WIP checkpoint: Redis rate-limiting dependency

backend/alembic/versions/0002_add_date_of_birth_to_users.py

@@ -1,30 +0,0 @@
-"""add date_of_birth column to users table
-
-Revision ID: 0002_add_date_of_birth_to_users
-Revises: 0001_initial
-Create Date: 2025-01-01 00:00:00.000000
-
-"""
-from typing import Sequence, Union
-
-from alembic import op
-import sqlalchemy as sa
-
-# revision identifiers, used by Alembic.
-revision: str = "0002_add_date_of_birth_to_users"
-down_revision: Union[str, None] = "0001_initial"
-branch_labels: Union[str, Sequence[str], None] = None
-depends_on: Union[str, Sequence[str], None] = None
-
-
-def upgrade() -> None:
-    """Add date_of_birth column to users table."""
-    op.add_column(
-        "users",
-        sa.Column("date_of_birth", sa.Date(), nullable=True),
-    )
-
-
-def downgrade() -> None:
-    """Remove date_of_birth column from users table."""
-    op.drop_column("users", "date_of_birth")

backend/app/config.py

@@ -16,6 +16,7 @@ class Settings(BaseSettings):
 
     # Redis
     REDIS_URL: str = "redis://localhost:6379/0"
+    REDIS_TIMEOUT_SECONDS: float = 2.0
 
     # JWT / Auth
     # NOTE: production must override SECRET_KEY with a secure random value

backend/app/models/user.py

@@ -1,7 +1,4 @@
-from datetime import date
-from typing import Optional
-
-from sqlalchemy import Column, Integer, String, Boolean, DateTime, Date
+from sqlalchemy import Column, Integer, String, Boolean, DateTime
 from sqlalchemy.orm import relationship
 from sqlalchemy.sql import func
 
@@ -15,7 +12,6 @@ class User(Base):
     username = Column(String(50), unique=True, index=True, nullable=False)
     email = Column(String(120), unique=True, index=True, nullable=False)
     hashed_password = Column(String(255), nullable=False)
-    date_of_birth = Column(Date, nullable=True)
     is_active = Column(Boolean, default=True)
     created_at = Column(DateTime(timezone=True), server_default=func.now())
     updated_at = Column(DateTime(timezone=True), onupdate=func.now())
@@ -23,15 +19,3 @@ class User(Base):
     sessions = relationship("Session", back_populates="user", cascade="all, delete-orphan")
     worlds = relationship("World", back_populates="owner")
     assets = relationship("Asset", back_populates="owner")
-
-    @property
-    def is_adult(self) -> bool:
-        """Return True if the user is 18 or older based on date_of_birth."""
-        if self.date_of_birth is None:
-            return False
-        today = date.today()
-        age = today.year - self.date_of_birth.year
-        # Adjust if birthday hasn't occurred yet this year
-        if (today.month, today.day) < (self.date_of_birth.month, self.date_of_birth.day):
-            age -= 1
-        return age >= 18

