# Internal Sirocco Meeting 27-11-2024

## Caffeine
Junkies get their dosis :raised_hands:

## Cleanup
clean up issues and PRs

## IR catch up / PR
Go over new IR design. with parameters or not?

## Milestones
Start defining [milestones](https://github.com/C2SM/Sirocco/milestones) and attaching issues to them.

### 1. Prorotype running
#### Features
- Just the sandbox
- No pre/post processing
- with restarts
#### Dev
- Integration with `aiida-workgraph`
- ICON task

### 2. Early users
#### Add the arbitrary task feature
- Use `aiida-shell`
- Interface design: how do users scripts access input data?
- Either through input arguments (requires users to write their scripts this way)
- Or through a yaml file Sirocco drops in the working directory containing the path to all input data. We would provide basic examples for how to parse such a file in `python` or `bash`.
#### CLI?
Maybe verdi is enough for now with some guidance
#### Status aware IR
Same remark

#### Have a machine running AiiDA 24/7
- CSCS authentication
- Hardware specs:
- CPU: multicore REALLY paralell, not necessarily performant
- memory: 8GB
- storage:
- able to grow
- First 1/2 TB
- suitable for postgres database

### 3. First release

#### status aware `core.Workflow`
for pretty printing and visualizing

=> verdi CLI with docs

#### Documentation

#### Enhance plugin for *arbitrary* scripts
- Allow environment activation
- Allow for *sources* as opposed to a single script

## Accelerating development?
ExClaim is already in need for the workflow tool. I can dedicate most of my time to it for the next months. What about Alex, Julian and Rico?

- Rico: I have an ongoing GT4Py development cycle until pretty much the end of the year (taking holidays into account). Next year I can dedicate at least 1-2 months from the beginning.
- For us at PSI, it's usually 50% SwissTwins, 50% general AiiDA dev

***

## Notes from the meeting

- Added `when:` keyword with `before, at, after`
- Parameters allowed `single` and `all` -> `all` is the default
- Think about using in-memory DB for parameters (might be much quicker for large graph, but overhead...)
- Materials Cloud graph visualization tool for large graphs
- Add status property to tasks, and for large graphs that are running shrink visualization to current cycle and one before and after
- Possibly use AiiDA links for graph subselection