Error log, retry notice, a waiting counter that keeps counting
CI / test (push) Successful in 1m8s
CI / build (push) Successful in 5m13s
CI / deploy (push) Skipped

- generation_errors: every failed attempt (model, attempt, outcome, duration); admin
  Hibanapló page and GET /v1/admin/errors; kept 90 days
- The waiting page counts from when the text was asked for (requested_at), not from
  the page load, so a reload no longer restarts it
- It says when an attempt failed and is retried, and when a failed text's unit was
  given back; admins see the error

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012vLHzh5qUhFtmJhzkcTbXL
This commit is contained in:
feco
2026-11-18 21:47:37 +00:00
co-authored by Claude Opus 5.5
parent cf2fb9d454
commit 02e4e225f0
22 changed files with 569 additions and 25 deletions
+9 -1
View File
@@ -148,7 +148,15 @@ impl Analyzer {
}
Err(e) => {
let can_fall_back = !job.on_fallback && job.fallback_model.is_some();
match next_step(&e, job.attempts, self.max_attempts, can_fall_back) {
let next = next_step(&e, job.attempts, self.max_attempts, can_fall_back);
let outcome = match next {
Next::Retry(_) => "retry",
Next::Fallback => "fallback",
Next::Fail => "failed",
};
let model = active_model(&job.model, job.fallback_model.as_deref(), job.on_fallback);
self.log_error("question", job.id, "question", model, job.attempts, outcome, e.message(), took).await;
match next {
Next::Retry(wait) => {
sqlx::query(
"UPDATE questions SET status = 'pending', error = $2, \