fix(security): use SHA-256 hash for rate limit key instead of token suffix

Co-Authored-By: Paperclip <noreply@paperclip.ing>

api/alembic/versions/009_add_gin_index_upc_variants.py

@@ -1,38 +0,0 @@
-"""Add GIN index on upc_variants and alter column to JSONB.
-
-Revision ID: 009_add_gin_index_upc_variants
-Revises: 008_create_domain_tables
-Create Date: 2026-04-14
-"""
-
-import sqlalchemy as sa
-from alembic import op
-
-revision = "009_add_gin_index_upc_variants"
-down_revision = "008_create_domain_tables"
-branch_labels = None
-depends_on = None
-
-
-def upgrade() -> None:
-    op.alter_column(
-        "normalized_products",
-        "upc_variants",
-        type_=sa.dialects.postgresql.JSONB(),
-        postgresql_using="upc_variants::jsonb",
-    )
-    op.create_index(
-        "ix_normalized_products_upc_variants_gin",
-        "normalized_products",
-        ["upc_variants"],
-        postgresql_using="gin",
-    )
-
-
-def downgrade() -> None:
-    op.drop_index("ix_normalized_products_upc_variants_gin", table_name="normalized_products")
-    op.alter_column(
-        "normalized_products",
-        "upc_variants",
-        type_=sa.JSON(),
-    )

api/src/cartsnitch_api/middleware/cors.py

@@ -11,6 +11,6 @@ def add_cors_middleware(app: FastAPI) -> None:
         CORSMiddleware,
         allow_origins=settings.cors_origins,
         allow_credentials=True,
-        allow_methods=["GET", "POST", "PUT", "DELETE", "PATCH", "OPTIONS"],
-        allow_headers=["Content-Type", "Authorization", "Accept", "Origin", "X-Requested-With"],
+        allow_methods=["*"],
+        allow_headers=["*"],
     )

api/src/cartsnitch_api/middleware/rate_limit.py

@@ -4,6 +4,7 @@ Uses in-memory sliding window as fallback, Redis/DragonflyDB when available.
 Per-IP limiting on public endpoints, per-token limiting on authenticated endpoints.
 """
 
+import hashlib
 import time
 from collections import defaultdict
 from threading import Lock
@@ -71,8 +72,8 @@ def _get_rate_limit_key(request: Request) -> tuple[str, _SlidingWindowCounter]:
     auth_header = request.headers.get("authorization", "")
     if auth_header.startswith("Bearer "):
         token = auth_header[7:]
-        # Use last 16 chars of token as key to avoid storing full tokens
-        return f"token:{token[-16:]}", _auth_limiter
+        token_hash = hashlib.sha256(token.encode()).hexdigest()
+        return f"token:{token_hash}", _auth_limiter
 
     # Fallback to IP for unauthenticated non-public endpoints
     return f"ip:{_get_client_ip(request)}", _public_limiter

api/tests/test_middleware/test_rate_limit.py

@@ -1,8 +1,10 @@
 """Tests for rate limiting middleware."""
 
+from unittest.mock import MagicMock
+
 import pytest
 
-from cartsnitch_api.middleware.rate_limit import _SlidingWindowCounter
+from cartsnitch_api.middleware.rate_limit import _SlidingWindowCounter, _get_rate_limit_key
 
 
 class TestSlidingWindowCounter:
@@ -53,3 +55,32 @@ async def test_health_skips_rate_limit(client):
     resp = await client.get("/health")
     assert resp.status_code == 200
     assert "x-ratelimit-limit" not in resp.headers
+
+
+class TestGetRateLimitKey:
+    def _make_request(self, auth_header: str = "") -> MagicMock:
+        req = MagicMock()
+        req.url.path = "/purchases"
+        req.headers = {"authorization": auth_header} if auth_header else {}
+        return req
+
+    def test_distinct_tokens_produce_distinct_keys(self):
+        req1 = self._make_request("Bearer token_alpha_12345")
+        req2 = self._make_request("Bearer token_beta_67890")
+        key1, _ = _get_rate_limit_key(req1)
+        key2, _ = _get_rate_limit_key(req2)
+        assert key1 != key2
+
+    def test_same_token_produces_same_key(self):
+        req1 = self._make_request("Bearer same_token_value_abc")
+        req2 = self._make_request("Bearer same_token_value_abc")
+        key1, _ = _get_rate_limit_key(req1)
+        key2, _ = _get_rate_limit_key(req2)
+        assert key1 == key2
+
+    def test_key_does_not_contain_raw_token_suffix(self):
+        raw_token = "my_secret_jwt_token_xyz"
+        req = self._make_request(f"Bearer {raw_token}")
+        key, _ = _get_rate_limit_key(req)
+        assert raw_token[-16:] not in key
+        assert raw_token not in key

