fix: replace direct psycopg DB connection with Supabase REST API

Removes psycopg dependency and direct PostgreSQL TCP connection which
caused IPv6 connection failures on HuggingFace Spaces. All DB operations
now go through supabase-py over HTTPS via RPC calls.

backend/database.py

@@ -1,15 +1,12 @@
 """
 Database connection helpers for Supabase.
-Uses psycopg (v3) for direct PostgreSQL/pgvector queries (search),
+Uses Supabase RPC for pgvector similarity search (works over HTTPS),
 and supabase-py for standard CRUD operations.
 """
 
 import os
-import psycopg
-from psycopg.rows import dict_row
 import numpy as np
 from supabase import create_client, Client
-from config import SUPABASE_URL, SUPABASE_KEY, DATABASE_URL
 
 # --- Supabase Client (for CRUD operations) ---
 
@@ -20,21 +17,13 @@ def get_supabase() -> Client:
     """Get or create the Supabase client for standard CRUD operations."""
     global _supabase_client
     if _supabase_client is None:
-        url = (os.environ.get("SUPABASE_URL") or SUPABASE_URL).strip()
-        key = (os.environ.get("SUPABASE_KEY") or SUPABASE_KEY).strip()
-        _supabase_client = create_client(url, key)
+        _supabase_client = create_client(
+            os.environ["SUPABASE_URL"].strip(),
+            os.environ["SUPABASE_KEY"].strip(),
+        )
     return _supabase_client
 
 
-# --- Direct PostgreSQL connection (for pgvector queries) ---
-
-def get_db_connection():
-    """Create a new psycopg3 connection for pgvector queries."""
-    db_url = (os.environ.get("DATABASE_URL") or DATABASE_URL).strip()
-    conn = psycopg.connect(db_url, row_factory=dict_row)
-    return conn
-
-
 def close_db_pool():
     """Placeholder for shutdown compatibility."""
     pass
@@ -53,37 +42,23 @@ def pgvector_to_embedding(pgvector_str: str) -> np.ndarray:
     return np.array([float(v) for v in values], dtype=np.float32)
 
 
-# --- pgvector similarity search ---
+# --- pgvector similarity search (via Supabase RPC) ---
 
 def search_similar_products(query_embedding: np.ndarray, top_k: int = 50):
     """
     Find the top_k most similar products to the query embedding
-    using pgvector cosine similarity search.
-    Returns list of dicts with product info + embedding.
+    using pgvector cosine similarity search via Supabase RPC.
     """
-    conn = get_db_connection()
-    embedding_str = embedding_to_pgvector(query_embedding)
-    
-    query = """
-        SELECT
-            id, seller_id, title, description, price, stock, images,
-            embedding::text as embedding_text,
-            1 - (embedding <=> %s::vector) as similarity
-        FROM products
-        WHERE is_active = true AND status = 'approved' AND stock > 0 AND embedding IS NOT NULL
-        ORDER BY embedding <=> %s::vector
-        LIMIT %s
-    """
-    
-    try:
-        with conn.cursor() as cur:
-            cur.execute(query, (embedding_str, embedding_str, top_k))
-            rows = cur.fetchall()
-    finally:
-        conn.close()
+    sb = get_supabase()
+    embedding_list = query_embedding.tolist()
+
+    response = sb.rpc("search_products_by_embedding", {
+        "query_embedding": embedding_list,
+        "match_count": top_k,
+    }).execute()
 
     results = []
-    for row in rows:
+    for row in response.data:
         results.append({
             "id": str(row["id"]),
             "seller_id": str(row["seller_id"]),
@@ -108,61 +83,23 @@ def search_similar_products_filtered(
     color: str = None,
 ):
     """
-    Find the top_k most similar products with optional structured filters.
-    Extends search_similar_products with WHERE clauses from query rewriting.
+    Find the top_k most similar products with optional structured filters
+    via Supabase RPC.
     """
-    conn = get_db_connection()
-    embedding_str = embedding_to_pgvector(query_embedding)
-    
-    # Build dynamic WHERE clause
-    conditions = ["is_active = true", "status = 'approved'", "stock > 0", "embedding IS NOT NULL"]
-    filter_params = []
-
-    if price_min is not None:
-        conditions.append(f"price > %s")
-        filter_params.append(price_min)
-
-    if price_max is not None:
-        conditions.append(f"price < %s")
-        filter_params.append(price_max)
-
-    if brand:
-        conditions.append(f"(title ILIKE %s OR description ILIKE %s)")
-        filter_params.append(f"%{brand}%")
-        filter_params.append(f"%{brand}%")
-
-    if color:
-        conditions.append(f"(title ILIKE %s OR description ILIKE %s)")
-        filter_params.append(f"%{color}%")
-        filter_params.append(f"%{color}%")
-
-    where_clause = " AND ".join(conditions)
-
-    # Params must match SQL placeholder order:
-    # 1) embedding for SELECT similarity, 2) filter params for WHERE,
-    # 3) embedding for ORDER BY, 4) top_k for LIMIT
-    params = [embedding_str] + filter_params + [embedding_str, top_k]
-
-    query = f"""
-        SELECT
-            id, seller_id, title, description, price, stock, images,
-            embedding::text as embedding_text,
-            1 - (embedding <=> %s::vector) as similarity
-        FROM products
-        WHERE {where_clause}
-        ORDER BY embedding <=> %s::vector
-        LIMIT %s
-    """
-
-    try:
-        with conn.cursor() as cur:
-            cur.execute(query, tuple(params))
-            rows = cur.fetchall()
-    finally:
-        conn.close()
+    sb = get_supabase()
+    embedding_list = query_embedding.tolist()
+
+    response = sb.rpc("search_products_by_embedding_filtered", {
+        "query_embedding": embedding_list,
+        "match_count": top_k,
+        "filter_price_min": price_min,
+        "filter_price_max": price_max,
+        "filter_brand": brand,
+        "filter_color": color,
+    }).execute()
 
     results = []
-    for row in rows:
+    for row in response.data:
         results.append({
             "id": str(row["id"]),
             "seller_id": str(row["seller_id"]),
@@ -179,17 +116,11 @@ def search_similar_products_filtered(
 
 
 def store_product_embedding(product_id: str, embedding: np.ndarray):
-    """Store/update the BERT embedding for a product."""
-    conn = get_db_connection()
+    """Store/update the BERT embedding for a product via Supabase RPC."""
+    sb = get_supabase()
     embedding_str = embedding_to_pgvector(embedding)
-    
-    query = """
-        UPDATE products SET embedding = %s::vector WHERE id = %s
-    """
-    
-    try:
-        with conn.cursor() as cur:
-            cur.execute(query, (embedding_str, product_id))
-        conn.commit()
-    finally:
-        conn.close()
+
+    sb.rpc("update_product_embedding", {
+        "p_product_id": product_id,
+        "p_embedding": embedding_str,
+    }).execute()

backend/requirements.txt

@@ -7,7 +7,6 @@ python-dotenv==1.0.1
 
 # Database
 supabase==2.7.2
-psycopg[binary]==3.3.2
 
 # ML / AI
 torch>=2.0.0