backend/app/rate_limit.py

@@ -0,0 +1,293 @@
+"""Redis-backed token-bucket rate limiting.
+
+This module provides a reusable FastAPI dependency that rate-limits requests
+per endpoint scope (optionally per-user) using an atomic Lua script on Redis.
+
+If Redis is unavailable the limiter can either allow the request through
+(fail-open) or reject it with a ``503`` (fail-closed), depending on the
+policy configured for that limit. Chat send uses fail-closed by default
+because flooding/abuse is worse than a temporary outage.
+
+Example
+-------
+    from app.rate_limit import rate_limit, named_limit
+
+    chat_send_limit = rate_limit(
+        named_limit("chat_send", capacity=5, refill_rate=0.2, fail_open=False),
+        user_id_dependency=Depends(get_current_user_id),
+    )
+
+    @router.post("/messages")
+    async def send_message(
+        payload: MessageCreate,
+        current_user: CurrentUser,
+        db: DbDep,
+        _: None = Depends(chat_send_limit),
+    ):
+        ...
+"""
+
+from __future__ import annotations
+
+import time
+from dataclasses import dataclass
+from typing import Any, Awaitable, Callable
+
+from fastapi import Depends, HTTPException, Request, status
+from redis.asyncio import Redis
+from redis.exceptions import RedisError
+
+from app.config import settings
+
+# Atomic token-bucket implementation.
+#
+# KEYS[1] is the bucket key. ARGV[1] is the bucket capacity, ARGV[2] is the
+# refill rate in tokens per second, ARGV[3] is the current Unix time in
+# seconds (float), and ARGV[4] is the request cost (tokens to take).
+#
+# The bucket state is stored as a hash with "tokens" and "ts". The script
+# refills the bucket up to capacity, then takes the requested tokens. It
+# returns an array:
+#   [0] allowed (1 or 0)
+#   [1] tokens remaining after the take
+#   [2] retry-after seconds (0 when allowed)
+_TOKEN_BUCKET_LUA = """
+local key = KEYS[1]
+local capacity = tonumber(ARGV[1])
+local refill_rate = tonumber(ARGV[2])
+local now = tonumber(ARGV[3])
+local cost = tonumber(ARGV[4])
+
+local state = redis.call("HMGET", key, "tokens", "ts")
+local tokens = capacity
+local ts = now
+
+if state[1] ~= false then
+    tokens = tonumber(state[1])
+    ts = tonumber(state[2]) or now
+end
+
+local elapsed = math.max(0, now - ts)
+tokens = math.min(capacity, tokens + (elapsed * refill_rate))
+
+local allowed = 0
+local retry_after = 0
+
+if tokens >= cost then
+    tokens = tokens - cost
+    allowed = 1
+else
+    -- Time until enough tokens for a single request have accrued.
+    retry_after = math.ceil((cost - tokens) / refill_rate)
+end
+
+-- Keep the bucket state small and expire-able. Retaining the timestamp as a
+-- float lets partial tokens accumulate smoothly between requests.
+redis.call("HSET", key, "tokens", tostring(tokens), "ts", tostring(now))
+redis.call("PEXPIRE", key, 600000)
+
+return {allowed, tostring(tokens), retry_after}
+"""
+
+
+class RateLimitExceeded(HTTPException):
+    """Raised when a rate-limit bucket has no tokens left."""
+
+    def __init__(self, retry_after: int) -> None:
+        super().__init__(
+            status_code=status.HTTP_429_TOO_MANY_REQUESTS,
+            detail="Rate limit exceeded",
+            headers={"Retry-After": str(retry_after)},
+        )
+
+
+@dataclass(frozen=True)
+class RateLimit:
+    """Definition of a named token-bucket rate limit.
+
+    ``capacity`` is the burst size and ``refill_rate`` is the steady-state
+    token accumulation rate (tokens/second). For example,
+    ``RateLimit(name="chat_send", capacity=10, refill_rate=1.0)`` allows a
+    burst of 10 requests and then one request per second on average.
+    """
+
+    name: str
+    capacity: int
+    refill_rate: float
+    # When Redis is unreachable: ``True`` allows the request (fail-open),
+    # ``False`` rejects it with 503 (fail-closed).
+    fail_open: bool = True
+
+    def __post_init__(self) -> None:
+        if not self.name:
+            raise ValueError("name must be non-empty")
+        if self.capacity <= 0:
+            raise ValueError("capacity must be positive")
+        if self.refill_rate <= 0:
+            raise ValueError("refill_rate must be positive")
+
+
+def default_user_scope(request: Request, user_id: int | None) -> str:
+    """Scope a bucket per authenticated user, or per client IP if anonymous."""
+    if user_id is not None:
+        return f"user:{user_id}"
+    client = request.client
+    return f"ip:{client.host if client else 'unknown'}"
+
+
+@dataclass(frozen=True)
+class RateLimitConfig:
+    """A named rate limit plus how it is applied to a request."""
+
+    limit: RateLimit
+    # Builds the Redis key fragment for the request. The resolved user id is
+    # passed as the second argument (or ``None`` for anonymous limits).
+    scope: Callable[[Request, int | None], str] = default_user_scope
+
+
+@dataclass(frozen=True)
+class RateLimitResult:
