Advertisement · 728 × 90

Posts by Kevin Whinnery

Likely unintentional- tracking g on a GitHub issue?

1 year ago 0 0 0 0
Preview
GitHub - openai/openai-realtime-console: React app for inspecting, building and debugging with the Realtime API React app for inspecting, building and debugging with the Realtime API - openai/openai-realtime-console

Before the break, I updated the OpenAI Realtime API console sample app to use WebRTC (and @fastify.dev with @astro.build 🚀) - should be pretty light weight, uses just the RTCPeerConnection directly.

Ask it for advice designing a color palette 🎨

github.com/openai/opena...

1 year ago 0 0 0 0
Preview
a black and white photo of a woman standing in front of a mirror and saying `` i mean ... '' . ALT: a black and white photo of a woman standing in front of a mirror and saying `` i mean ... '' .

I like Christmas... but I find no fault in this plan either 💡

1 year ago 1 0 1 0
Kevin Whinnery Kevin Whinnery is a software developer and developer experience expert based in Minneapolis, MN.

First holiday TODO ✅

bare-minimum landing page at kevin.mn

1 year ago 3 0 2 0
Post image

Thanksgiving break mode engaged for middle teenager.

1 year ago 2 0 0 0
Preview
Function calling in four acts. A simple but complete soup-to-nuts example of function calling with the OpenAI API. Function calling in four acts. A simple but complete soup-to-nuts example of function calling with the OpenAI API. - example.js

This flow is how you can provide a model external data to use for generations, or allow a model to control any external system programmatically. Here's the full example I just showed, usable in Node or @deno.land:

Code: gist.github.com/kwhinnery/a6...

Docs:
platform.openai.com/docs/guides/...

1 year ago 0 0 0 0
Post image

4️⃣ Finally, make another call to the model containing the previous prompts and messages we sent, plus the function call message sent by the model after the first request, and our response.

Now the model can generate a response for the user!

1 year ago 0 0 1 0
Post image

3️⃣ In this case, the model correctly decides it should call the "generate_horoscope" function. Instead of a text response, the model returns a message containing a JSON string of data that should be passed in to a function call in your code.

1 year ago 0 0 1 0
Post image

2️⃣ Generate output from the model based on a prompt. Depending on the content of the prompt, the model may decide to call a function that you provide to it.

1 year ago 0 0 1 0
Post image

1️⃣ Define a list of functions the model could decide to call based on a prompt. We provide only one function that expects an astrological sign in an input object.

1 year ago 0 0 1 0
Advertisement
Post image

Simple (but complete) walkthrough of function calling in the OpenAI API in four acts. If you haven't tried it before, this is how you connect OAI models to external data or systems. Link to full code at the end of the 🧵

1 year ago 2 0 2 0

Changi airport lives up to the hype - I get why people plan to go there early before flying out. I didn't have time to see the butterfly garden, but a sweet old lady gave me some katsu curry I watched her make from scratch for $8 and it was amazing.

1 year ago 1 0 0 0
Running evals in the OpenAI dashboard
Running evals in the OpenAI dashboard YouTube video by Kevin Whinnery

Did you know you can run evals on your prompts in the OpenAI dashboard? Here's a quick walkthrough of uploading a test data set and comparing GPT-4o model outputs to ground truth data in a test dataset.

youtu.be/DtFCGsqwtGY

1 year ago 4 0 0 0
Post image

By convention I always post the doggo first on a new social platform ☁️

1 year ago 3 0 1 0