Week of April 15th

Hi all, here's my first attempt at writing a newsletter for the weekly updates I have been pushing.

TL;DR:

• We're going to be at #PyCon in Pittsburgh in May as part of Start-Up Row!
• New Hamilton release 1.58.0 driven by OS contributions, with one coming from a user at 安海斯 - 布希英博 !
• Meet-up recording: learn about how Thierry Jean approached building ML models & evaluation with Hamilton as well as a deeper dive into loading & saving data with Hamilton.
• New blog & Hamilton example show casing a lightweight and flexible #ETL stack using dltHub with Ibis; pure python libraries that don't require systems to run!
• New Burr release 0.14.0 with with UI updates, as well as documentation explaining the multi-agent "divide-and-conquer" example.

We're excited to be accepted to PyCon as part of Start-Up row. We'll have a booth for the duration of the conference; come grab a sticker!

Are you going to be there? If so, let us know as we're trying to organize an informal meet-up / happy-hour.

------------------------

Hamilton Release 1.58.0

-------------------------

pip install sf-hamilton==1.58.0

?? New features - driven by OS contributors

We continue to get more coverage for our data savers & loaders!

• Add Pandas reader for SPSS by Swapnil Dewalkar in #814
• YAMLDataLoader and YAMLDataSaver as default data adapters by Walber Moreira at AB-InBev (his first contribution ?? ) in #819.

??? Fixes:

• No Output Requested with .materialize? should lead to a clearer error by Swapnil Dewalkar in #810

Documentation / Examples:

• We updated the ChatGPT & DALLE3 image telephone example code. It’s now simpler to run, and is a cute example that uses both Hamilton and Burr.

-----------------------------

Hamilton Meet up Recording

-----------------------------

Catch the recording of this month's meetup here. In it we covered:

1. Thierry Jean 's experience using Hamilton for managing ML Experiments and what led to the creation of the lightweight experiment manager that Hamilton ships with.
2. We covered recent feature releases and contributions to Hamilton.
3. We went deeper into the different ways you can structure loading/saving data with Hamilton. Going over the "Data savers & loaders" concept and when they're most appropriate.

---------------

From our blog

---------------

In this past week's blog we wrote about how you can create a lightweight and flexible #ETL stack using dltHub , Ibis, and Hamilton. The example used shows how one could pull slack data and transform it.

The bigger picture/theme is that all of these tools work within a python process and are therefore quite lightweight and portable; say good-bye to infrastructure lock-in for your pipelines!

See the code here.

------------------------

Burr Release 0.14.0

-------------------------

pip install burr==0.14.0

?? New features:

• Collapsable views of rendered state:

??? Fixes:

• Better error messaging for when no inputs are provided but are required.

?? Documentation:

• Want to see an agent example with Burr? We've now add more documentation to our "divide-and-conquer" example that shows two "agents" working together to complete a task. There are two implementations: one in LangChain the other in Hamilton.

That's all for this week -- let me know what you think of this format!