Upload chatbot/retriever.py with huggingface_hub

chatbot/retriever.py

@@ -1,9 +1,4 @@
-"""
-Retriever wrapping the embedding pipeline.
-
-Loads MedEmbed + Qdrant client once and exposes a simple `retrieve(query)` interface.
-Filters out low-value sections (References, Application) that pollute results.
-"""
+"""Policy retrieval via MedEmbed embeddings and Qdrant vector search."""
 
 import sys
 import time
@@ -158,12 +153,14 @@ class PolicyRetriever:
                 page_end=p.get("page_end", 0),
             ))
 
-        # Boost Coverage Rationale so it always survives deduplication.
-        # This ensures the authoritative coverage/non-coverage statement
-        # is present even when Clinical Evidence chunks score higher.
-        COVERAGE_BOOST = 0.02
+        SECTION_BOOST = {
+            "Coverage Rationale": 0.04,
+            "Coverage Summary": 0.03,
+            "Benefit Considerations": 0.01,
+            "Documentation Requirements": 0.01,
+        }
         for r in results:
-            if r.section == "Coverage Rationale":
-                r.score += COVERAGE_BOOST
+            r.score += SECTION_BOOST.get(r.section, 0.0)
 
+        results.sort(key=lambda r: r.score, reverse=True)
         return results