test: add retry behavior tests for AI provider

Author: arshdeepkaur-code

backend/tests/test_ai_provider.py

@@ -21,7 +21,6 @@
 import pytest
 
 
-
 def _make_llm_response(text: str) -> MagicMock:
     """Return a fake httpx.Response with an OpenAI-compatible JSON body."""
     resp = MagicMock()
@@ -35,15 +34,15 @@ def _make_llm_response(text: str) -> MagicMock:
 def _make_error_response(status_code: int = 500) -> MagicMock:
     """Return a fake httpx.Response whose raise_for_status() raises."""
     resp = MagicMock()
-    resp.status_code = status_code  
-    
+    resp.status_code = status_code
+
     mock_response = MagicMock()
-    mock_response.status_code = status_code  
-    
+    mock_response.status_code = status_code
+
     resp.raise_for_status.side_effect = httpx.HTTPStatusError(
         message=f"HTTP {status_code}",
         request=MagicMock(),
-        response=mock_response, 
+        response=mock_response,
     )
     return resp
 
@@ -64,12 +63,11 @@ def _reload_module(env: dict):
     """Reload ai_provider so module-level env vars are re-evaluated."""
     with patch.dict(os.environ, env, clear=False):
         import app.services.ai_provider as mod
+
         importlib.reload(mod)
         return mod
 
 
-
-
 @pytest.fixture()
 def enabled_env():
     """Env vars that enable the LLM provider."""
@@ -199,6 +197,7 @@ async def test_whitespace_only_response_stripped_to_empty(self, enabled_env):
             patcher.stop()
         assert result == ""
 
+
 class TestCallLlmPayload:
 
     @pytest.mark.asyncio
@@ -267,6 +266,7 @@ async def test_calls_correct_ollama_endpoint(self, ollama_env):
         url_called = mock_client.post.call_args[0][0]
         assert url_called == "http://localhost:11434/v1/chat/completions"
 
+
 class TestCallLlmErrors:
 
     @pytest.mark.asyncio
@@ -316,7 +316,9 @@ async def test_returns_none_on_timeout(self, enabled_env):
     async def test_returns_none_on_network_error(self, enabled_env):
         mod = _reload_module(enabled_env)
         mock_client = AsyncMock()
-        mock_client.post = AsyncMock(side_effect=httpx.ConnectError("connection refused"))
+        mock_client.post = AsyncMock(
+            side_effect=httpx.ConnectError("connection refused")
+        )
 
         with patch("app.services.ai_provider.httpx.AsyncClient") as MockCls:
             MockCls.return_value.__aenter__ = AsyncMock(return_value=mock_client)
@@ -356,3 +358,69 @@ async def test_returns_none_on_empty_choices(self, enabled_env):
         finally:
             patcher.stop()
         assert result is None
+    
+    @pytest.mark.asyncio
+    async def test_retries_on_500_error(self,enabled_env):
+        mod = _reload_module(enabled_env)
+        
+        mock_client = AsyncMock()
+        mock_client.post = AsyncMock(
+            side_effect=httpx.HTTPStatusError(
+                "HTTP 500",
+                request=MagicMock(),
+                response=MagicMock(status_code=500),
+            )
+        )
+
+        with patch("app.services.ai_provider.httpx.AsyncClient") as MockCls:
+            MockCls.return_value.__aenter__ = AsyncMock(return_value=mock_client)
+            MockCls.return_value.__aexit__ = AsyncMock(return_value=False)
+
+            with patch("app.services.ai_provider.asyncio.sleep", new=AsyncMock()):
+                result = await mod.call_llm("sys", "usr")
+            
+        assert result is None
+        assert mock_client.post.call_count == mod.LLM_MAX_RETRIES + 1
+    @pytest.mark.asyncio
+    async def test_retries_on_429_error(self,enabled_env):
+        mod=_reload_module(enabled_env)
+        
+        mock_client=AsyncMock()
+        mock_client.post=AsyncMock(
+            side_effect=httpx.HTTPStatusError(
+                "HTTP 429",
+                request=MagicMock(),
+                response=MagicMock(status_code=429),
+            )
+        )
+
+        with patch("app.services.ai_provider.httpx.AsyncClient") as MockCls:
+            MockCls.return_value.__aenter__=AsyncMock(return_value=mock_client)
+            MockCls.return_value.__aexit__=AsyncMock(return_value=False)
+
+            with patch("app.services.ai_provider.asyncio.sleep", new=AsyncMock()):
+                result=await mod.call_llm("sys","usr")
+
+        assert result is None
+        assert mock_client.post.call_count == mod.LLM_MAX_RETRIES + 1
+
+    @pytest.mark.asyncio
+    async def test_retries_on_401_error(self,enabled_env):
+        mod=_reload_module(enabled_env)
+
+        mock_client=AsyncMock()
+        mock_client.post=AsyncMock(
+            side_effect=httpx.HTTPStatusError(
+                "HTTP 401",
+                request=MagicMock(),
+                response=MagicMock(status_code=401),
+            )
+        )
+
+        with patch("app.services.ai_provider.httpx.AsyncClient") as MockCls:
+            MockCls.return_value.__aenter__=AsyncMock(return_value=mock_client)
+            MockCls.return_value.__aexit__=AsyncMock(return_value=False)
+
+            result = await mod.call_llm("sys", "usr")
+        assert result is None
+        assert mock_client.post.call_count == 1