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

services/shared/llm_protocol.py

@@ -0,0 +1,44 @@
+"""LLM client protocol for provider abstraction.
+
+Defines the structural interface that both OllamaClient and VLLMClient
+must satisfy, using typing.Protocol for duck-typing compatibility.
+
+Requirements: 1.1, 1.2
+"""
+from __future__ import annotations
+
+from typing import TYPE_CHECKING, Protocol, runtime_checkable
+
+if TYPE_CHECKING:
+    from services.extractor.client import ExtractionAttempt
+
+
+@runtime_checkable
+class LLMClient(Protocol):
+    """Protocol defining the contract for LLM inference clients.
+
+    Both OllamaClient and VLLMClient satisfy this protocol via
+    structural subtyping — no inheritance required.
+    """
+
+    async def call_llm(
+        self,
+        prompts: dict[str, str],
+        json_schema: dict[str, object],
+        document_text: str = "",
+    ) -> ExtractionAttempt:
+        """Send a chat completion request and return an extraction attempt.
+
+        Args:
+            prompts: Dict with 'system' and 'user' prompt strings.
+            json_schema: JSON schema hint for structured output.
+            document_text: Optional raw document text for context.
+
+        Returns:
+            An ExtractionAttempt with raw output, validation, and error info.
+        """
+        ...
+
+    async def close(self) -> None:
+        """Release underlying HTTP resources."""
+        ...