telega/idempotency

Webhook/update idempotency: drop updates Telegram re-delivers.

Telegram retries an update (same update_id) when it does not receive a 200 in time — on a slow response, a redeploy, or a network blip. Without deduplication that means a command runs twice, which is a real problem for non-idempotent actions (sending an invoice, charging Stars, granting a reward).

deduplicate builds a pre-router middleware that remembers each update_id in a KeyValueStorage for a TTL window and stops any update whose id was already seen. Because pre-router middleware runs sequentially inside the single bot actor, the read-then-write below is race-free even under concurrent re-delivery.

import telega
import telega/idempotency
import telega/storage/ets

let assert Ok(store) = ets.new(name: "telega_dedup")

telega.new(token:, url:, webhook_path:, secret_token:)
|> telega.use_pre_handler(idempotency.deduplicate(
  storage: store,
  // Keep ids for an hour — comfortably longer than Telegram's retry window.
  ttl_ms: 3600_000,
))
|> telega.with_router(router)
|> telega.init()

Use a persistent backend (Postgres/SQLite/Redis) when you run more than one bot node or want dedup to survive a restart — the in-memory ets store is per-node and cleared on VM restart.

pub fn deduplicate(
  storage storage: storage.KeyValueStorage(storage_error),
  ttl_ms ttl_ms: Int,
) -> fn(bot.PreContext(dependencies)) -> bot.PreRouterResult