Short version: check the live health files first, then pick the channel that matches what you have.
Most "it stopped working" reports are the pipeline behaving exactly as designed. Free configs die constantly — that is measured here, not hidden.
| Question | Where the answer already is |
|---|---|
| How many configs passed the real client checks in the last run? | health.json → cascade.buckets |
| Is an upstream source down? | health.json → summary and the per-source list |
| What does each tier/file contain right now? | index.json |
| Did the last run even succeed? | Actions tab |
| Everything at a glance | Live dashboard |
If health.json shows the last run finished and the buckets are non-zero, the
pipeline is fine — the specific server you tried is simply one of the many that
died since publication. Pull the subscription again and your client will fetch
a fresh list.
| You have… | Go to |
|---|---|
| A question — "which tier should I use?", "does this work with client X?" | Discussions → Q&A |
| An idea — a new output format, a tier, a docs improvement | Discussions → Ideas |
| A reproducible bug — the pipeline, the workflow, or a published file is wrong | New issue → Bug report |
| A whole protocol/tier failing in a client | New issue → Config problem |
| An upstream source to add | New issue → Suggest an upstream source |
| A security problem | SECURITY.md — report privately, not as a public issue |
| General chat, outage notices, quick questions | @Raydikalx on Telegram |
| Want to change the code | CONTRIBUTING.md |
The difference between a report that gets fixed and one that stalls is almost always numbers.
Not very useful:
The configs don't work.
Useful:
Every
vlessentry inprotocols/vless.txtis rejected by v2rayNG 1.9.x withinvalid reality settings. 20/20 I tried failed the same way.singbox.jsonentries for the same servers import fine.
The second one names the file, the client, the error, and how many you tried — which is enough to reproduce without a single follow-up question.
This is a single-maintainer project run in spare time. There is no SLA.
Issues with a reproduction get looked at first, because they can actually be
acted on. Questions that are already answered by health.json or the README
may just get a link back — that is not a brush-off, it is the fastest correct
answer.
Nothing here costs money and nothing is gated. If you want to support the work, that is entirely optional.