Show HN: Formula8.ai – A formula-based approach to AI prompts
formula8.aiWe just launched our new product, which we developed ourselves out of the need to have a better abstraction layer for ChatGPT.
Yes, we know; another AI content tool.
As a marketing agency, we produce different types of content such as newsletters for large public companies. We found ourselves typing the same prompts over and over again, even though the structure was often very similar or identical, except for the actual content or topic.
Formula8 offers us an easy way to develop parameter-based prompts for this and combine them in templates and make them smarter with helper functions, such as automatically crawling the content of an external URL for article teaser excerpts.
Frankly, we didn't really plan to turn it into a real product originally, but since our team adapted it super fast and we can certainly give those who have the same problem a useful tool with it, we're introducing it today.
Happy if it seems to be helpful to anybody and to collect your input or ideas. Stephan here, lead-developer of Formula8. For all who are interested in the technical implementation, I am available for exchange. Quick overview of the stack:
- Remix React Application with Server-Side-Rendering (TypeScript)
- Supabase Backend & Database (Postgres)
- AWS Serverless Functions for batch-processing & fronted delivery
- AWS Cloudfront as CDN UI and continuous deployment? ;-) We use https://github.com/architect/architect for provisioning and deployment of the functional web app via GitHub Actions (…whenever they work — availability was not that great the recent weeks) Tests are based on Cypress (https://www.cypress.io). For the UI/UX we work with https://tailwindui.com and paid them for their great work.. saves a lot of time building React components.