Fast SEO Platform
Intelligent SEO content automation: from strategy to publishing, with AI and human oversight
The aerial surveying business leans heavily on local SEO to attract clients.
Every completed project is a potential case study. But the technical team has no time to write, design images, optimize SEO, and publish each article manually. The problem had multiple layers: generating content that sounds natural, automated illustration without generic stock, on-page optimization, AI hallucination detection, publication queue management, and human oversight without reviewing every line.
- 01
Architecture
Monorepo with pnpm workspaces: content pipeline (bot), operations interface (dashboard), conversational assistant (planner), and MCP server. Each module independent with shared types.
- 02
Content Pipeline
5 chained stages: writer (generates the post), illustrator (generates images with Imagen 3), linker (adds internal links), QA (validates quality), and publisher (uploads to Payload CMS). Each stage atomic and transactional.
- 03
Operations Dashboard
Next.js interface with a publication queue, real-time monitoring, AI provider settings with encrypted credentials (AES-256-GCM), and a prompt editor.
- 04
Conversational Planner
AI assistant with function calling (12 tools) to manage content via chat. We evolved from MCP to function calling after evaluating with 7 experts.
- 05
Hardening
Security audit with 18 findings remediated (3 critical, 6 high), full TypeScript migration, 800+ tests, and Turborepo build optimization (30s down to 310ms).
Automate the entire pipeline while keeping editorial quality and brand consistency.
Monorepo with 4 apps and 2 shared packages.
5-stage atomic pipeline: writer, imager (Imagen 3), linker, QA, publisher (Payload CMS).
Next.js dashboard with publication queue, real-time monitoring, encrypted credentials (AES-256-GCM).
Conversational planner with 12 function-calling tools.
800+ tests, security audit with 18 findings remediated, builds optimized from 30s to 310ms with Turborepo.
Complete SEO content automation platform with a 5-stage pipeline, real-time operations dashboard, conversational planner with 12 tools, 800+ tests, 18 security findings remediated, and builds optimized from 30s to 310ms.
Monorepo Architecture
Structure of 4 apps + 2 shared packages with centralized types and validation.
packages: - apps/dashboard # Next.js 16, 7 rutas - packages/bot # Pipeline de contenido - packages/shared # Tipos, validación Zod - packages/mcp # Herramientas externas
Content Pipeline
5-stage flow: writer → imager → linker → QA → publisher. Each stage atomic and transactional.
type PipelineStage = "writing" | "illustrating" | "linking" | "qa" | "publishing";
async function runPipeline(post: QueueItem) {
for (const stage of stages) {
const result = await executeStage(stage, post);
if (!result.ok) {
await rollbackPartial(post, stage);
return { ok: false, error: result.error };
}
}
return { ok: true, value: post };
}Planner Function Calling
12 Zod-validated tools for managing content via chat with the operator.
const tools = [
{
name: "schedule_post",
description: "Programa un post para publicación",
parameters: z.object({
title: z.string().min(1).max(200),
surface: z.enum(["blog", "case-study"]),
scheduled_date: z.string().datetime(),
}),
},
];Interested in building something like this?
[Start a Project]