feat: Redis-backed rate limiting with stricter auth limits (#194)

feat: Redis-backed rate limiting with stricter auth limits

src/cartsnitch_api/config.py

@@ -32,6 +32,9 @@ class Settings(BaseSettings):
 
     rate_limit_requests: int = 60
     rate_limit_window_seconds: int = 60
+    rate_limit_auth_requests: int = 5
+    rate_limit_auth_window_seconds: int = 60
+    rate_limit_redis_enabled: bool = True
     rate_limit_enabled: bool = True
 
     _PLACEHOLDER_VALUES = {"change-me-in-production"}
@@ -72,7 +75,9 @@ class Settings(BaseSettings):
     def normalize_database_url(self):
         """Normalize postgresql:// → postgresql+asyncpg:// for the asyncpg driver."""
         if self.database_url.startswith("postgresql://"):
-            self.database_url = self.database_url.replace("postgresql://", "postgresql+asyncpg://", 1)
+            self.database_url = self.database_url.replace(
+                "postgresql://", "postgresql+asyncpg://", 1
+            )
         return self
 
 

src/cartsnitch_api/middleware/rate_limit.py

@@ -5,18 +5,31 @@ Per-IP limiting on public endpoints, per-token limiting on authenticated endpoin
 """
 
 import hashlib
+import logging
 import time
+import uuid
 from collections import defaultdict
 from threading import Lock
+from typing import Protocol
 
 from fastapi import FastAPI, Request, status
 from fastapi.responses import JSONResponse
+from redis.asyncio import Redis, RedisError
 from starlette.middleware.base import BaseHTTPMiddleware
 
 from cartsnitch_api.config import settings
 
+logger = logging.getLogger(__name__)
 
-class _SlidingWindowCounter:
+
+class RateLimitBackend(Protocol):
+    """Protocol for rate limit backends."""
+
+    async def is_allowed(self, key: str) -> tuple[bool, int, int]:
+        """Check if request is allowed. Returns (allowed, remaining, retry_after)."""
+
+
+class InMemorySlidingWindow:
     """Thread-safe in-memory sliding window rate limiter."""
 
     def __init__(self, max_requests: int, window_seconds: int) -> None:
@@ -25,13 +38,12 @@ class _SlidingWindowCounter:
         self._hits: dict[str, list[float]] = defaultdict(list)
         self._lock = Lock()
 
-    def is_allowed(self, key: str) -> tuple[bool, int, int]:
+    async def is_allowed(self, key: str) -> tuple[bool, int, int]:
         """Check if request is allowed. Returns (allowed, remaining, retry_after)."""
         now = time.monotonic()
         cutoff = now - self.window_seconds
 
         with self._lock:
-            # Prune expired entries
             self._hits[key] = [t for t in self._hits[key] if t > cutoff]
 
             current_count = len(self._hits[key])
@@ -44,15 +56,84 @@ class _SlidingWindowCounter:
             return True, remaining, 0
 
