🧠 How I made HubermanGPT

Hello from the heights!

It's Luke Skyward here from Smoothing the Learning Curve!

In this issue, we're diving deep into the world of GPTs, specifically my latest creation - HubermanGPT. 🧠

🚀 Ready to explore? Strap in and let's launch into the knowledge cosmos!

The Journey of Creating HubermanGPT

Creating a custom GPT isn't a walk in the park. There are challenges like:

• Misleading AI responses

• Limitations in knowledge integration

• The struggle for accuracy

But, where there's a challenge, there's also an opportunity. That's where HubermanGPT shines. 🌟

Crafting HubermanGPT - A Step Beyond

What sets HubermanGPT apart? The answer: Retrieval Augmented Generation (RAG).

✨ RAG combines a language model with an external database, pulling in precise information dynamically.

For HubermanGPT, this meant creating an extensive, relevant database from the Huberman Lab Podcast, surpassing the standard 20-file limit.

🔍 We used the YouTube Transcription API for data extraction, followed by summarization and embedding creation for accurate retrieval.

The Heart of HubermanGPT

Our GPT doesn't just answer; it understands and retrieves.

• It analyzes health-related questions.

• Fetches relevant podcast summaries using a FastAPI endpoint.

• Integrates these summaries into comprehensive responses with links for further exploration.

Here's a sneak peek into how it's done:

Inside the GPT, the API (i.e. AWS Lambda Function) is called exactly at this moment:

Configuring Actions in Custom GPTs

Setting up Actions in Custom GPTs is straightforward, especially if you have a FastAPI endpoint ready.

All you have to do is copy and paste into the Action box (inside GPTs UI) the OpenAPI JSON from your FastAPI docs.

You can access it here:

This integration allows for seamless interaction between the GPT and the external knowledge base.

Conclusion

The approach taken in HubermanGPT represents an interesting approach to custom GPT development.

It demonstrates the power of combining large language models with a well-curated knowledge base.

Thanks for joining this episode of Smoothing the Learning Curve. I'll see you in the next issue. Until then, keep reaching for the stars! 🌌

Clear skies,

Luke Skyward