v1.0.0 cutoff

[Melissari1997]

Hey @brendancol

From the README I see that the project is working toward its first major release, and now the focus is mostly on features stabilization, bug fixing and other details.
I was wondering if you have a rough idea of what features are needed for the release of version 1.0.0, and if possible, to get an idea of where we are on the roadmap

Thanks!

[brendancol]

@Melissari1997 Great question and I'm hoping to box the effort to get the release out.

The rough plan is:

1. No new tools or subtools until after 1.0.0 - basically just to limit to what is there now...basically what is in the README.md feature matrix, but there may be some undocumented dark corners to add or drop from that feature matrix, but things already in the codebase...no new high-level modules.
2. Each public function / function group is free of CRITICAL and HIGH category bug as discovered by the .claude/commands/deep-sweep.md which is a composite command that runs each of the "sweep-{}" commands on a specified area of the codebase. The most important two areas for me are accuracy and security, then performance, api-consistency and documentation. Some of the issues you've been responding to were generated during a deep-sweep of the contours module for example. The deep sweeps are not deterministic and different models find different things, so Ive been using various models to do the deep sweep, but primarily Opus 4.x and gpt 5.5. As new models come out, deep sweeps should be re-ran across the codebase to find new issues and vulnerabilities.
3. Each public function / function group has a decent notebook examples, and potentially a stand alone python script example.
4. We ("xarray-spatial contributors") like the high-level spellings in the API and feel they are worthy of LTS

In terms of dividing this up to finish faster, I'm thinking it would be great if different maintainers essentially "owned" a module, at least until 1.0.0 release. For example, if you wanted to continue running deep-sweeps, or other ways to bring the contours module to "production" ready, that would be ideal. I'm especially concerned about the geotiff module because the "blast radius" is so large if there is a bug there. I've gone through dozens of deep-sweeps using Opus 4.6, Opus 4.7, and Opus 4.8 and now am only see MEDIUM and LOW categories issues. When I ask codex to scan the geotiff module, it continues to find inconsistencies and backend parity issues which I've been triaging.

My day is about over here, but happy to chat more and would love to do a 1.0.0 release planning session if youre interested.

Hope all is great with you and thanks for your participation over the past week

[brendancol]

I've made the following project board to help organize the vision above: https://github.com/orgs/xarray-contrib/projects/4/views/2

[Melissari1997]

Thank you @brendancol for the detailed explaination. Really useful for me, and I agree with your approach.
A few questions:

• Do you think using Frontier models might yield results that are very different from those of open models? Have you already tried them out and decided to keep using them?
• It’s perfectly fine with me to focus on the contours module, and I can also take on other tasks (e.g., I often work with GeoTIFFs, but also ZARR, IceChunk, etc.). I can also help write some documentation.

I can't see the project board 😞

[brendancol]

Do you think using Frontier models might yield results that are very different from those of open models? Have you already tried them out and decided to keep using them?

The open models are less capable, but depending on how they are used, still capable of creating value for xarray-spatial. I big reason the project needs human is because of this question of different results. Different results between models, and different results from the same model over time! I still think that gpt 5.5 with codex is better than claude opus 4.7, but I personally like the claude code harness better, and the difference isn't big enough for me to completely switch. I saw claude say "All Clean" on a deep sweep, when then codex (with a be brutally honest prompt) found additional security bugs that later claude verified...

It’s perfectly fine with me to focus on the contours module, and I can also take on other tasks (e.g., I often work with GeoTIFFs, but also ZARR, IceChunk, etc.). I can also help write some documentation.

Let's be clear here, Zarr and Icechunk are a big improvement over GeoTIFF IMHO. But I also work with GeoTIFFs and wanted a solution there...

I just broke out contours into its own 1.0.0 tracking issues. Its a good one to focus on.

I can't see the project board 😞

@Melissari1997 I added you as a collaborator so hopefully you can see the board now.

[Melissari1997]

I still got 404 over https://github.com/orgs/xarray-contrib/projects/4/views/2

And if I click on "projects" I see no boards

[Melissari1997]

[brendancol]

@Melissari1997 Sorry, I may not have sufficient rights to give you visibility to the board. Still troubleshooting, but for the time being, we can at least use "assign" and you should be able to open close issues.

