sampaiodev.
PTEN
Own productIn progress

OrkestraAI

Started from n8n workflows and is evolving into a multi-tenant SaaS with AI agents, RAG, queues and product UI.

React 19HonoNestJSBullMQRedispgvectorCloudflareDokploy

Technical reading

Quick case read

Problem

Replies in comments and DMs require consistency, context and operations. Manual workflow scales poorly and an agent without queue/log becomes a black box.

Solution

The proposal is to turn n8n automations into a product: onboarding, webhooks, BullMQ/Redis queues, Docker API, Cloudflare gateway and RAG base with pgvector.

Stack

React 19 · Hono · NestJS · BullMQ

Status

In progress

Architecture

Workflow becomes product with queue, context and operations

Confirmed public state: landing live at orkestraia.com and app in development. Private details stay off the page.

Technical flow

  • Input
  • Processing
  • Queue
  • Storage
  • Output
Flowchart
Rendering flowchart…

Step-by-step flow

01Input
Instagram/OAuth/webhooksComments, DMs and authorized events
02Edge
Gateway/BFFOAuth, webhooks and normalization
03Processing
Product APIProduct rules and tenancy
04Queue
Redis/BullMQAsync processing of automations
05AI
AI agents/RAGContext, vector base and response generation
06Output
Comments/DMsReply sent or prepared for sending
07Product
Panel/landingLanding live; app in development
Diagram source (Mermaid)
Paste into any Mermaid renderer
flowchart LR
  instagram["Instagram/OAuth/webhooks"]
  gateway["Gateway/BFF"]
  api["Product API"]
  queue["Redis/BullMQ"]
  rag["AI agents/RAG"]
  reply["Comments/DMs"]
  panel["Panel/landing"]
  instagram --> gateway
  gateway --> api
  api --> queue
  queue --> rag
  rag --> reply
  reply --> panel
1. Instagram/OAuth/webhooks (Input) → Gateway/BFF. 2. Gateway/BFF (Edge) → Product API. 3. Product API (Processing) → Redis/BullMQ. 4. Redis/BullMQ (Queue) → AI agents/RAG. 5. AI agents/RAG (AI) → Comments/DMs. 6. Comments/DMs (Output) → Panel/landing. 7. Panel/landing (Product).

Stack and responsibilities

  • React 19 + Vite

    Product interface in development.

  • Hono / Cloudflare Worker

    Gateway for OAuth, webhooks and edge.

  • NestJS/Fastify

    Main API in Docker.

  • BullMQ + Redis

    Automation queue and async processing.

  • Postgres + pgvector

    Product database and RAG base.

Technical decisions

  • Product separate from service

    OrkestraAI is SaaS; Maestro/n8n done-for-you stays separate in public communication.

  • Queue before agent

    Message automation needs operational control, not direct calls without traceability.

  • Private repo

    There is no public code link in this phase; only confirmed public landing.

Operations

  • documentedLanding

    Live at `https://orkestraia.com`.

  • in developmentApp

    In development, without public production claim.

  • not publicCode

    Private repository; no link exposed.

Status and next steps

  • Finish the app before publishing sensitive technical details.
  • Keep the landing as the honest public product entrypoint.

Contact

Want something similar for your process?

If there is repeated input, a manual step and an expected output, there is probably a system to design.

Talk about a project