We’re glad you could make it out to Index!
Today is all about trying out some of the tooling we’ve built and deep diving on a new way of thinking about databases. Think of it as somewhere between a workshop and a hack-a-thon.
You'll also get the chance to listen to some folks talk about their own work, divulge secrets, etc.
What is Bog?
Bog is a database runtime that makes every attempt to do as much work as possible as early as possible to make reads incredibly fast. In Bog's case, this means compiling queries into functions that eagerly update their output as mutations occur. Bog and its constituent components are built in Rust with high performance and flexible deployment targets in mind.
Today you'll be working with some of the internal libraries that power Bog:
- Fold: Our take on an incremental programming framework, and Bog's "engine"
- ESE: Or, "Embedded Static Embeddings", our first take on a "compiler oriented" approach to static embedding.
- ANNy: Or, "Approximate Nearest Neighbots ...yeah", a very fast crate for creating HNSWs.
How to build Bog things
For BOG-A-THON 2, we've established a few hack categories we'll be judging and presenting awards for. These categories feel pretty Bog-aligned.
To help you get your bearings, we've created a few github repos to help you get started with domain-specific projects that correlate to the hack categories:
We’ve created a cargo workspace that contains some of the Bog tooling we’ve built, clone this repo and use the project creation script to frame out your project.
Bog-A-Thon Schedule
| Time | Activity |
|---|---|
| 1:00pm | Welcome |
| 30min | Placeholder |
| 1:00pm–6:00pm | Project hacking, small group discussion |
| 6:00pm–7:00pm | Pizza and demos |
Potential Bogcore project ideas (Click to expand)
PANTRY-4000: Auto-filter recipes based on bought or consumed grocery items
Use a recipe data set to create a searchable system that only shows you recipes that are valid with your current pantry items. Use a bog style db to auto update which recipes are valid as new grocery items are added to the pantry, and as items are consumed by cooking a recipe.
Selective Memory Wiper: delete ideas from agent memory / context
Create a speech-to-text system that listens to conversations, indexes the context using ese / anny, and also has an interface for searching / selecting / inputing concepts to delete. Make the corpus interact-able to make sure things are being forgotten.
Skip the speech to text by kickstarting with an existing dataset:
Interesting Datasets (Streaming, etc.)
- Foursquare Open Places
- MTA Transit Data APIs
- Bluesky Firehose Documentation
- Bluesky Jetstream Repository
- Flower Office Weather Station
- Magnifica Humanitas Text Dataset
Staff
Questions, concerns, anything really:
Flower Co. people (Ed, Sam, or William) will be wearing obvious white Bog shirts.