dire cape
We have been using completions API to power query and plot generations at MakerDojo (https://makerdojo.io) since mid last year. We found ourselves needing a way to quickly experiment with prompt variations, versioning the prompts, templating and caching of responses in production. We built the infrastructure needed to do all of this. Realizing this is a common workflow for everyone building application using LLMs, we built a platform out of this infrastructure and are calling it Promptly (https://trypromptly.com). We have now opened it up for everyone to try out.
Using Promptly, we can create high level endpoints on top of Open AI APIs with prompt versioning, templating and caching. Users can call promptly APIs instead of direct Open AI APIs to use versioned prompts with template values. Both completions and image generation APIs are currently supported. Another advantage with using something like Promptly is the ability to switch providers without any changes to application code. Coming soon is support for Stability APIs and integrated testing. Please give it a try and let me know what you think. Are there any APIs that you'd like to be supported?
hearty badge
echo solar