Get responses streamed from 20+ AI models, in the exact same format as OpenAI. Never face issues with rate limits. Get 10x lower error rates verses OpenAI. No need to manage multiple API keys for different models.
Hey Product Hunt fam 👋 😊
Super pumped to launch Merlin's Unified API that supports all major AI models (Gemini, Mistral, Claude-2, llama-2-70b, GPT-4 etc) in an OpenAI like API format and without rate limits.
If you are an app builder (like us at getMerlin.in), you too would someday face problems of working with multiple AI providers and facing challenges with error rates as you scale.
What typically happens is this:
> You see a new API is launched, & you want to push it in production.
> It takes 3-7 days to introduce the new model in your app.
> The delay was mostly attributed to quirks of different SDKs, talking with sales teams to increase quotas, and never-ending efforts of managing different streams, token calculations, context windows, rate limits etc.
As you scale these problems just become worse. And this sucks. :(
That's why we decided to open up our existing infra for anyone to use. This infra has helped us scale to over 1M+ users and we hope this can do the same for you too. We have also kept an affordable and flexible price plan, so that this is truly scalable for you. And the price doesn't bug. 🫶
If you are an AI app builder, I hope this helps. Or if you know an AI app builder, please do share with them. 🤠
That's all folks 👋, lets keep building and shipping 💪🚀
P.S. Kudos to OG builders Ser Puneet, Ser Lakshay, Ser Sourabh and the greater Merlin team for shipping this amazing product in almost no time. 🫶❤️
Report
@pratyush_r8 Cool launch! When we'll try to create apps in our studio, we will definitely test your service. Followed you on Twitter and here :)
@eugeneyarovoii thanks Eugene. My email is pratyush@foyer.work if you ever need any support. We will keep improving the API in the coming days, but if you need anything, just shoot an email / note. Wish you best with your app studio. 😊🚀
@pratyush_r8 I love this idea. It might be good to chat live about it when you have some time. There's lots that can be done with respect to your Unified API concept here! Congrats.
Really relevant problem statement! Have a question - do we specify the alternative AI models in case our primary model(e.g GPT4) is not working, or will that we internally prioritised?
@rohan_singhvi1 Right now, we have added an internal logic to serve this. We first do Azure GPT-4, then OpenAI and then serve a different model. We'll make this logic configurable in future too.
With LLMs booming, this can be a key player for swiftly integrating different LLM models and shipping to prod without any unnecessary hassle. Best Wishes!
Brilliant and Fantastically built, Congratulations to the ?makers and everyone involved. I am going to see how i can best utilise this for my clients whom i advise on B2B Marketing. @pratyush_r8 Well done 👍
This is going to be really useful. Now it exists it's really obvious that it needed to exist, which is normally the sign of something big :)
I've signed up to pay as you go and I'll be giving it a test pretty soon.
Well done guys!
Merlin - ChatGPT powered chrome extension (Cmd+M)