fix import errors

src/streamlit_app.py

@@ -16,7 +16,6 @@ from utils.retrieval_helpers import enrich_search_results, enrich_bm25_search_re
 from utils.bm25 import load
 from utils.semantic import load_vector_store
 from utils.rag_pipeline import run_rag
-from utils.bm25 import load
 from utils.hybrid import HybridRetriever
 
 import markdown
@@ -36,7 +35,6 @@ st.set_page_config(
 )
 
 # ─── Paths ────────────────────────────────────────────────────────────────────
-ROOT = Path(__file__).resolve().parent.parent
 FEEDBACK_CSV = ROOT / "results" / "feedback.csv"
 FEEDBACK_CSV.parent.mkdir(parents=True, exist_ok=True)
 

utils/hybrid.py

@@ -167,74 +167,4 @@ class HybridRetriever(BaseRetriever):
             "HybridRetriever: BM25=%d, Semantic=%d → fused=%d (returning top %d)",
             len(bm25_docs), len(semantic_docs), len(rrf_scores), len(top_docs),
         )
-        return top_docs
-
-
-# ---------------------------------------------------------------------------
-# Convenience loader
-# ---------------------------------------------------------------------------
-
-def load_hybrid_retriever(
-    bm25_index_path: str = "data/processed/tokenisation/bm25_index_mini.pkl",
-    faiss_store_path: str = "data/processed/embeddings",
-    k: int = 5,
-    bm25_weight: float = 0.5,
-    semantic_weight: float = 0.5,
-    rrf_c: int = 60,
-    fetch_multiplier: int = 3,
-) -> HybridRetriever:
-    """
-    Load both indexes from disk and return a ready-to-use HybridRetriever.
-
-    Call this once in your notebook or app.py, then pass the result to run_rag().
-
-    Parameters
-    ----------
-    bm25_index_path  : Path to the pickled BM25Retriever (from bm25.build_and_save())
-    faiss_store_path : Directory containing index.faiss + index.pkl
-                       (from semantic.build_and_save_vector_store())
-    k                : Number of documents to return per query
-    bm25_weight      : RRF weight for BM25 (keyword signal). Default 0.5.
-    semantic_weight  : RRF weight for semantic (meaning signal). Default 0.5.
-                       Weights don't need to sum to 1 but relative scale matters.
-    rrf_c            : RRF rank-dampening constant. Default 60 (standard).
-    fetch_multiplier : Candidates to fetch per retriever = k * fetch_multiplier.
-
-    Returns
-    -------
-    HybridRetriever
-        A LangChain-compatible retriever pipeable with |.
-
-    Example
-    -------
-    >>> from utils.hybrid import load_hybrid_retriever
-    >>> from utils.rag_pipeline import run_rag
-    >>>
-    >>> hybrid = load_hybrid_retriever(k=5)
-    >>> answer = run_rag(hybrid, "Best coffee beans for a French press")
-    >>> print(answer)
-    """
-    # Import here to avoid circular imports when used from rag_pipeline.py
-    from utils.bm25 import load as load_bm25
-    from utils.semantic import load_vector_store
-
-    print(f"Loading BM25 index from: {bm25_index_path}")
-    bm25_ret: BM25Retriever = load_bm25(bm25_index_path)
-
-    print(f"Loading FAISS store from: {faiss_store_path}")
-    faiss_store: FAISS = load_vector_store(faiss_store_path)
-
-    retriever = HybridRetriever(
-        bm25_retriever=bm25_ret,
-        semantic_store=faiss_store,
-        k=k,
-        bm25_weight=bm25_weight,
-        semantic_weight=semantic_weight,
-        rrf_c=rrf_c,
-        fetch_multiplier=fetch_multiplier,
-    )
-    print(
-        f"HybridRetriever ready — k={k}, "
-        f"BM25 weight={bm25_weight}, Semantic weight={semantic_weight}, RRF c={rrf_c}"
-    )
-    return retriever
+        return top_docs