Summary

Add optional accountId field to cron delivery configuration, allowing cron jobs to target a specific channel account for delivery.

Problem

When multiple bot accounts are configured on the same channel (e.g., a main Telegram bot and an alerts bot), cron job delivery always uses the account inferred from the session store (the last-used account). There is no way to explicitly route cron notifications to a different account.

For example, with two Telegram bots configured:

telegram:
  accounts:
    main:
      botToken: "..."
    alerts:
      botToken: "..."

A cron job with delivery: { mode: "announce", channel: "telegram", to: "123456" } would always deliver via whichever account was last used in the main session, with no way to force delivery through the alerts account.

Solution

Add an optional accountId field to the delivery config:

{
  "delivery": {
    "mode": "announce",
    "channel": "telegram",
    "to": "123456",
    "accountId": "alerts"
  }
}

When accountId is set, it overrides the account inferred by resolveDeliveryTarget() from the session store.

Changes

src/cron/types.ts

• Add accountId?: string to CronDelivery type (flows to CronDeliveryPatch via Partial<>)

src/gateway/protocol/schema/cron.ts

• Add accountId: Type.Optional(NonEmptyString) to CronDeliverySchema and CronDeliveryPatchSchema

src/cron/delivery.ts

• Add accountId?: string to CronDeliveryPlan type
• Extract accountId from delivery object in resolveCronDeliveryPlan()
• Include it in the returned plan when source is "delivery"

src/cron/isolated-agent/run.ts

• After resolveDeliveryTarget(), override resolvedDelivery.accountId when deliveryPlan.accountId is set
• This ensures the explicit config takes precedence over session-inferred account

src/cron/normalize.ts

• Preserve accountId through coerceDelivery() normalization (trim + validate like other string fields)

Backward Compatible

• accountId is optional everywhere — existing jobs without it behave exactly as before
• No migration needed for existing jobs.json files
• The field is simply ignored if the specified account doesn't exist (existing error handling in resolveOutboundTarget covers this)

Greptile Overview

Greptile Summary

• Adds an optional delivery.accountId field to cron job config/types and gateway schemas so users can select a specific channel account for outbound delivery.
• Normalizes/trims delivery.accountId in cron job input parsing.
• Threads the new field into the cron isolated agent delivery plan and overrides the resolved delivery target’s accountId before sending.
• Overall fits into the existing delivery-target resolution flow (session-inferred defaults + explicit overrides) but currently overrides the account after target validation.

Confidence Score: 3/5

• This PR is close to mergeable, but it changes delivery target selection in a way that can bypass existing target/account validation.
• The schema/type additions and normalization are straightforward, but overriding accountId after resolveDeliveryTarget() means the new accountId is not validated/docked via resolveOutboundTarget(), so misconfigured accountId may only fail later or behave inconsistently with existing error handling.
• src/cron/isolated-agent/run.ts; src/cron/isolated-agent/delivery-target.ts

_{(2/5) Greptile learns from your feedback when you react with thumbs up/down!}