:hedgehog: PostHog is the leading platform for building self-driving products. Our developer tools – AI observability, analytics, session replay, flags, experiments, error tracking, logs, and more – capture all the context agents need to diagnose problems, uncover opportunities, and ship fixes. Steer it all from Slack, web, desktop, or the MCP.
Python ab-testingai-analyticsanalyticscdpdata-warehouseexperimentsfeature-flagsjavascriptproduct-analyticspythonreactsession-replaysurveystypescriptweb-analytics
import { MessageKey } from '~/common/kafka/producer'
import { logger } from '~/common/utils/logger'
import { IngestionOutput } from './ingestion-output'
import { IngestionOutputMessage } from './types'
/**
* Immutable container of resolved ingestion outputs.
*
* Each output maps a typed name (e.g. `'events'`, `'heatmaps'`) to an `IngestionOutput`.
* Pipeline steps use `produce()` and `queueMessages()` to send messages by output name —
* they never access the underlying producer or topic directly.
*
* @see `IngestionOutputsBuilder` for constructing instances from config.
*/
export class IngestionOutputs<O extends string> {
constructor(private outputs: Record<O, IngestionOutput>) {}
async produce(output: O, message: IngestionOutputMessage & { key: MessageKey }): Promise<void> {
await this.outputs[output].produce(message)
}
async queueMessages(output: O, messages: IngestionOutputMessage[]): Promise<void> {
await this.outputs[output].queueMessages(messages)
}
/**
* Check that all producers can reach their brokers.
*
* @param timeoutMs - Timeout for each broker connectivity check.
* @returns Output names whose producers failed the check.
*/
async checkHealth(timeoutMs = 5000): Promise<string[]> {
const failures: string[] = []
for (const outputName in this.outputs) {
try {
await this.outputs[outputName].checkHealth(timeoutMs)
} catch {
failures.push(outputName)
}
}
if (failures.length > 0) {
logger.error('🔴', `Health check failed for outputs: ${failures.join(', ')}`)
}
return failures
}
/**
* Check that all non-empty topics exist on their brokers.
*
* @param timeoutMs - Timeout for each topic metadata check.
* @returns Output names whose topics failed the check.
*/
async checkTopics(timeoutMs = 10000): Promise<string[]> {
const failures: string[] = []
for (const outputName in this.outputs) {
try {
await this.outputs[outputName].checkTopicExists(timeoutMs)
} catch {
failures.push(outputName)
}
}
if (failures.length > 0) {
logger.error('🔴', `Topic check failed for outputs: ${failures.join(', ')}`)
}
return failures
}
}
/**
* Scope owner for an `IngestionOutputs` derived from already-running
* infrastructure (typically a shared Kafka producer registry). `start()`
* resolves the outputs via the `build` callback and verifies that every
* output's topic is reachable; `stop()` is a no-op because the producer
* registry's own component owns the connection lifetimes.
*/
export class IngestionOutputsComponent<O extends string> {
constructor(private readonly build: () => IngestionOutputs<O>) {}
async start(): Promise<{ value: IngestionOutputs<O>; stop: () => Promise<void> }> {
const outputs = this.build()
const failures = await outputs.checkTopics()
if (failures.length > 0) {
throw new Error(`Output topic verification failed for: ${failures.join(', ')}`)
}
return { value: outputs, stop: () => Promise.resolve() }
}
}