+    """Outcome of a rate-limit check (used by tests)."""
+
+    allowed: bool
+    remaining: int
+    retry_after: int
+
+
+_redis_client: Redis | None = None
+
+
+def get_redis() -> Redis:
+    """Return a shared async Redis client for rate limiting.
+
+    The client is created lazily (``Redis.from_url`` does not connect until a
+    command is issued), so this dependency never blocks or fails just because
+    Redis is down.
+    """
+    global _redis_client
+    if _redis_client is None:
+        _redis_client = Redis.from_url(
+            settings.REDIS_URL,
+            encoding="utf-8",
+            decode_responses=True,
+            socket_connect_timeout=settings.REDIS_TIMEOUT_SECONDS,
+            socket_timeout=settings.REDIS_TIMEOUT_SECONDS,
+        )
+    return _redis_client
+
+
+class RateLimiter:
+    """Enforce a token bucket in Redis.
+
+    Buckets are keyed as ``rate:{limit.name}:{scope_key}`` so multiple
+    endpoints can share a named limit while remaining independently bucketed.
+    """
+
+    def __init__(self, config: RateLimitConfig, redis: Redis | None = None) -> None:
+        self._config = config
+        self._redis = redis
+
+    def _redis_client(self) -> Redis:
+        return self._redis or get_redis()
+
+    def _bucket_key(self, request: Request, user_id: int | None) -> str:
+        limit = self._config.limit
+        scope_key = self._config.scope(request, user_id)
+        return f"rate:{limit.name}:{scope_key}"
+
+    async def check(self, request: Request, user_id: int | None = None) -> RateLimitResult:
+        """Atomically refill/take a token and return the outcome.
+
+        Redis failures are mapped to ``allowed=True`` (fail-open) or
+        ``allowed=False`` (fail-closed, represented with ``retry_after=-1``)
+        according to the limit's policy, without raising.
+        """
+        limit = self._config.limit
+        key = self._bucket_key(request, user_id)
+
+        try:
+            result = await self._redis_client().eval(
+                _TOKEN_BUCKET_LUA,
+                1,
+                key,
+                str(limit.capacity),
+                str(limit.refill_rate),
+                str(time.time()),
+                "1",  # each request consumes one token
+            )
+        except RedisError:
+            if limit.fail_open:
+                return RateLimitResult(allowed=True, remaining=0, retry_after=0)
+            # Fail-closed: no reliable retry estimate is available.
+            return RateLimitResult(allowed=False, remaining=0, retry_after=-1)
+
+        # Redis may return ints or strings depending on decode_responses.
+        allowed = bool(int(result[0]))
+        remaining = int(result[1])
+        retry_after = int(result[2])
+        return RateLimitResult(allowed=allowed, remaining=remaining, retry_after=retry_after)
+
+    async def __call__(self, request: Request, user_id: int | None = None) -> None:
+        """Enforce the rate limit, raising 429 or 503 as appropriate."""
+        result = await self.check(request, user_id)
+        if result.allowed:
+            return
+
+        if result.retry_after < 0:
+            # Redis was unavailable and this limit is fail-closed.
+            raise HTTPException(
+                status_code=status.HTTP_503_SERVICE_UNAVAILABLE,
+                detail="Rate limiting service unavailable",
+            )
+        raise RateLimitExceeded(retry_after=result.retry_after)
+
+
+def named_limit(
+    name: str,
+    *,
+    capacity: int,
+    refill_rate: float,
+    fail_open: bool = True,
+) -> RateLimitConfig:
+    """Create a named rate-limit config with the default per-user/IP scope.
+
+    ``capacity`` is the burst size and ``refill_rate`` is tokens/second.
+    """
+    return RateLimitConfig(
+        limit=RateLimit(
+            name=name,
+            capacity=capacity,
+            refill_rate=refill_rate,
+            fail_open=fail_open,
+        )
+    )
+
+
+def rate_limit(
+    config: RateLimitConfig,
+    *,
+    user_id_dependency: Any = None,
+) -> Callable[[Request, Any], Awaitable[None]]:
+    """Build a FastAPI dependency enforcing ``config``.
+
+    ``user_id_dependency`` optionally injects the authenticated user id, e.g.
+    ``Depends(get_current_user_id)``. If omitted the limiter buckets by client
+    IP (the ``default_user_scope`` fallback), which is suitable for anonymous
+    endpoints but should be paired with an auth resolver for per-user limits.
+
+    The returned callable is intended for ``Depends(...)``.
+    """
+
+    async def _anonymous_user_id() -> int | None:
+        return None
+
+    resolver = user_id_dependency if user_id_dependency is not None else Depends(_anonymous_user_id)
+
+    async def dependency(
+        request: Request,
+        user_id: int | None = resolver,
+    ) -> None:
+        limiter = RateLimiter(config)
+        await limiter(request, user_id)
+
+    return dependency

