Evenfire is an intelligence layer platform for running LLM agents that take real actions, on your infrastructure
TypeScript
let registryIdentityCacheGeneration = 0
export class RegistryIdentityChangedError extends Error {
constructor() {
super('registry_identity_changed')
this.name = 'RegistryIdentityChangedError'
}
}
export function invalidateRegistryIdentityCaches(): void {
registryIdentityCacheGeneration += 1
}
export function getRegistryIdentityCacheGeneration(): number {
return registryIdentityCacheGeneration
}
export function isRegistryIdentityCacheGenerationCurrent(generation: number): boolean {
return registryIdentityCacheGeneration === generation
}
export async function withCurrentRegistryIdentity<T>(
read: (generation: number) => Promise<T>,
opts: { staleError?: () => Error } = {}
): Promise<T> {
for (let attempt = 0; attempt < 2; attempt += 1) {
const generation = getRegistryIdentityCacheGeneration()
try {
const value = await read(generation)
if (isRegistryIdentityCacheGenerationCurrent(generation)) return value
} catch (err) {
if (isRegistryIdentityCacheGenerationCurrent(generation)) throw err
}
}
throw opts.staleError?.() ?? new RegistryIdentityChangedError()
}
export function __resetRegistryIdentityCacheGenerationForTests(): void {
registryIdentityCacheGeneration = 0
}