feat: add remote vLLM support with provider abstraction layer

- LLMClient Protocol for provider-agnostic inference
- VLLMClient for OpenAI-compatible /v1/chat/completions API
- LLM client factory with provider routing (ollama/vllm)
- VLLMConfig with VLLM_* environment variable loading
- Updated extractor worker with health check and provider switching
- Updated event classifier to use LLMClient protocol
- Helm values for vLLM configuration
- 18 unit tests + 6 property-based tests
- Full backward compatibility preserved

tests/test_event_classifier.py

@@ -274,19 +274,19 @@ class TestParseClassificationResponse:
 
 class TestClassifyGlobalEvent:
     def _make_mock_client(self, raw_output: str, error: str | None = None):
-        """Create a mock OllamaClient with configurable response."""
+        """Create a mock LLMClient with configurable response."""
         client = MagicMock()
         client._config = MagicMock()
         client._config.model = "llama3.1:8b"
-        client._max_retries = 2
-        client._base_delay = 0.01
-        client._max_delay = 0.1
-        client._backoff_multiplier = 2.0
+        client._config.max_retries = 2
+        client._config.retry_base_delay = 0.01
+        client._config.retry_max_delay = 0.1
+        client._config.retry_backoff_multiplier = 2.0
 
         attempt = MagicMock()
         attempt.raw_output = raw_output
         attempt.error = error
-        client._call_ollama = AsyncMock(return_value=attempt)
+        client.call_llm = AsyncMock(return_value=attempt)
         return client
 
     @pytest.mark.asyncio
@@ -314,7 +314,7 @@ class TestClassifyGlobalEvent:
         assert event.severity == "critical"
         assert event.confidence == 0.9
         assert event.source_document_id == "doc-123"
-        client._call_ollama.assert_called_once()
+        client.call_llm.assert_called_once()
 
     @pytest.mark.asyncio
     async def test_retries_on_error(self):
@@ -340,11 +340,11 @@ class TestClassifyGlobalEvent:
         success_attempt.error = None
 
         client = self._make_mock_client("")
-        client._call_ollama = AsyncMock(side_effect=[fail_attempt, success_attempt])
+        client.call_llm = AsyncMock(side_effect=[fail_attempt, success_attempt])
 
         event = await classify_global_event("text", "doc-456", client)
         assert event.severity == "high"
-        assert client._call_ollama.call_count == 2
+        assert client.call_llm.call_count == 2
 
     @pytest.mark.asyncio
     async def test_raises_after_exhausted_retries(self):
@@ -353,12 +353,12 @@ class TestClassifyGlobalEvent:
         fail_attempt.error = "timeout"
 
         client = self._make_mock_client("")
-        client._call_ollama = AsyncMock(return_value=fail_attempt)
+        client.call_llm = AsyncMock(return_value=fail_attempt)
 
         with pytest.raises(ValueError, match="Event classification failed"):
             await classify_global_event("text", "doc-789", client)
 
-        assert client._call_ollama.call_count == 3  # initial + 2 retries
+        assert client.call_llm.call_count == 3  # initial + 2 retries
 
     @pytest.mark.asyncio
     async def test_minio_persistence_called(self):