What we shipped, when, and why.

Meta deauthorize & data-deletion webhooks

LittlePoster.ai now handles the official Meta and Threads deauthorize and data-deletion callbacks. If a user revokes our app from their Meta or Threads account, the matching channels are removed automatically and a confirmation page is published at /legal/data-deletion.

• Page and Instagram webhook subscriptions now happen automatically at connect time.
• A new "Data deletion request status" page lets people verify completion using the confirmation code Meta returned to them.

Onboarding: explicit channel + slot setup

After signup, new workspaces now walk through a calmer four-step onboarding: pick channels, set posting slots, define a brand voice, and write the first draft. The whole thing takes about three minutes.

• You can skip any step and resume later from Settings.
• Brand voice samples are stored separately from posts and never used to train external models.

AI image generation in compose

The composer can now generate images via Gemini "Nano Banana Pro" — useful for placeholder visuals, A/B variants, and quick mockups. Generations count against your workspace AI quota; image generation is disabled when no Gemini API key is configured.

Worker on Node 22, hot-reload via tsx

The worker now loads its TypeScript source through tsx instead of ts-node, which fixes a long-running issue where some adapters silently re-exported stale modules after a hot reload. Production worker is on Node 22.

Threads: posting and webhook receiver

Threads is now a first-class network. You can connect a Threads account, post text + media, and receive comment webhooks for normalized inbox.

• Cross-post from a single composer draft to Threads + X + Bluesky.

Refresh sweep: fewer needs_reauth surprises

A background job now refreshes channel tokens 24 hours before they expire, instead of refreshing only at publish time. Channels stuck in needs_reauth dropped by 71% in our beta cohort over the first week.

Per-channel costs surfaced for X

X (Twitter) is pay-per-use under the v2 API. Each post now records the cents charged in PublishResult.metadata.costCents and rolls up into a monthly per-workspace total in Settings → Billing.

Compose: AI suggestions are debounced

Earlier, AI suggestions could fire on every keystroke if you typed quickly inside the per-channel variant fields. Suggestions are now debounced and respect a per-workspace monthly token cap.