We read every piece of feedback, and take your input very seriously.
To see all available qualifiers, see our documentation.
docs(distributed): document the best-effort Workers KV store (TK-812)
docs(operations): reference the stable metrics contract + recordDecisionOnSpan (TK-819/820)
docs(operations): document the analytics tap (tapDecisions) (TK-821)
docs(frameworks): document the NestJS @RateLimit decorator (TK-816) Show the ergonomic per-route form (@RateLimit + a global createRateLimitGuard via APP_GUARD) alongside the existing guard, incl. duration strings, a quota() route, and the class/handler override rule.
docs(migrating): full guides for express-rate-limit, rate-limiter-flexible, @upstash/ratelimit (TK-815) Each gets a concept-mapping table (options, methods, units) and the key behavioural shift — throw-on-deny → Decision, fixed-window → GCRA default, and (for Upstash) vendor-locked → same strategy on any backend. Adds 0.8.1 peek()/quota() to the recipes. API names cross-checked against the code (emit families, fromUpstash, trustProxy, peek).
docs(operations): add per-store outage/recovery table (TK-814) Condensed failure matrix + twoTier L2-outage behaviour, linking the full docs/FAILURE-MODES.md.
docs(performance): refresh benchmark numbers (2026-05-28, TK-813) Sync into the wiki the current figures from SCOREBOARD: checkSync ~186 ns/op (5.4M ops/s), async ~3.5M; async in-memory now edges past rate-limiter-flexible (express-rate-limit's bare counter ~1.4x faster); Postgres twoTier(leased) ~37x throughput win. Note the Docker-on-Windows latency caveat and the machine spec.
docs(getting-started): document peek & forecast introspection (TK-811) A non-consuming peek (current Decision) and forecast ({ spendableNow, nextReplenishAt, fullAt }) — works in-memory and over a distributed store via a read-only round trip that never runs the consuming Lua.
docs(strategies): document the quota() billing-period strategy (TK-810) Add quota to the strategy table plus a dedicated section: the calendar/ fixed/rolling cadences, the leap-correct civil-calendar boundary mirrored in Redis Lua, and the honest fixed-offset (non-DST) timezone caveat.
docs: document 0.8.0 stores, adapters, and distributedTokenBudget
Mark GALE Pillars 2-3 and TALE Layers 2-3 shipped (0.7.0) learnedReservation/predictiveReservation and leaseSizer/predictiveLeaseSizer now ship as first-class API. Research.md status markers flipped to Shipped + the shipped-surface paragraph updated; new admission-pacing section on the Overload page; new 'Sizing the lease' subsection on the Distributed page; Home breadth bullet notes the LLM cost-control stack.
Document tokenBudget + provability rebrand (0.6.0)
Add ThrottleKit wiki: guides, distributed model, research
Initialize wiki