Skip to content
BaysSystems

Automations/N8N / AI AUTOMATION

Personal Brand AI Engine

An AI content and publishing workflow built around editorial memory, semantic anti-repetition, AI content generation, image generation, automated branding, scheduling, LinkedIn publishing and history persistence. It runs in your own n8n instance rather than behind someone else's SaaS.

Category
n8n / AI Automation
Runs in
Your own n8n instance
Publishes to
LinkedIn
Editions
3
Live

Most AI content automations are a prompt on a timer. They post confidently, repeat themselves within a fortnight, and have no idea what they already said.

The problem

A publishing workflow that cannot remember is not an assistant, it is a generator. Without an editorial memory it re-treads the same three ideas; without a same-day guard it double-posts; without persistence it starts from zero on every run. The hard part of automated publishing is not generating text — it is everything around the generation step.

What it does

  • Reads its own publishing history before it writes anything new
  • Compares candidate topics against what has already been published and rejects semantic near-repeats
  • Refuses to publish twice in the same day, by design rather than by luck
  • Generates the post, formats it for LinkedIn, and derives a visual direction from the content itself
  • Produces a branded image and publishes through the LinkedIn API
  • Writes the result back to history so the next run knows it happened

System/EXECUTION PATH

How the run happens.

Read before written, guarded before generated, persisted after published. The expensive steps never run for output that would be rejected anyway.

Pipeline

00 / 11 STAGES

  1. Recall

    1. Schedule / Manual Trigger

      Scheduled run, or fired by hand.

    2. Content History

      Prior output is loaded before anything is written.

    3. Editorial Memory

      History becomes working context for this run.

    4. Same-Day Guard

      Fails closed if today has already been published.

  2. Generate

    1. AI Content

      The post is generated against editorial memory.

    2. Platform Formatting

      Shaped for LinkedIn, the platform it publishes to.

  3. Compose

    1. Visual Direction

      An image brief is derived from the content itself.

    2. Image Generation

      The brief drives generation, not a random prompt.

    3. Brand Renderer

      Output is composed into a consistent branded frame.

  4. Publish

    1. LinkedIn API

      Published through the platform API.

    2. History Persistence

      The run is written back. The loop closes.

System architecture

The workflow is built as a linear pipeline with explicit state at both ends: history is read before generation and written after publication. Guards sit between recall and generation so the expensive steps never run for output that would be rejected anyway. Each stage is a discrete n8n node boundary, which is what makes a swap — a different model, a different image generator — a change at one boundary rather than a rewrite. The published workflow publishes to LinkedIn; other destinations are an extension point, not something it does today.

Current scope. The published workflow publishes to LinkedIn. Its stage boundaries are designed so another destination can be added, but that is extensibility rather than current behaviour — nothing here claims multi-platform publishing today.

Capabilities

What the system actually holds.

  • 01

    Editorial memory

    Published content is persisted and read back as context on every run. The workflow's knowledge of itself is durable state, not something reconstructed from a prompt each time.

  • 02

    Semantic anti-repetition

    Candidate topics are checked against prior output semantically rather than by string match, so a rephrased version of last week's post is caught instead of shipped.

  • 03

    Same-day guard

    An explicit gate prevents a second publish inside the same day. Scheduling mistakes and manual re-runs fail closed rather than producing duplicate posts.

  • 04

    Content and visual generation

    Copy is generated, formatted for LinkedIn, and given a visual direction derived from the content — which then drives image generation rather than a random illustration.

  • 05

    Automated branding

    A brand renderer composes the generated image into a consistent branded output, so visual identity is applied by the system instead of by hand each time.

  • 06

    History persistence

    Every completed run is written back to the content history. The loop closes, and the next run is better informed than the last.

Who it is for

  • Operators and founders who publish consistently and want the pipeline to hold the editorial memory instead of their head

  • Technical teams who would rather own the workflow in their own n8n instance than rent it from a SaaS

  • Engineers who want a real reference architecture for an AI pipeline with memory, guards and persistence — not a demo prompt chain

What you need to run it

  • n8n (self-hosted or cloud)
  • An LLM provider of your choice
  • An image generation provider
  • LinkedIn API credentials
  • A persistence target for content history

Credentials and provider accounts stay yours. Nothing routes through BaysSystems infrastructure.

Editions/AVAILABILITY

What you can get today.

Editions that are not finished carry no price and no checkout. When there is something to buy, it will appear here with both.

  • 01

    Open Source Edition

    Open source

    The sanitised workflow, published publicly. Import it into your own n8n instance, connect your own credentials and run it.

    • Importable workflow definition
    • The full pipeline: memory, guards, generation, branding, publishing
    • Runs entirely in your own n8n instance
    • No BaysSystems account or key required
    View the repository

    Free and open source. Bring your own model, image and LinkedIn credentials.

  • 02

    Pro Edition

    Coming soon

    An extended edition is being prepared. It is not finished, not priced and not for sale yet — and this page will not pretend otherwise.

    • Scope is still being defined
    • No price has been set
    • No purchase path exists yet

    Not available yet

    Coming soon. There is nothing to buy today; when there is, it will appear here with a real price and a real checkout.

  • 03

    Custom Implementation

    Scoped engagement

    We adapt the engine to your stack, your editorial rules, your destination platforms and your brand system, and hand it over running.

    • Adapted to your models, providers and destinations
    • Your editorial rules and brand system encoded into the pipeline
    • Deployed into your own infrastructure
    • Handed over with the architecture explained, not as a black box
    Start a conversation

    Scoped per engagement. Start with a conversation about the objective.

Continue

Implementation//SCOPED WORK

Want this running on your stack?

A custom implementation adapts the engine to your models, your editorial rules and your destinations — and is handed over explained, not as a black box.

Start a Project

Direct

Bring the objective rather than the feature list — the system design follows from it.