Open Resources

Open Resources collects community-maintained research assets with contributor acknowledgement, review status, and optional small file uploads.

Motivation

Materials science moves faster when useful resources are easy to find, cite, reuse, and improve. Open Resources is a community-wide index for high-quality datasets, benchmark activities, code, models, reading lists, software, tutorials, synthesis resources, and directories that materials researchers can trust.

Important resources are often scattered across papers, GitHub repositories, lab pages, personal notes, benchmark websites, and supplementary files. Open Resources gives the community a curated place to collect downloadable files or durable URL links, with contributor acknowledgement, citation text, license information, source links, tags, and notes about why a resource is useful.

Resource Tabs

• Datasets: data files, curated data links, formats, licenses, and citation notes.
• Benchmark: benchmark activities and leaderboards such as Matbench, Matbench Discovery, and generative-model benchmarks. Benchmark entries use the same contribution, review, search, URL, citation, and small-file upload workflow as datasets.
• Code: repositories, scripts, notebooks, and implementation resources.
• Models: model cards, checkpoint links, and hosted model resources. Model files are expected to be linked rather than uploaded.
• Software & Tools: web apps, GUIs, packages, workflow managers, visualization tools, calculators, and other reusable utilities.
• Tutorials: courses, workshop pages, notebooks, videos, teaching modules, and practical how-to resources.
• Synthesis: synthesis recipes, precursor resources, experimental protocols, failed-synthesis knowledge, and condition datasets.
• Directory: labs, research groups, societies, mailing lists, workshops, conferences, summer schools, and community initiatives.
• Reading List: papers, books, tutorials, reviews, and other reading material organized by topic.

Benchmark Entries

Use Benchmark for reusable evaluation activities rather than one-off datasets. Good entries include a task definition, metric, train/test split or evaluation protocol, leaderboard URL, citation, license, and contributor acknowledgement.

Supported optional benchmark upload formats match Dataset uploads: .csv, .tsv, .json, .jsonl, .txt, and .zip.

Additional Resource Categories

Software & Tools, Tutorials, Synthesis, and Directory entries use the same contribution, review, search, detail-page, URL, citation, acknowledgement, topic/tag, and optional small-file workflow as Dataset entries. Use Software & Tools for reusable materials-science utilities, Tutorials for learning materials, Synthesis for experimental knowledge and protocols, and Directory for community lists such as labs, societies, conferences, and mailing lists.

AI URL Parsing

Each Add/Edit resource form includes an AI parse button beside the URL field. The parser fetches the URL metadata, asks the configured LLM to infer a concise title, description, and tags, and fills those fields in the form. If the LLM is unavailable, the app falls back to local parsing from the page title, meta description, and URL domain.

The Description or Summary field in every Open Resources Add/Edit form supports Markdown with a Write/Preview composer and toolbar for headings, emphasis, links, code, quotes, lists, and uploaded images. The Image toolbar button uploads a small image and inserts a Markdown image link such as Image: caption at the cursor. Markdown rendering is used consistently for datasets, code, models, benchmarks, software/tools, tutorials, synthesis resources, directory entries, and reading items.

Author Acknowledgement Emails

Every Open Resources Add/Edit form includes an optional Author / maintainer emails field. Submitters can enter multiple email addresses separated by commas, semicolons, spaces, or new lines. The backend validates and deduplicates these addresses before saving.

Author / maintainer emails are used only for acknowledgement notifications. When an admin approves a new resource or an approved-resource revision, MaterialsAtlas sends a short email to those addresses with the resource title, type, submitter, and public detail-page link. Drafts, pending submissions, rejected entries, and ordinary edits do not trigger these emails.

Community Voting

Approved Open Resources entries can be voted up from both the list view and the detail page. Votes are public community signals that help useful datasets, code, models, benchmarks, reading items, software, tutorials, synthesis resources, and directory entries rise in the default listing. Voting is separate from recommending an item to another user and from any future private save/bookmark workflow.

Admin Approval Queue

Open Resources submissions with pending status are centrally visible from Admin TODO. The TODO card Public resources need approval (#) links to #/admin/public-resource-approvals, which lists pending datasets, code, models, benchmarks, software/tools, tutorials, synthesis resources, directory entries, and reading entries in one queue.

The approval queue supports search, resource-type filtering, pagination, preview/open, approve, edit, delete, and reject actions. It is backed by GET /api/admin/open-resources/approvals and uses the existing admin review endpoint POST /api/admin/open-resources/{resource_id}/review.

Approved entries remain editable by their original submitter, but those edits do not overwrite the public version immediately. A submitter edit to an approved entry creates a pending revision linked to the approved entry. The current approved version stays visible until an admin reviews the revision. Pending revisions appear in the same Public resources need approval queue as new submissions; approving a revision applies it to the existing public entry, while rejecting it leaves the current public entry unchanged.

Detail Pages

Selecting Open from a resource list opens a dedicated detail page instead of expanding a crowded block in the list view. The detail route is #/research/open-resources/detail?resource=<id> and includes a Kaggle-style tab row:

• Overview: the current resource card, description, citation, links/files, tags, acknowledgement, review status, and admin/user actions.
• Code: related notebooks, scripts, repositories, demos, and optional small code artifacts.
• Discussion: questions, comments, usage notes, reproducibility notes, and troubleshooting.
• Suggestions: dataset/model/benchmark improvement proposals, corrections, and curation requests.

Phase one implemented the dedicated detail page and Overview tab using existing Open Resources data.

Phase two adds backend-backed detail posts for the Code, Discussion, and Suggestions tabs. Each post belongs to one Open Resources entry and stores title, body, optional URL, optional citation, tags, author, and optional small file attachment for Code posts. The tabs support search, pagination, inline expansion, creation by logged-in users, and deletion by the post author or an admin. Discussion posts also support threaded replies so users can answer questions and continue resource-specific Q&A without leaving the detail page.

Detail post routes:

• GET /api/open-resources/{resource_id}/detail-posts?post_type=code|discussion|suggestion&q=...
• POST /api/open-resources/{resource_id}/detail-posts
• DELETE /api/open-resources/detail-posts/{post_id}
• POST /api/open-resources/detail-posts/{post_id}/replies
• DELETE /api/open-resources/detail-post-replies/{reply_id}

Detail posts and replies can be added only to approved Open Resources entries. If an Open Resources entry is deleted, its attached detail posts, replies, and uploaded child files are removed with it.