Mukala — AI-powered shop operations

Live proof-of-life from the mukala-test node.

Live status

Loading…

Grok Orchestrator



Deployed


Architecture
01APIrequest comes in02Grok 4.3 Orchestratorpicks up the request, decides what to do03Workflow Enginereads the workflow definition from Postgres, executes steps in orderSTEP 1scan analyticsSTEP 2adjust tomorrow’s postsSTEP 3attach media via FFmpeg04Postgres 16updated by workflow steps · state, schedules, logs05Backlog queuework items extracted from conversations
A request comes in through the API. Grok 4.3 plans what needs to happen and hands it to the Workflow Engine, which runs each step in order against Postgres and internal tools like FFmpeg. Every step’s results and state are logged to Postgres, and follow-up work lands in the backlog queue.


How Mukala Works
01Shop ownerbrowser · app.mukala.ai / prod.mukala.ai02Cloudflarefronts every requestCDN / edge cachestatic assets at the edgeAccess sign-ingate for app.mukala.aiDNSapp via tunnel · prod direct03Mukala VM (this box)the brain and the databasenginx :443TLS, proxies to APIMukala APIevery request lands hereGrok 4.3orchestratorWorkflow Engineruns steps in orderPostgres 16shops · connectors · routines · workflowsFFmpegvideo / media prepWidget Stagepreviews & widgets04Every shop load pulls from Postgres on the VM.Workflows trigger connector calls (Typefully, ElevenLabs, Etsy, etc.) from the VM.
Cloudflare fronts everything: caching, sign-in and DNS. The VM is the brain and the database. The front end never talks to the database directly; every request goes through the Mukala API.