: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 { ProcessedPluginEvent, RetryError } from '~/plugin-scaffold'
import { LegacyDestinationPluginMeta } from '../../types'
export interface ReplicatorMetaInput {
config: {
host: string
project_api_key: string
replication: string
events_to_ignore: string
disable_geoip: 'Yes' | 'No'
}
}
type StrippedEvent = Omit<ProcessedPluginEvent, 'team_id' | 'ip' | 'person'>
const reverseAutocaptureEvent = (autocaptureEvent: StrippedEvent) => {
const { elements, properties, ...event } = autocaptureEvent
const $elements = elements?.map((el) => {
// $el_text and attributes are the only differently named parts
const { attributes, text, ...commonProps } = el
return {
...commonProps,
$el_text: text,
...attributes,
}
})
return {
...event,
properties: $elements
? {
...properties,
$elements,
}
: properties,
}
}
export const onEvent = async (
event: ProcessedPluginEvent,
{ config, fetch, logger }: LegacyDestinationPluginMeta
): Promise<void> => {
const replication = parseInt(config.replication) || 1
if (replication > 1) {
// This is a quick fix to make sure we don't become a spam bot
throw Error('Replication factor > 1 is not allowed')
}
const eventsToIgnore = new Set(
config.events_to_ignore && config.events_to_ignore.trim() !== ''
? config.events_to_ignore.split(',').map((event: string) => event.trim())
: null
)
if (eventsToIgnore.has(event.event)) {
return
}
const { team_id, ...sendableEvent } = { ...event, token: config.project_api_key }
if (config.disable_geoip === 'Yes') {
sendableEvent.properties.$geoip_disable = true
}
const finalSendableEvent =
sendableEvent.event === '$autocapture' ? reverseAutocaptureEvent(sendableEvent) : sendableEvent
const batch = []
for (let i = 0; i < replication; i++) {
batch.push(finalSendableEvent)
}
if (batch.length > 0) {
const batchDescription = `${batch.length} event${batch.length > 1 ? 's' : ''}`
await fetch(`https://${config.host.replace(/\/$/, '')}/e`, {
method: 'POST',
body: JSON.stringify(batch),
headers: { 'Content-Type': 'application/json' },
// TODO: add a timeout signal to make sure we retry if capture is slow, instead of failing the export
}).then(
(res) => {
if (res.status >= 200 && res.status < 300) {
logger.log(`Flushed ${batchDescription} to ${config.host}`)
} else if (res.status >= 500) {
// Server error, retry the batch later
logger.error(
`Failed to submit ${batchDescription} to ${config.host} due to server error: ${res.status}`
)
throw new RetryError(`Server error: ${res.status}`)
} else {
// node-fetch handles 300s internaly, so we're left with 400s here: skip the batch and move forward
// We might have old events in ClickHouse that don't pass new stricter checks, don't fail the whole export if that happens
logger.warn(`Skipping ${batchDescription}, rejected by ${config.host}: ${res.status}`)
}
},
(err) => {
if (err.name === 'AbortError' || err.name === 'FetchError') {
// Network / timeout error, retry the batch later
// See https://github.com/node-fetch/node-fetch/blob/2.x/ERROR-HANDLING.md
logger.error(`Failed to submit ${batchDescription} to ${config.host} due to network error`, err)
throw new RetryError(`Target is unreachable: ${(err as Error).message}`)
}
// Other errors are rethrown to stop the export
logger.error(`Failed to submit ${batchDescription} to ${config.host} due to unexpected error`, err)
throw err
}
)
}
}