Package org.openidentityplatform.openig.ai.filter

Class TokenRateLimiter

java.lang.Object

org.openidentityplatform.openig.ai.filter.TokenRateLimiter

public class TokenRateLimiter
extends Object

Identity-aware, in-memory token-bucket rate limiter for LLM prompt tokens.

Configuration

   "rate": {
     "numberOfTokens": 10000,
     "duration": "1 minute"
   }
 

Constructor Summary

Constructors

Constructor	Description
TokenRateLimiter(long numberOfTokens, org.forgerock.util.time.Duration duration)	Minimal constructor, with system ticker and no bucket eviction
TokenRateLimiter(long numberOfTokens, org.forgerock.util.time.Duration duration, com.github.benmanes.caffeine.cache.Ticker ticker, org.forgerock.util.time.Duration cleaningInterval)	Full constructor, matching the pattern of TokenBucketThrottlingStrategy(Ticker, ScheduledExecutorService, Duration).

Method Summary

Modifier and Type	Method	Description
long	availableTokens(String sub)
void	stop()
long	tryConsume(String sub, long cost)	Try to consume N tokens (e.g. costly endpoints) for identity

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

TokenRateLimiter

public TokenRateLimiter(long numberOfTokens,
                        org.forgerock.util.time.Duration duration)

Minimal constructor, with system ticker and no bucket eviction

Parameters:

numberOfTokens - maximum tokens per window (= burst capacity)

duration - window size — OpenIG Duration syntax

TokenRateLimiter

public TokenRateLimiter(long numberOfTokens,
                        org.forgerock.util.time.Duration duration,
                        com.github.benmanes.caffeine.cache.Ticker ticker,
                        org.forgerock.util.time.Duration cleaningInterval)

Full constructor, matching the pattern of TokenBucketThrottlingStrategy(Ticker, ScheduledExecutorService, Duration).

When cleaningInterval is non-null, a periodic eviction task removes buckets that have been idle for at least one full window, preventing unbounded heap growth in long-running deployments.

Parameters:

numberOfTokens - maximum tokens per window (= burst capacity)

duration - window size — OpenIG Duration syntax

ticker - monotonic time source; use Ticker.systemTicker() in production and FakeTicker in tests

cleaningInterval - eviction period; ignored when executor is null

Method Detail

tryConsume

public long tryConsume(String sub,
                       long cost)

Try to consume N tokens (e.g. costly endpoints) for identity

availableTokens

public long availableTokens(String sub)

stop

public void stop()