Building Generative AI apps on Google Cloud with LangChain

This was a great session to do at Cloud Next 2024! Thanks Harrison Chase, Steren Giannini, and Thomas Menard. I really enjoyed this! The recording of the session is now up (find it below).

In this session:

• Harrison (LangChain founder) introduces LangChain.
• Steren (Group Product Manager Cloud Run) talks about Cloud Run.
• Thomas demos L’Oreal GPT.
• I show how to build a LangChain app from the ground up.

L’Oreal GPT is an interesting project. Large companies want to give their employees access to LLM tools, but they don’t want them to use the consumer versions. They need some guarantees that their input isn’t used for training and can potentially leak internal data, that’s why. At L’Oreal they’ve built an internal version of Gemini/ChatGPT to solve this problem. For example, they use Gemini through Vertex AI and that means these terms apply.

According to Thomas, many employees use L’Oreal GPT and they can even ask questions over internal data sources. It’s really interesting to see what they were able to build in just three months. You should see the demo.

I think my demo went pretty well too. I show how to use LangChain to ask questions over the Cloud Run release notes. I build the whole app from scratch and deploy it to a public URL. I’ll share a codelab later so you can reproduce my demo at home! Here’s how I walk you through the code:

You can find the full recording of the talk at Cloud Next 2024 here: