Skip to content
James Foss, jamesfoss.me home
P.06In development· Running, actively expanding

AI Content Automation

An automated workflow system for content generation, publishing, and multi-platform distribution.

Category
Automation pipelines
Status
In development
Core stack
AI Automation · TypeScript · Node.js · REST APIs · Application Integrations
Fig. 06 — Illustrative cover artwork, not a product screenshot
01

Overview

AI Content Automation is a set of pipelines that run alongside LWS Dashboard. Language models draft long-form content through the dashboard's provider layer, and the pipelines turn finished material such as blog posts, tips, and quotes into captions, branded cards, and short-form video published on a schedule.

Captions and hashtags come from deterministic templates, so every post stays on-brand. Each pipeline runs as its own worker with its own settings, scheduling, and verification, so a failure in one does not stall the others.

  • 01Blog campaigns that select posts and render promo cards and short videos
  • 02Template-driven reel rendering assembled with ffmpeg
  • 03Publishing to Facebook Reels and YouTube with generated metadata and thumbnails
  • 04Separate worker processes with scheduling and automated tests
02

Objective

Keep a steady, multi-platform publishing cadence for the agency without manual production work.

03

Technical approach

  1. A.

    A pipeline per format

    Blog campaigns, tip reels, and quote reels are separate pipelines built from shared steps: selection, captioning, rendering, publishing, and verification.

  2. B.

    Rendering on the server

    Cards are rendered from JSX templates to images with Satori, then assembled into vertical video with ffmpeg.

  3. C.

    Platform-native publishing

    Each destination gets native metadata: captions and hashtags for Facebook, and titles, descriptions, and thumbnails for YouTube.

  4. D.

    Models behind an interface

    Language model calls go through a provider layer, so the model behind a task can change without touching the pipelines that use it.

04

Architecture

  1. Sources

    • Blog posts
    • Tip and quote libraries

  2. Generation

    • Model provider layer
    • Caption templates
    • JSX card templates

  3. Rendering

    • Satori to PNG
    • ffmpeg video assembly

  4. Distribution

    • Facebook Pages and Reels
    • YouTube
    • Job scheduling

  5. Runtime

    • PM2 workers
    • PostgreSQL job state
05

Technologies

  • AI Automation
  • TypeScript
  • Node.js
  • REST APIs
  • Application Integrations
  • Database-driven Applications
  • Satori
  • ffmpeg
  • Linux
06

Visuals

Private system. Output samples and screenshots will be added once approved.

The artwork above is an illustration of the system, not an interface capture.

07

Challenges & solutions

  1. Challenge 01

    Knowing whether a published reel actually finished processing on the platform.

    Solution

    A verification step checks published reels and records diagnostics, so failures surface in the dashboard instead of going unnoticed.

Have a system like this in mind?

Get in touch