GloxPost

• 

  Main Title

Inspiration

Creating content for different platforms sounds easy until you actually have to do it. A single video can require multiple titles, descriptions, captions, keywords, and a lot of rewriting. We wanted to turn that repetitive process into something simple: give the AI one idea and let it handle the rest.

What it does

GloxPost takes a title and some context from the user and turns them into a complete content package. It generates improved YouTube titles, evaluates their performance, creates YouTube descriptions and Instagram captions, and produces a structured TXT/PDF report.

How we built it

We built GloxPost in Python using OpenRouter for the AI layer. We separated the project into different stages for analysis, generation, evaluation, workflow management, and report formatting. This made it easier to keep each part of the agent focused on one job.

Challenges we ran into

One of the biggest challenges was getting the AI to consistently return structured JSON that our Python code could actually process. We also had to deal with formatting long descriptions, keeping generated content relevant to the original context, and making sure reports were created with unique filenames instead of being overwritten.

Accomplishments that we're proud of

We're proud that GloxPost isn't just a chatbot that generates text. It's an actual workflow that analyzes an idea, generates multiple pieces of content, evaluates the results, and turns everything into usable reports. Getting the entire pipeline working end-to-end was probably the most satisfying part.

What we learned

We learned a lot about building practical AI workflows rather than just making individual API calls. We also learned how important structured outputs, prompt design, error handling, and modular project architecture are when working with LLMs.

What's next for GloxPost

We'd like to take GloxPost beyond content generation by adding analytics-based optimization, thumbnail generation, support for more platforms, scheduling, and eventually a feedback loop where the agent can learn which types of content perform best.

Built With

• git
• llms
• openrouter
• python
• python-dotenv
• reportlab
• workflows