Social Image & Caption Generator

Description

I built an automated pipeline that generates social images and captions, then publishes them to Instagram and Facebook with minimal manual effort. The core challenge was reliability. Creative outputs can look good in a test run, but scheduling, API quirks, and token issues can break a live pipeline. I wanted a system that separates content creation from publishing, supports review, and still runs on a predictable schedule.

The workflow is built in C# with Redis backed background jobs so content generation does not block publishing. Captions are created with OpenAI LLM prompts tuned to a specific voice, and images are generated with Flux image models. A scheduling layer decides when posts go out, and the publishing layer talks to the Instagram and Facebook Graph APIs. I added token refresh, retry logic, and clear state tracking so the pipeline can recover from transient failures without duplicating posts.

I containerized the pipeline with Docker so the same setup runs locally and in production. Config is driven by simple environment variables for schedules, API keys, and prompt templates. I also added a dry run mode that generates assets without posting, which helps me validate a batch before it goes live.

I kept the system builder friendly by making each stage visible. Generated assets can be reviewed before publishing, and logs show exactly what was posted and what is still pending. The pipeline uses simple status states like draft, scheduled, and published, which makes it easier to reason about outcomes. I also focused on clear failure messages, because debugging API errors can be time consuming. The goal was to make it trustworthy enough to run without daily babysitting. I track each batch with a simple run id for traceability.

The result is a low touch publishing workflow that keeps content flowing while preserving quality control. It reduces manual work and creates a repeatable routine for social output. The architecture also leaves room for future additions like approvals, A/B testing, or multi channel expansion. This project is a practical example of how automation and AI can support content operations without sacrificing oversight or brand quality.