Help shape this resource
The catalog, guide, recipes, and AI-scientist tracker are kept up to date by automated curators that run daily. But what they cover is driven by the working scientists, engineers, and clinicians who use the site. Three lightweight Issue Forms let you tell the curators what to write next.
Three things you can file
Ask a recipe question — “How should I do X?”
Use this when the cookbook doesn’t already have a recipe for what you’re trying to do. A bot replies in-thread within a few minutes pointing to the closest existing recipes or tools. The next daily Recipes curator run writes a durable recipe page if one doesn’t exist yet, then comments on your issue with a direct link to the new page and closes it.
Good prompts here are concrete and goal-shaped: “I have a counts matrix from a bulk RNA-seq experiment with batch effects across 4 sites — how should I run DE?” — not “how do I do bioinformatics?”
Share feedback on a recipe — “I tried X and…”
Worked great, worked but slow, got stuck, found a better way, something else — every report tightens the evidence label on the recipe page, adds a field-report note, or flags the recipe for review if multiple people hit the same wall.
Concrete details are what make these reports useful: the commands you ran, dataset size, hardware, wall-clock, exact error messages, any workaround you found.
Share feedback on a catalog tool — “I installed Y and…”
Same idea, scoped to one tool from the catalog. This is how last_verified stays honest and how install-path notes accumulate across operating systems and Python versions. Mention OS, install path, errors, and any workarounds.
What happens after you file
- Within minutes — a bot reads your issue and posts an in-thread comment with the closest existing recipes or tools (or a best-effort answer if no good match exists), and queues your request for the next curator run.
- Within ~24h — the next daily scheduled curator run (Recipes runs at 10:00 UTC; Catalog at 07:00 UTC) ships any durable change: a new recipe page, an updated tool note, an evidence-label bump, or a flag for review.
- At loop-close — the curator posts a comment on your issue with a direct link to the rendered page and closes the issue. You’ll get a GitHub notification when this happens.
If a request needs more than one run — for example because the curator wants to read a paper before publishing — it stays in the queue and is retried next run. Nothing gets dropped.
What you need
- A GitHub account (free). Sign up here if you don’t have one.
- No Markdown knowledge required. The forms ask plain-text questions and dropdown choices.
- No special permissions. Anyone can file.
What the bot can and can’t do
The in-thread responder is read-only on the site content — it only reads existing pages and posts a comment. Durable changes are made by the scheduled curator agents, which run with the full evidence rules, simplicity ladder, and source-verification machinery applied. That keeps the rules consistent on every change, regardless of whether the trigger was a user request or the daily directed pass.
The bot will not invent tools or recipes that don’t exist. If a question reaches beyond what’s catalogued today, it’ll say so plainly and queue the gap for the curator to consider in the next run.
Other ways to engage
- General discussion — open-ended questions and chatter that don’t fit a form belong in Discussions.
- Track changes — the easiest way to follow along is the Weekly digest: one prioritized summary of the week’s catalog, guide, tracker, and recipe activity, posted every Sunday. Open it and click Subscribe (top-right) to get it by email. For every change as it lands instead, watch the repository.
- See the machinery — the About page describes the four scheduled curator agents and how the schedules are wired up.