July 28, 2023, 4:07 p.m.

LangChain seems to be a semi-standard to start building LLM experiences

• I have heard people abandoning it for production
• documentation and the code seem messy
• it’s flexible enough to use whatever LLM I decide on, so fine-tuning a model from Hugging Face

This blog post does a good job outlining basic building blocks. Follow the links to the external colab notebook, without reading through these the post is less useful.

Memory looks like a decent way to inject truth and correct false or outdated facts for a small set of data. It is good to see that I can use PostgreSQL for storage.

>> Today I Learned: Large Language Models