Add model wrapper modules (embedding, reranker, llm, preference_extractor)

Add Python wrappers for:
- Qwen3/Nemotron embedding models
- BGE/Qwen3 rerankers
- vLLM/Llama/Qwen LLM backends
- GPT-4o/LLM-based preference extractors

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

1 files changed, 16 insertions, 0 deletions

diff --git a/src/personalization/models/reranker/base.py b/src/personalization/models/reranker/base.py
new file mode 100644
index 0000000..34cf6ce
--- /dev/null
+++ b/src/personalization/models/reranker/base.py
@@ -0,0 +1,16 @@
+from typing import List, Protocol
+
+class Reranker(Protocol):
+    def score(
+        self,
+        query: str,
+        docs: List[str],
+        **kwargs,
+    ) -> List[float]:
+        """
+        Score multiple candidate documents for the same query.
+        Higher score indicates higher relevance.
+        Returns a list of floats with length equal to len(docs).
+        """
+        ...
+