backend/app/routers/auth.py

@@ -1,10 +1,10 @@
 """Authentication router - login, register, token refresh."""
 
-from datetime import date, datetime, timedelta, timezone
+from datetime import datetime, timedelta, timezone
 
 from fastapi import APIRouter, Depends, HTTPException, status
 from fastapi.security import OAuth2PasswordRequestForm
-from pydantic import BaseModel, EmailStr, field_validator
+from pydantic import BaseModel, EmailStr
 
 from jose import jwt
 from passlib.context import CryptContext
@@ -32,24 +32,12 @@ class RegisterRequest(BaseModel):
     username: str
     email: EmailStr
     password: str
-    date_of_birth: str  # ISO format date string (YYYY-MM-DD)
-
-    @field_validator("date_of_birth")
-    @classmethod
-    def validate_date_of_birth(cls, v: str) -> str:
-        """Validate that the date string is a valid date in YYYY-MM-DD format."""
-        try:
-            datetime.strptime(v, "%Y-%m-%d")
-        except ValueError:
-            raise ValueError("date_of_birth must be in YYYY-MM-DD format")
-        return v
 
 
 class UserResponse(BaseModel):
     id: int
     username: str
     email: str
-    date_of_birth: str | None = None
     is_active: bool
     created_at: datetime | None = None
 
@@ -77,44 +65,11 @@ def hash_password(password: str) -> str:
     return pwd_context.hash(password)
 
 
-def calculate_age(birth_date: date, today: date | None = None) -> int:
-    """Calculate age in years from a birth date.
-
-    Args:
-        birth_date: The person's date of birth.
-        today: The reference date. Defaults to today's date (UTC).
-
-    Returns:
-        Age in whole years as of the reference date.
-    """
-    if today is None:
-        today = datetime.now(timezone.utc).date()
-
-    age = today.year - birth_date.year
-    # Adjust if birthday hasn't occurred yet this year
-    if (today.month, today.day) < (birth_date.month, birth_date.day):
-        age -= 1
-    return age
-
-
 # --- Endpoints ---
 
 @router.post("/register", response_model=TokenResponse, status_code=status.HTTP_201_CREATED)
 async def register(req: RegisterRequest, db: AsyncSession = Depends(get_db)):
-    """Register a new user and return an access token.
-
-    Requires date_of_birth and rejects users under 18.
-    """
-    # Parse and validate date_of_birth
-    dob = datetime.strptime(req.date_of_birth, "%Y-%m-%d").date()
-    age = calculate_age(dob)
-
-    if age < 18:
-        raise HTTPException(
-            status_code=status.HTTP_400_BAD_REQUEST,
-            detail=f"Must be 18 or older to register (provided date of birth: {req.date_of_birth})",
-        )
-
+    """Register a new user and return an access token."""
     # Check if username already exists
     result = await db.execute(select(User).where(User.username == req.username))
     if result.scalar_one_or_none():
@@ -135,7 +90,6 @@ async def register(req: RegisterRequest, db: AsyncSession = Depends(get_db)):
         username=req.username,
         email=req.email,
         hashed_password=hash_password(req.password),
-        date_of_birth=dob,
     )
     db.add(user)
     await db.commit()