[brendancol]

@Melissari1997 Since the Github project access never really worked out, we now have a sweeps-report command which uses the sweep state to generate a report. Here is the current one:

Sweeps Report — 2026-07-02

8 of 12 sweeps have state; 49 module-units in scope. Never-run: benchmarks, error-handling, reference-validation (commands exist, no CSV), plus dependencies (keyed by unit, not module — no data).

1. Module × sweep staleness grid

Legend: -- never inspected · ok ≤30d & nothing >MEDIUM · NNd days since · ! = HIGH/CRITICAL at last look. Sorted most-neglected first. Columns: acc=accuracy, api=api-consistency, doc=documentation, meta=metadata, perf=performance, sec=security, sty=style, test=test-coverage. (bench, err, ref are empty for all — omitted below.)

module               acc   api  doc  meta  perf   sec  sty  test
gpu_rtx/              --    --   --    --    --   64d!   --    --
bump                  --    --   --    --   77d   71d!   --    --
convolution           --    --   --    --   93d   70d!   --    --
erosion               --    --   --    --   93d   68d!   --    --
geodesic              --    --   --    --   93d   66d!   --    --
surface_distance      --    --   --    --   93d   65d!   --    --
pathfinding           --    --   --    --   78d   60d    --    --
terrain_metrics      63d    --   --    --   93d    --    --    --
glcm                 62d!   --   --    --   93d   69d!   --    --
sky_view_factor      62d!   --   --    --   93d   65d!   --    --
bilateral            62d    --   --    --   93d   70d!   --    --
emerging_hotspots    63d    --   --    --   93d   68d!   --    --
hydro/               63d    --   --    --   62d   60d!   --    --
kde                  80d    --   --    --   79d   66d!   --    --
sieve                80d    --   --    --   79d   65d!   --    --
balanced_allocation  79d    --   --    --   77d   70d    --    --
curvature            94d    --   --    --   93d   68d    --    --
edge_detection       62d    --   --    --   93d   68d    --    --
hillshade            83d    --   --    --   77d   66d    --    --
normalize            62d    --   --    --   93d   66d    --    --
terrain              83d    --   --    --   93d   60d    --    --
worley               62d    --   --    --   93d   65d    --    --
dasymetric           ok     --   --    --   ok    68d!  12d!  12d!
diffusion            62d    --   --    --   93d   66d!  12d!  12d!
corridor             62d    --   --   10d!  93d   69d    --   10d!
polygon_clip         22d!   --   --    --   ok    ok    22d!  ok
flood                ok     --   7d!   --   ok    60d    --   ok
interpolate/          --    ok   --    ok   ok     --   20d!  ok
slope                 --   34d   --    --   93d   65d   34d!  34d
viewshed             34d!   --   --   34d   88d   71d!  34d!  34d!
contour              62d    --   --   34d!  93d   70d!  34d!  ok
cost_distance        16d!   --   --    ok   ok    68d!  17d!  ok
morphology           63d!   ok   --    --   ok    69d!  12d!  ok
fire                 ok     --   ok    --   93d   68d   13d!  7d!
mahalanobis          62d    --   ok    --   93d   66d!   2d!  ok
aspect               ok     --   --   34d   34d   70d    ok   ok
rasterize            14d!  23d!  --    ok   ok    20d!  36d!  14d!
mcda/                ok    22d!  --   22d!  ok     ok    22d!  22d!
reproject/           20d!  23d!  --    ok   ok    46d   23d!  ok
resample             34d!  36d   --   36d   78d   65d!  36d   34d!
zonal                36d    ok   --   34d   ok    36d!  36d!  22d!
focal                ok     ok   --    ok   34d    ok   34d!  22d!
multispectral        94d    ok   --    ok   61d   66d!  12d!  ok
polygonize           34d!   ok   --    ok   ok     ok   36d!  ok
visibility           80d    ok   --   22d!  ok    65d    ok   22d!
classify              --    ok   ok    ok   ok    69d   12d!  ok
proximity            ok    23d!  --   34d   ok    71d    ok   ok
perlin               83d    ok   ok    ok   ok    71d!   9d!   9d!
geotiff/             ok     ok   ok    ok   ok     ok    ok   ok

Coverage per sweep: perf 48/49 · sec 47/49 · acc 39/49 · test 27/49 · sty 25/49 · meta 19/4tiff/ is the only unit green across every sweep it's been through.