Gorilla: An API Appstore for LLMs

[@aadarshlalchandani3487]

Please make a video on chromadb and vector databases in general to explain there usage in production

[@mayorc]

Agree.

[@interspacer4277]

I've been combining this with GPT4 and smol-dev with great success. Basically prompt GPT4 to generate a prompt for gorilla, then feed the gorilla+gpt specced-reply to smol-dev. Then iterate between gpt+browsing (tbh in some cases Bard works better) and smol-dev and voila... you pump out a ready-made app with api attached.

[@engineerprompt]

Wow, that's great to hear. Would love to see if you have an example code somewhere.

[@interspacer4277]

@@engineerprompt Just been doing it manually via cut+paste for now, but could easily be coded up. It's not foolproof and smol-dev isn't perfect (debug always necessary) but it's a nice flow given how reliant we are on APIs currently. At the end of the day, given how often APIs are updated, you still many times have to feed GPT or smol-dev the API references but at least the scaffolding is there before you do that.

[@ozzy1987mr]

@@interspacer4277 tendras algun ejemplo que puedas compartir para mirarlo??

[@user-pg7gd9gu1y]

Thank you for making it easy to understand Gorilla announcement!

[@74L]

Gorilla has huge potential

[@attilavass6935]

Great video, I can't wait a video about Langchain implementation of this!

[@kevon217]

Great overview and demo. Very useful model!

[@mrmoffett7034]

Great stuff, can't wait to see what you're going to show us next.

[@volkandurmaz3164]

🎯 Key Takeaways for quick navigation: 00:14 🧠 Gorilla is a language model trained on API calls, creating an "API app store for llms." 00:29 🌐 Language models like GPT require external tools for interacting with the physical world, solved using APIs. 01:24 📚 Gorilla's models are open source, specialized in making API calls, outperforming GPT-4 by 20% and ChatGPT by 10%. 02:58 🧩 The model is fine-tuned on a new dataset containing 1600 API calls, enabling more accurate interaction with external tools. 03:53 🤖 Gorilla's approach improves API-based performance, enhancing interaction between language models and external tools. Made with HARPA AI

[@ivantang5795]

Awesome introduction to Gorilla. Thanks for sharing. I wonder if we could even automate the execution of the code spitted out by Gorilla?. Probably we can using Langchain's Python agent but I'm not sure.

[@engineerprompt]

Yes, that can be done.

[@mytechnotalent]

Incredibly powerful!

[@adriangabriel3219]

How do you use Gorilla for information retrieval? Say you have a vector database with good ol' embeddings?

[@pfcwells522]

More interested in seeing how these work for running local model. Like ks anyone working on a GPT that will run locally but be able to search the internet when appropriate with out leaking the docs you have it ingest for analysis.

[@engineerprompt]

There are so many different possibilities here. Will be exploring them here.

[@sebastiansosa3072]

how does this measure up against the function calling models

[@engineerprompt]

This is specialized for api calls. I haven't done a comparison with function calling models so can't really say which one will be better.

[@fraugdib3834]

🎯 Key Takeaways for quick navigation: Made with HARPA AI

[@shubhamguptachannel3853]

😅

[@fraugdib3834]

🎯 Key Takeaways for quick navigation: Made with HARPA AI