Release interface workflows, AI pipelines, and research environment

[ShaneLogic]

Promotes the accumulated development work to main, adding AI component generation, active-learning orchestration, coherent interface construction, and CP2K/SLURM preparation to the research workflow. Provides a reproducible ARM64 CPU environment with InterfaceML Web, JupyterLab, and a persistent research workspace.

The release also includes the earlier fullerene/perovskite model work, web and batch interface refactoring, packaging fixes, documentation updates, and lint/format cleanup. Coherent interface construction preserves site metadata and termination ordering; DFT preparation retains fixed-atom constraints and returns actionable failures.

Scope

• Source: devel at 131e91a36e4dbda67c7eb15cbe62a91cdcf8f41c; base: main at 67a4bf51346eaa6a0dbcdf25cc0dec5cd323edad.
• Includes 8 earlier development commits, the 3 commits reviewed in Add interface construction, DFT preparation, and research containers #2, and the Add interface construction, DFT preparation, and research containers #2 merge commit.
• The full release changes 230 paths: 165 added, 34 modified, and 31 removed. The recent interface/environment/lint work accounts for 52 of these paths.
• Historical content includes 40 checkpoint files already tracked on devel, one dataset CSV update, and removal of tracked caches and temporary files. Those historical assets are part of this release scope.

Validation

• The Add interface construction, DFT preparation, and research containers #2 merge tree exactly matches its approved head, 645ff74c3038e123ddb163e19107d6d0a2cb3e2a.
• Post-merge devel CI passed: Python 3.9-3.13, ML tests, and Ruff lint/format checks. The mypy job completed under the existing advisory configuration.
• The reviewed source passed 104 tests in an isolated research container; all 42 checked Python files passed formatting and Ruff reported zero findings.
• Previously verified browser flows include FAPbI3/C70 input, a 250-atom interface, interactive 3D rendering, and DFT archive download on desktop and mobile.
• The ARM64 research image, notebook execution, dependency consistency, service health, and workspace persistence were verified.
• This PR's merge CI also passed all 8 jobs against the proposed merge into main.

Validation Limits

• Existing model weights and historical training claims have not been scientifically revalidated in this release process. Model checkpoints are not bundled into the research image and must be mounted explicitly when needed.
• The research environment targets ARM64 CPU. CP2K execution, HPC submission, and accelerator training were not validated.
• Arbitrary film rotation is a geometric operation; periodic commensurability after rotation is not certified.
• Mypy remains informational (continue-on-error); a successful workflow does not establish a clean type-check result.

[sourcery-ai]

Sorry, we are unable to review this pull request

The GitHub API does not allow us to fetch diffs exceeding 20000 lines

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 131e91a36e

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Import the perovskite API under a distinct module name

The normal generate_round order imports the fullerene implementation first, which registers fullerene_e3gen/api.py as sys.modules["api"]. Inserting perovskite_e3gen at the front of sys.path here does not replace that cached module, so this import looks for PerovskiteAPI in the fullerene module and raises ImportError before any perovskite structures or interfaces can be generated. Use package-qualified imports or load each implementation under a unique module name.

Useful? React with 👍 / 👎.

[chatgpt-codex-connector]

Point the active-learning defaults at available checkpoints

A stock python -m active_learning run passes these exact paths to FullereneAPI and PerovskiteAPI, but neither fullerene_e3gen/checkpoints/best_model.pt nor any perovskite_e3gen/checkpoints/best_model.pt is included in this commit. The generation phase therefore fails with a missing-checkpoint error before the workflow can produce components; ship the intended models, select the tracked PaiNN checkpoint where appropriate, or make the required external checkpoint locations explicit and configurable.

Useful? React with 👍 / 👎.