Will Dean

I think both options would be very useful. I've been playing around with the GitHub CLI commands required here and it seems like there might be some permission issues when...

Hi @arnold-c I have implemented this in #609 . 1 and 2 are both available for telescope. However, I haven't implemented 2 for other pickers

Yes, there are options... basic one https://github.com/pymc-labs/pymc-marketing/blob/2a4dd0362804577b971482ea192bf5e817a8eb43/tests/conftest.py?plain=1#L128-L162 One that includes some sample_stats ( maybe not needed) https://github.com/pymc-labs/pymc-marketing/blob/2a4dd0362804577b971482ea192bf5e817a8eb43/scripts/run_notebooks/injected.py?plain=1#L8-L36 https://github.com/pymc-devs/pymc-examples/blob/69563da49ceef589127140d3675a05b2ba4c3ae0/scripts/run_notebooks/injected.py?plain=1#L9-L69

Would be something for `pymc.testing` module? Maybe just the mock_sample but not the fixture. That's quite specific and small

> [@williambdean](https://github.com/williambdean) this is not including numba / jax CI? Would have to check. Is that included here? https://github.com/pymc-devs/pytensor/actions/runs/15284667195

I am anti having to maintain various backends and **very** anti having every function having if backend == 'matplotlib': ... elif backend == 'plotly': ... etc. If arviz supports multiple...

I personally dont use brackets. Seems imprecise. Does this warn with idata["unlikely_key"]?

Does it always just make a graph with only the data nodes? If only this did that, but the edges are all dropped ```python data_nodes = [var.name for var in...

So is the difference here that you'd want to edit the template before the creation of the PR (including when it is a draft)? Related to #596

Splitting by file was an easy win