Rate Limiting & Observability¶

Rate Limiting¶

fast-agent-stack provides Redis-backed fixed-window rate limiting via fastapi-redis-sdk.

pip install "fast-agent-stack[rate-limit]"

Enable Rate Limiting¶

class Settings(BaseSettings):
    include_rate_limit: bool = True
    rate_limit_requests: int = 100    # max requests per window
    rate_limit_period: int = 60       # window size in seconds
    redis_url: str = "redis://localhost:6379"

Rate limiting is applied globally to all routes when enabled. The limit is per client IP, enforced via a Redis Lua script (atomic INCR + EXPIRE).

Response Headers¶

Clients receive standard rate-limit headers on every response:

X-RateLimit-Limit: 100
X-RateLimit-Remaining: 87
X-RateLimit-Reset: 1751836800

When the limit is exceeded, the server returns 429 Too Many Requests.

Observability (OpenTelemetry)¶

fast-agent-stack auto-instruments with OpenTelemetry. Every request generates a trace span with HTTP metadata.

pip install "fast-agent-stack[tracing]"

Enable Tracing¶

class Settings(BaseSettings):
    tracing_enabled: bool = True
    otel_exporter_endpoint: str = "http://localhost:4317"   # OTLP gRPC

Jaeger (default backend)¶

docker run -p 16686:16686 -p 4317:4317 jaegertracing/all-in-one

Open http://localhost:16686 to view traces.

Custom Exporter¶

Set otel_exporter_endpoint to any OTLP-compatible backend (Grafana Tempo, Honeycomb, Datadog OTLP endpoint, etc.).

What Gets Traced¶

Every HTTP request (method, path, status code, duration)
Database queries (SQLAlchemy instrumentation)
External service calls via the tracer context manager

Adding Custom Spans¶

from opentelemetry import trace

tracer = trace.get_tracer(__name__)

async def my_handler():
    with tracer.start_as_current_span("my-operation") as span:
        span.set_attribute("user.id", user_id)
        result = await do_work()
    return result