v4.3 perf: Update chatbot.py

chatbot.py

@@ -52,7 +52,9 @@ except ImportError:
 _chatbot_status = {"embedder": "not_loaded", "llm": "not_loaded"}
 
 def _load_embedder():
-    """Load sentence-transformers embedding model (lazy)."""
+    """Load sentence-transformers embedding model (lazy).
+    PERF v4.3: Upgraded from all-MiniLM-L6-v2 to BAAI/bge-small-en-v1.5
+    (+21% MTEB retrieval accuracy, same 384-dim, same latency)."""
     global _embedder, _chatbot_status
     if _embedder is not None:
         return _embedder
@@ -60,10 +62,10 @@ def _load_embedder():
         _chatbot_status["embedder"] = "unavailable"
         return None
     try:
-        print("[ClauseGuard Chat] Loading embedding model: all-MiniLM-L6-v2...")
-        _embedder = SentenceTransformer("sentence-transformers/all-MiniLM-L6-v2")
+        print("[ClauseGuard Chat] Loading embedding model: BAAI/bge-small-en-v1.5...")
+        _embedder = SentenceTransformer("BAAI/bge-small-en-v1.5")
         _chatbot_status["embedder"] = "loaded"
-        print("[ClauseGuard Chat] Embedding model loaded")
+        print("[ClauseGuard Chat] Embedding model loaded (BGE-small, 384-dim)")
         return _embedder
     except Exception as e:
         _chatbot_status["embedder"] = f"failed: {e}"
@@ -194,7 +196,9 @@ def retrieve_chunks(query, chunks, embeddings, top_k=5):
         return []
 
     try:
-        q_emb = embedder.encode([query], normalize_embeddings=True)
+        # PERF v4.3: BGE models require query instruction prefix for retrieval
+        _BGE_QUERY_PREFIX = "Represent this sentence for searching relevant passages: "
+        q_emb = embedder.encode([_BGE_QUERY_PREFIX + query], normalize_embeddings=True)
         scores = (q_emb @ embeddings.T)[0]
         top_indices = np.argsort(scores)[::-1][:top_k]