error"
+    assert provider.calls == 1
+    assert delays == []
+
+
+@pytest.mark.asyncio
+async def test_chat_with_retry_prefers_structured_retry_after(monkeypatch) -> None:
+    provider = ScriptedProvider([
+        LLMResponse(
+            content="429 rate limit, retry after 99s",
+            finish_reason="error",
+            error_retry_after_s=0.2,
+        ),
+        LLMResponse(content="ok"),
+    ])
+    delays: list[float] = []
+
+    async def _fake_sleep(delay: float) -> None:
+        delays.append(delay)
+
+    monkeypatch.setattr("nanobot.providers.base.asyncio.sleep", _fake_sleep)
+
+    response = await provider.chat_with_retry(messages=[{"role": "user", "content": "hello"}])
+
+    assert response.content == "ok"
+    assert delays == [0.2]
+
+
 @pytest.mark.asyncio
 async def test_persistent_retry_aborts_after_ten_identical_transient_errors(monkeypatch) -> None:
     provider = ScriptedProvider([
@@ -263,4 +357,3 @@ async def test_persistent_retry_aborts_after_ten_identical_transient_errors(monk
     assert provider.calls == 10
     assert delays == [1, 2, 4, 4, 4, 4, 4, 4, 4]
 
-