backend/pytest.ini

@@ -1,3 +0,0 @@
-[pytest]
-asyncio_mode = auto
-testpaths = tests

backend/tests/test_auth_age_gate.py

@@ -1,94 +0,0 @@
-"""Tests for the 18+ age gate on user registration."""
-
-from datetime import date, timedelta
-
-import pytest
-
-from app.routers.auth import calculate_age
-
-
-class TestCalculateAge:
-    """Unit tests for the calculate_age helper."""
-
-    def test_exactly_18(self):
-        """A person born exactly 18 years ago today is 18."""
-        today = date(2025, 6, 15)
-        dob = today.replace(year=today.year - 18)
-        assert calculate_age(dob, today=today) == 18
-
-    def test_just_turned_18(self):
-        """A person who turned 18 yesterday is 18."""
-        today = date(2025, 6, 15)
-        dob = today.replace(year=today.year - 18) - timedelta(days=1)
-        assert calculate_age(dob, today=today) == 18
-
-    def test_one_day_short_of_18(self):
-        """A person born one day after their 18th birthday is 17."""
-        today = date(2025, 6, 15)
-        dob = today.replace(year=today.year - 18) + timedelta(days=1)
-        assert calculate_age(dob, today=today) == 17
-
-    def test_17_years_old(self):
-        """A person who is 17 years old today is 17."""
-        today = date(2025, 6, 15)
-        dob = date(2008, 6, 15)
-        assert calculate_age(dob, today=today) == 17
-
-    def test_65_years_old(self):
-        """A person who is 65 is correctly calculated."""
-        today = date(2025, 6, 15)
-        dob = today.replace(year=today.year - 65)
-        assert calculate_age(dob, today=today) == 65
-
-    def test_newborn(self):
-        """A newborn is 0."""
-        today = date(2025, 6, 15)
-        dob = today
-        assert calculate_age(dob, today=today) == 0
-
-    def test_birthday_not_yet_this_year(self):
-        """Person born Dec 31, 2000 — on Jan 1, 2025 they are still 24."""
-        dob = date(2000, 12, 31)
-        assert calculate_age(dob, today=date(2025, 1, 1)) == 24
-
-    def test_birthday_today(self):
-        """Person born Dec 31, 2000 — on Dec 31, 2024 they turn 24."""
-        dob = date(2000, 12, 31)
-        assert calculate_age(dob, today=date(2024, 12, 31)) == 24
-
-
-class TestRegisterRequestDateValidation:
-    """Tests for the RegisterRequest date_of_birth field validator."""
-
-    @pytest.mark.parametrize(
-        "dob",
-        [
-            "2000-01-01",
-            "1990-12-31",
-            "2004-02-29",  # leap day
-        ],
-    )
-    def test_valid_dates(self, dob):
-        """Valid YYYY-MM-DD dates pass validation."""
-        from app.routers.auth import RegisterRequest
-
-        req = RegisterRequest(username="u", email="u@example.com", password="p", date_of_birth=dob)
-        assert req.date_of_birth == dob
-
-    @pytest.mark.parametrize(
-        "dob",
-        [
-            "01-01-2000",
-            "2000/01/01",
-            "2000-13-01",
-            "2000-00-10",
-            "not-a-date",
-            "",
-        ],
-    )
-    def test_invalid_dates_rejected(self, dob):
-        """Invalid dates are rejected by the validator."""
-        from app.routers.auth import RegisterRequest
-
-        with pytest.raises(ValueError):
-            RegisterRequest(username="u", email="u@example.com", password="p", date_of_birth=dob)

