Sellanto

The Problem

Running social media for a business is a full-time job — and most businesses are doing it wrong.

A typical small to medium business spends 6 to 10 hours every single week writing captions, designing visuals, scheduling posts across 4+ platforms, manually replying to customer messages, and trying to make sense of performance data.

The tools that exist — Hootsuite, Buffer, Sprout Social — charge between $99 and $499 per month, and they still require significant human effort. None of them generate content at a professional level. None of them handle customer conversations intelligently.

The Solution

Sellanto replaces the entire social media workflow with one AI-powered platform.

Generate content in seconds — type a brief, and Sellanto produces a polished caption tailored to the platform, an AI-generated image, and even a short video clip ready to post.

Schedule and publish automatically — posts go live across Facebook, Instagram, Twitter/X, and LinkedIn exactly when scheduled.

Let AI handle customer messages — Sellanto's Messenger chatbot reads the business's website and documents, then uses RAG-powered knowledge to answer customer questions 24/7.

Understand what's working — performance analytics, A/B testing, and competitor tracking are built in.

Architecture

The platform covers every layer of a production SaaS:

• Multi-tenant workspace isolation — zero data bleed between clients
• Role-based access control — Owner → Admin → Creator → Viewer
• AI content pipeline — GPT-4o for captions, DALL-E 3 for images, Gemini for video
• RAG chatbot — website crawling → PDF extraction → embeddings → vector search → contextual response
• Async job processing — AI calls and scheduled posts run via background workers
• OAuth token management — per-platform refresh with error recovery
• Subscription enforcement — feature gating at the API middleware level

Technical Highlights

Area	Decision	Why It Matters
Database isolation	Tenant-scoped queries at ORM level	Prevents data leakage between workspaces
RAG pipeline	Embeddings + vector search	Semantic understanding, not keyword matching
Async processing	Background job queue	AI calls never block the UI
Token management	Per-platform OAuth refresh	Robust against token expiry
Post lifecycle	State machine with role-based transitions	Enterprise content approval workflows
Multi-model orchestration	5 AI models, 2 providers	Right model per task

Why It Matters

This is not a tutorial project. This is a fully operational SaaS platform — the kind of system that companies raise funding to build with a team of 10+ engineers.

It was designed, architected, and shipped by one developer, covering every layer: backend API, frontend dashboard, AI integrations, chatbot pipeline, scheduling engine, analytics system, and admin panel.

Stack Used

• Backend: Django + Django REST Framework + PostgreSQL
• Frontend: React
• AI Models: OpenAI GPT-4o, DALL-E 3, Google Gemini, Whisper, TTS
• Async Processing: Celery + Redis
• Infrastructure: Cloud hosting with background worker architecture

Read more about how I built this Hootsuite competitor as a solo developer and see my full development toolkit.