ElementX2 · Applied AI systems · Live
ElementX2[email protected]
Index W4Workshop build · SaaS product

PostBear — a month of social content from one URL

Most small businesses know they should post consistently and simply don’t have anyone whose job that is. PostBear turns one input — the business’s website URL — into a month of on-brand social content, rendered and published automatically.

Laravel · React · OpenRouter · Meta Graph API · PostgreSQL

Ref 01 · The problem

The problem

  • Consistent social presence is a real commercial signal, but producing it is a daily creative chore with no owner in most SMEs.
  • Generic AI content is worse than nothing — output has to sound like the business, grounded in what the business actually says about itself.
  • Scheduling posts is easy; publishing them reliably through Meta’s APIs, with tokens, quotas and retries, is the hard part nobody sees.

Ref 02 · What we built

What we built

01Business URL02Brand profile03Content plan04Render05PublishROUTE · SURVEYED
The content route
01

From URL to brand profile

The pipeline scrapes the business site with Firecrawl and derives a brand profile — voice, offerings, audience — via LLM, with a grounding guard that constrains generation to what the source material supports rather than letting the model invent services.

02

A month of content, rendered properly

From the profile it plans and writes a month of posts (12–28 depending on chosen frequency) and renders carousel graphics server-side through a dedicated render service, so output is imagery, not just captions.

03

Real automated publishing

A scheduler publishes due posts to Facebook and Instagram through the Meta Graph API every minute, with atomic job claiming and retry classification so a flaky API call can’t double-post. Meta system-user tokens are held encrypted.

04

The product scaffolding

Consent flows for using a real person’s likeness in generated imagery — with an audit trail — plus subscription billing and plan quotas via Paddle, and a test suite in the hundreds keeping the pipeline honest.

Ref 03 · Stack

Stack

  • Laravel
  • React
  • OpenRouter
  • Meta Graph API
  • PostgreSQL

Ref 04 · What it demonstrates

What it demonstrates

  • A complete AI content pipeline: ingest → understand → generate → render → publish
  • Grounded generation — engineering against hallucinated marketing claims
  • Production discipline around third-party APIs: idempotency, retries, encrypted credentials