backend/tests/test_rate_limit.py

@@ -0,0 +1,193 @@
+"""Tests for the Redis-backed token-bucket rate limiter.
+
+These tests use fakeredis (Lua enabled via lupa) so no live Redis server is
+required.
+"""
+
+import asyncio
+import sys
+
+import pytest
+from fastapi import Depends, FastAPI, Request
+from fastapi.testclient import TestClient
+
+sys.path.insert(0, "/workspace")
+
+import fakeredis.aioredis
+from redis.asyncio import Redis
+from redis.exceptions import ConnectionError
+
+from app.rate_limit import (
+    RateLimit,
+    RateLimitConfig,
+    RateLimiter,
+    default_user_scope,
+    named_limit,
+    rate_limit,
+)
+
+
+@pytest.fixture()
+def fake_redis():
+    return fakeredis.aioredis.FakeRedis(decode_responses=True)
+
+
+def make_request(client_ip: str = "127.0.0.1") -> Request:
+    scope = {
+        "type": "http",
+        "method": "GET",
+        "path": "/",
+        "headers": [],
+        "client": (client_ip, 12345),
+    }
+    return Request(scope)
+
+
+@pytest.mark.asyncio
+async def test_token_bucket_allows_burst_then_exhausts(fake_redis):
+    limiter = RateLimiter(
+        RateLimitConfig(
+            limit=RateLimit(name="burst", capacity=3, refill_rate=0.1),
+        ),
+        redis=fake_redis,
+    )
+    request = make_request()
+
+    results = []
+    for _ in range(4):
+        result = await limiter.check(request, user_id=42)
+        results.append(result)
+
+    assert [r.allowed for r in results] == [True, True, True, False]
+    assert results[0].remaining == 2
+    assert results[1].remaining == 1
+    assert results[2].remaining == 0
+    assert results[3].retry_after > 0
+
+
+@pytest.mark.asyncio
+async def test_retry_after_reflects_refill_rate(fake_redis):
+    limiter = RateLimiter(
+        RateLimitConfig(
+            limit=RateLimit(name="slow", capacity=1, refill_rate=0.5),
+        ),
+        redis=fake_redis,
+    )
+    request = make_request()
+
+    assert (await limiter.check(request, user_id=7)).allowed is True
+    denied = await limiter.check(request, user_id=7)
+    assert denied.allowed is False
+    # 1 token / 0.5 tokens per second = 2 seconds.
+    assert denied.retry_after == 2
+
+
+@pytest.mark.asyncio
+async def test_separate_scopes_get_separate_buckets(fake_redis):
+    limiter = RateLimiter(
+        RateLimitConfig(
+            limit=RateLimit(name="scoped", capacity=1, refill_rate=0.1),
+        ),
+        redis=fake_redis,
+    )
+    request = make_request()
+
+    assert (await limiter.check(request, user_id=1)).allowed is True
+    assert (await limiter.check(request, user_id=2)).allowed is True
+    assert (await limiter.check(request, user_id=1)).allowed is False
+
+
+@pytest.mark.asyncio
+async def test_fail_open_returns_allowed_when_redis_unavailable():
+    class BrokenRedis(Redis):
+        async def eval(self, *args, **kwargs):
+            raise ConnectionError("down")
+
+    limiter = RateLimiter(
+        RateLimitConfig(
+            limit=RateLimit(name="open", capacity=1, refill_rate=0.1, fail_open=True),
+        ),
+        redis=BrokenRedis(),
+    )
+    result = await limiter.check(make_request(), user_id=1)
+    assert result.allowed is True
+
+
+@pytest.mark.asyncio
+async def test_fail_closed_returns_denied_with_negative_retry_after():
+    class BrokenRedis(Redis):
+        async def eval(self, *args, **kwargs):
+            raise ConnectionError("down")
+
+    limiter = RateLimiter(
+        RateLimitConfig(
+            limit=RateLimit(name="closed", capacity=1, refill_rate=0.1, fail_open=False),
+        ),
+        redis=BrokenRedis(),
+    )
+    result = await limiter.check(make_request(), user_id=1)
+    assert result.allowed is False
+    assert result.retry_after == -1
+
+
+@pytest.mark.asyncio
+async def test_fail_closed_call_raises_503(fake_redis):
+    class BrokenRedis(Redis):
+        async def eval(self, *args, **kwargs):
+            raise ConnectionError("down")
+
+    limiter = RateLimiter(
+        RateLimitConfig(
+            limit=RateLimit(name="closed", capacity=1, refill_rate=0.1, fail_open=False),
+        ),
+        redis=BrokenRedis(),
+    )
+    with pytest.raises(Exception) as exc_info:
+        await limiter(make_request(), user_id=1)
+    assert exc_info.value.status_code == 503
+
+
+def test_named_limit_builds_expected_config():
+    config = named_limit(
+        "chat_send",
+        capacity=5,
+        refill_rate=0.2,
+        fail_open=False,
+    )
+    assert config.limit.name == "chat_send"
+    assert config.limit.capacity == 5
+    assert config.limit.refill_rate == 0.2
+    assert config.limit.fail_open is False
+    assert config.scope is default_user_scope
+
+
+def test_invalid_limit_definition_rejected():
+    with pytest.raises(ValueError):
+        named_limit("bad", capacity=0, refill_rate=1.0)
+    with pytest.raises(ValueError):
+        named_limit("bad", capacity=1, refill_rate=0)
+
+
+def test_dependency_integration_returns_429_with_retry_after():
+    async def fake_user_id() -> int:
+        return 99
+
+    chat_limit = rate_limit(
+        named_limit("chat_send_test", capacity=2, refill_rate=0.1, fail_open=False),
+        user_id_dependency=Depends(fake_user_id),
+    )
+
+    app = FastAPI()
+
+    @app.get("/probe")
+    async def probe(_: None = Depends(chat_limit)):
+        return {"ok": True}
+
+    client = TestClient(app)
+    assert client.get("/probe").status_code == 200
+    assert client.get("/probe").status_code == 200
+
+    denied = client.get("/probe")
+    assert denied.status_code == 429
+    assert denied.headers["retry-after"]
+    assert denied.json() == {"detail": "Rate limit exceeded"}