Recorded walkthrough

How DueCare works, in one walkthrough.

A guided tour of the problem, the system, and a worked example. No setup required, just press play.

Trouble loading the embed? Watch on YouTube.

Chapters Approximate
  1. 01 The problem we're addressing 00:00
  2. 02 Design constraints we chose 00:45
  3. 03 System overview & harness 01:30
  4. 04 Worked example, end to end 02:30
  5. 05 How it integrates with partners 04:30
  6. 06 Open questions & what's next 05:30
Reproduce it yourself

Run the same pipeline yourself.

The walkthrough above is the fastest path. If you'd rather see the code execute, fork the public notebook or clone the repository. Both run the same harness.

01 · Notebook In-browser GPU
# Open the notebook # File → Copy & Edit # Settings → Accelerator → GPU   !pip install duecare from duecare import Runtime rt = Runtime.load("default") rt.run("<your scenario>")

Public reproducible notebook

A fork-and-run notebook with the runtime wired to a free in-browser GPU. Modify the prompts, swap rule packs, and inspect the harness output cell by cell.

  • Free GPU runtime (subject to host quotas)
  • Same pipeline shown in the video
  • Logged outputs & pack versions visible inline
↗ Open notebook ~10 min to first run
02 · Source Your hardware
# Clone the public repo git clone https://github.com/<org>/duecare cd duecare   # Install & pull a default pack make install make pack-default   # Run a scenario duecare run --scenario examples/<file>

Source repository

If you have GPU access (workstation, lab, or cloud), the README walks you through install, default pack download, and running an example end to end.

  • GPU recommended for inference
  • Public packs ship as samples
  • Issues & discussions open for feedback
↗ View on source repo ~30 min for first install
i

The runtime requires a GPU for inference, so we don't host a public 24/7 endpoint. The video above is the easiest way to see the system in action; the notebook and source let you run it.