common/src/cartsnitch_common/models/product.py

@@ -3,7 +3,6 @@
 from typing import TYPE_CHECKING
 
 from sqlalchemy import JSON, String
-from sqlalchemy.dialects.postgresql import JSONB
 from sqlalchemy.orm import Mapped, mapped_column, relationship
 
 from cartsnitch_common.constants import ProductCategory, SizeUnit
@@ -27,9 +26,7 @@ class NormalizedProduct(UUIDPrimaryKeyMixin, TimestampMixin, Base):
     brand: Mapped[str | None] = mapped_column(String(200))
     size: Mapped[str | None] = mapped_column(String(50))
     size_unit: Mapped[SizeUnit | None] = mapped_column(String(10))
-    upc_variants: Mapped[list[str] | None] = mapped_column(
-        JSON().with_variant(JSONB(), "postgresql"), default=list
-    )
+    upc_variants: Mapped[list[str] | None] = mapped_column(JSON, default=list)
 
     # Relationships
     purchase_items: Mapped[list["PurchaseItem"]] = relationship(back_populates="normalized_product")

nginx.conf

@@ -9,12 +9,6 @@ server {
     gzip_types text/plain text/css application/json application/javascript text/xml application/xml application/xml+rss text/javascript image/svg+xml;
     gzip_min_length 256;
 
-    # Security headers
-    add_header X-Frame-Options "SAMEORIGIN" always;
-    add_header X-Content-Type-Options "nosniff" always;
-    add_header Referrer-Policy "strict-origin-when-cross-origin" always;
-    add_header Content-Security-Policy "default-src 'self'; script-src 'self'; style-src 'self' 'unsafe-inline'; img-src 'self' data: https:; font-src 'self'; connect-src 'self' https://*.cartsnitch.com https://*.farh.net; frame-ancestors 'self'" always;
-
     # Health endpoint for K8s probes
     location /health {
         access_log off;

package-lock.json

@@ -9805,9 +9805,9 @@
       }
     },
     "node_modules/vite": {
-      "version": "6.4.2",
-      "resolved": "https://registry.npmjs.org/vite/-/vite-6.4.2.tgz",
-      "integrity": "sha512-2N/55r4JDJ4gdrCvGgINMy+HH3iRpNIz8K6SFwVsA+JbQScLiC+clmAxBgwiSPgcG9U15QmvqCGWzMbqda5zGQ==",
+      "version": "6.4.1",
+      "resolved": "https://registry.npmjs.org/vite/-/vite-6.4.1.tgz",
+      "integrity": "sha512-+Oxm7q9hDoLMyJOYfUYBuHQo+dkAloi33apOPP56pzj+vsdJDzr+j1NISE5pyaAuKL4A3UD34qd0lx5+kfKp2g==",
       "devOptional": true,
       "license": "MIT",
       "dependencies": {

receiptwitness/src/receiptwitness/pipeline/normalization.py

@@ -5,14 +5,12 @@ Matches products across retailers by:
 2. Fuzzy name matching via token-based Jaccard similarity (lower confidence)
 """
 
-import json
 import re
 from dataclasses import dataclass
 from enum import StrEnum
 
 from cartsnitch_common.models.product import NormalizedProduct
-from sqlalchemy import cast, func, select, String
-from sqlalchemy.dialects.postgresql import JSONB
+from sqlalchemy import select
 from sqlalchemy.orm import Session
 
 
@@ -98,24 +96,17 @@ def jaccard_similarity(a: str, b: str) -> float:
 def match_by_upc(session: Session, upc: str) -> MatchResult | None:
     """Find a normalized product by exact UPC match.
 
-    Uses PostgreSQL JSONB containment (@>) for production efficiency.
-    Falls back to LIKE on SQLite for test compatibility.
+    Loads products with upc_variants and checks membership in Python
+    for cross-database compatibility (works on both PostgreSQL and SQLite).
     """
-    dialect_name = session.bind.dialect.name if session.bind else "default"
-    if dialect_name == "postgresql":
-        stmt = select(NormalizedProduct).where(
-            cast(NormalizedProduct.upc_variants, JSONB).op("@>")(
-                func.cast(json.dumps([upc]), JSONB)
-            )
-        )
-    else:
-        stmt = select(NormalizedProduct).where(
-            NormalizedProduct.upc_variants.is_not(None),
-            cast(NormalizedProduct.upc_variants, String).contains(upc),
-        )
-    product = session.execute(stmt).scalars().first()
-    if product:
-        return MatchResult(product=product, confidence=1.0, method=MatchMethod.UPC)
+    # TODO: Use PostgreSQL JSON containment query (@>) for production.
+    # Current approach loads all products into memory — acceptable for tests
+    # and small datasets, but will not scale.
+    stmt = select(NormalizedProduct).where(NormalizedProduct.upc_variants.is_not(None))
+    products = session.execute(stmt).scalars().all()
+    for product in products:
+        if product.upc_variants and upc in product.upc_variants:
+            return MatchResult(product=product, confidence=1.0, method=MatchMethod.UPC)
     return None