← All work

Public build · Technical case study

Content production pipeline

A public multi-stage content system with research, ideation, channel-specific writing and Google Sheets records.

Node.jsGemini adaptersGoogle Sheets

PUBLIC IMPLEMENTATION / SOURCE WALKTHROUGH

Content production pipeline

Reviewed public source: separated research and analysis modules, integration adapters and a pipeline entry point.

  1. 01

    Collect source material through API and feed adapters

    Input and entry point

  2. 02

    Pass topic context into a separate ideation step

    Implementation stage

  3. 03

    Prepare channel-specific text through writing modules

    Implementation stage

  4. 04

    Record content and run information in Google Sheets

    Output and operational boundary

IMPLEMENTATION EXCERPT / agents/researcher.js

async function runResearcher(runLabel = 'morning') {
  // … source collection and topic selection
}

// Separate analysis stage: agents/analyst.js
async function runAnalyst() {
  // … inspect stored topic and hook records
}

Selected implementation lines with explanatory comments. Omitted work is marked with “…”. Inspect the complete source below.

Evidence boundary

A trend is not proof of customer demand. The existing analyst reasons from stored topics and hooks; it should not be presented as measured conversion or engagement analysis.

THE PROBLEM

Where the work gets stuck.

Content preparation involves gathering material, choosing an angle, adapting the format and keeping a record of what was produced.

Follow the implementation

  1. Collect source material through API and feed adapters
  2. Pass topic context into a separate ideation step
  3. Prepare channel-specific text through writing modules
  4. Record content and run information in Google Sheets

Technical scope

Source review covers the separate researcher, ideator, scripter and analyst modules, channel adapters and Sheets-backed run records.

The decision that matters

A trend is not proof of customer demand. The existing analyst reasons from stored topics and hooks; it should not be presented as measured conversion or engagement analysis.

Inspect the evidence

Public source reviewed October 1, 2026. Pinned implementation links preserve the version reviewed. This walkthrough does not reproduce private client details.

Have a similar repetitive task?

Share the input, the tools and the decision that should stay with a person.

Discuss your workflow ↗