Configuration¶

Configuration may be supplied directly or through environment variables.

from riotskillissue import PlatformRoute, RiotClient, RiotClientConfig

config = RiotClientConfig(
    api_key="RGAPI-...",
    default_route=PlatformRoute.EUW1,
    read_timeout=20.0,
    max_retries=3,
    max_rate_limit_retries=2,
)

riot = RiotClient(config=config)

The primary environment variables are:

Variable	Purpose
`RIOT_API_KEY`	Riot development or production API key
`RIOT_DEFAULT_ROUTE`	Default platform, regional, or VALORANT route
`RIOT_CONNECT_TIMEOUT`	Connection timeout in seconds
`RIOT_READ_TIMEOUT`	Response-read timeout in seconds
`RIOT_WRITE_TIMEOUT`	Request-write timeout in seconds
`RIOT_POOL_TIMEOUT`	Connection-pool timeout in seconds
`RIOT_MAX_RETRIES`	Network and server retry limit
`RIOT_MAX_RATE_LIMIT_RETRIES`	HTTP 429 retry limit
`RIOT_CACHE_TTL`	Default cache lifetime
`RIOT_LOG_LEVEL`	Library log level

Empty API keys are rejected. RSO credentials use a token provider and are not accepted as endpoint arguments.

Network and server retries apply to idempotent HTTP methods, including GET, PUT, and DELETE. POST and PATCH retry connection failures that occur before the request is sent, but do not retry read/write failures or server errors that could have occurred after a mutation was applied. HTTP 429 responses use the separate rate-limit retry budget for every method.

Closing a client also closes the Redis rate limiter created from redis_url. If you pass a shared rate limiter directly to HttpClient, close that limiter yourself after all clients using it have closed.

Caching¶

Pass an AbstractCache implementation to either client. RSO operations are not cached unless explicitly enabled. If enabled, cache entries are partitioned by a non-reversible credential fingerprint.

from riotskillissue import MemoryCache, RiotClient

riot = RiotClient(cache=MemoryCache(max_size=1024))

The optional RedisCache stores its entries under riot:cache:. Its clear() method removes only entries in that namespace, preserving rate-limit counters and unrelated application data in the same database. Existing entries from versions using unprefixed cache keys expire according to their original TTL; the first request after upgrading repopulates the cache.