 
-# Module-level counters — one for public (per-IP), one for auth (per-token)
-_public_limiter = _SlidingWindowCounter(
-    max_requests=settings.rate_limit_requests,
-    window_seconds=settings.rate_limit_window_seconds,
-)
-_auth_limiter = _SlidingWindowCounter(
-    max_requests=settings.rate_limit_requests * 5,  # 300/min for authenticated users
-    window_seconds=settings.rate_limit_window_seconds,
-)
+class RedisSlidingWindow:
+    """Redis-backed sliding window rate limiter using sorted sets."""
+
+    def __init__(self, redis: Redis, max_requests: int, window_seconds: int) -> None:
+        self.redis = redis
+        self.max_requests = max_requests
+        self.window_seconds = window_seconds
+
+    async def is_allowed(self, key: str) -> tuple[bool, int, int]:
+        """Check if request is allowed. Returns (allowed, remaining, retry_after)."""
+        try:
+            now = time.monotonic()
+            cutoff = now - self.window_seconds
+            now_ms = int(now * 1000)
+            cutoff_ms = int(cutoff * 1000)
+
+            pipe = self.redis.pipeline()
+            pipe.zremrangebyscore(key, 0, cutoff_ms)
+            pipe.zcard(key)
+            results = await pipe.execute()
+
+            current_count = results[1]
+
+            if current_count >= self.max_requests:
+                oldest = await self.redis.zrange(key, 0, 0, withscores=True)
+                if oldest:
+                    retry_after = int((oldest[0][1] - cutoff) / 1000) + 1
+                else:
+                    retry_after = self.window_seconds
+                return False, 0, retry_after
+
+            member = f"{now_ms}:{uuid.uuid4().hex[:8]}"
+            pipe = self.redis.pipeline()
+            pipe.zadd(key, {member: now_ms})
+            pipe.expire(key, self.window_seconds)
+            await pipe.execute()
+
+            remaining = self.max_requests - current_count - 1
+            return True, remaining, 0
+
+        except RedisError as e:
+            logger.warning("Redis rate limit error, falling back to in-memory: %s", e)
+            in_memory = InMemorySlidingWindow(self.max_requests, self.window_seconds)
+            return await in_memory.is_allowed(key)
+
+
+_redis_client: Redis | None = None
+_use_redis = False
+
+if settings.rate_limit_redis_enabled:
+    try:
+        _redis_client = Redis.from_url(settings.redis_url)
+        _use_redis = True
+        logger.info("Rate limiting will use Redis at %s", settings.redis_url)
+    except Exception as e:
+        logger.warning("Failed to connect to Redis for rate limiting, using in-memory: %s", e)
+        _use_redis = False
+
+if _use_redis and _redis_client:
+    _public_limiter = RedisSlidingWindow(
+        _redis_client, settings.rate_limit_requests, settings.rate_limit_window_seconds
+    )
+    _auth_limiter = RedisSlidingWindow(
+        _redis_client, settings.rate_limit_requests * 5, settings.rate_limit_window_seconds
+    )
+    _auth_strict_limiter = RedisSlidingWindow(
+        _redis_client, settings.rate_limit_auth_requests, settings.rate_limit_auth_window_seconds
+    )
+else:
+    _public_limiter = InMemorySlidingWindow(
+        settings.rate_limit_requests, settings.rate_limit_window_seconds
+    )
+    _auth_limiter = InMemorySlidingWindow(
+        settings.rate_limit_requests * 5, settings.rate_limit_window_seconds
+    )
+    _auth_strict_limiter = InMemorySlidingWindow(
+        settings.rate_limit_auth_requests, settings.rate_limit_auth_window_seconds
+    )
 
 
 def _get_client_ip(request: Request) -> str:
@@ -63,30 +144,30 @@ def _get_client_ip(request: Request) -> str:
     return request.client.host if request.client else "unknown"
 
 
-def _get_rate_limit_key(request: Request) -> tuple[str, _SlidingWindowCounter]:
+def _get_rate_limit_key(request: Request) -> tuple[str, RateLimitBackend]:
     """Determine rate limit key and which limiter to use."""
     if request.url.path.startswith("/public"):
         return f"ip:{_get_client_ip(request)}", _public_limiter
 
-    # For authenticated endpoints, use Bearer token as key if present
+    if request.url.path.startswith("/auth/") and request.method == "POST":
+        return f"ip:{_get_client_ip(request)}", _auth_strict_limiter
+
     auth_header = request.headers.get("authorization", "")
     if auth_header.startswith("Bearer "):
         token = auth_header[7:]
         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
 
 
 class RateLimitMiddleware(BaseHTTPMiddleware):
     async def dispatch(self, request: Request, call_next):
-        # Skip rate limiting when disabled (e.g. in tests) or for health checks
         if not settings.rate_limit_enabled or request.url.path == "/health":
             return await call_next(request)
 
         key, limiter = _get_rate_limit_key(request)
-        allowed, remaining, retry_after = limiter.is_allowed(key)
+        allowed, remaining, retry_after = await limiter.is_allowed(key)
 
         if not allowed:
             return JSONResponse(