templates/bill.html: кнопка Остановить расчет, серийный режим при ошибке, 3 попытки

[konard]

Summary

Fixes #15

Three changes to templates/bill.html requested in the issue title:

1. Кнопка «Остановить расчет» — adds a red outline button that appears
   while processArtists is running. Clicking it sets stopRequested = true,
   which short-circuits the parallel loop, the per-artist task, and any
   pending retry delay. The button hides automatically when processing
   finishes (success, stop, or critical error).
2. Параллель не больше 1 запроса, пока есть ошибка — once any error is
   encountered, maxConcurrent drops to 1 and is not restored to its
   initial value (2) for the rest of the run, so every following request
   (retry or new artist) is serial while the system is unhealthy. The
   previous behavior (introduced in Fix: Issue #9 - Make maxConcurrent variable and restore after retry #10) restored concurrency after a single
   retry. processWithLimit now reads the current maxConcurrent on each
   iteration so the drop takes effect immediately for the next artist.
3. 3 попытки перерасчета после ошибки — replaces the isRetry boolean
   with a retryCount counter. The function recurses with retryCount + 1
   up to maxRetries = 3, keeping the existing 5-second delay between
   attempts. After exhausting the 3 retries the error is propagated as
   before.

Changes

• templates/bill.html:
  • New #stopBill link below the bill table.
  • New module-level stopRequested flag, reset to false at the start of
    each processArtists run.
  • processArtist(artist, retryCount = 0) with a 3-retry loop and stop
    checks before/after the retry delay.
  • maxConcurrent is forced to 1 on the first error of a run and stays at
    1; the post-retry restoration is removed.
  • processWithLimit reads the closure variable maxConcurrent per
    iteration instead of taking it as a parameter.
  • End-of-run code path handles the stopped, errored, and successful
    states and hides the Stop button in all of them.
• experiments/test_issue15_logic.js: standalone Node script that asserts
  (a) up to 3 retries occur before an error is thrown, (b) maxConcurrent
  becomes 1 on the first failure and is never restored, and (c)
  stopRequested = true halts processing.

Test plan

• node experiments/test_issue15_logic.js passes all 4 cases.
• Manual smoke test on a live /bill page:
  • Click "Остановить расчет" while artists are processing — progress
    label switches to "Остановлено пользователем" and no more requests fire.
  • Trigger a failing request — the log shows three retries (1/3, 2/3, 3/3)
    with 5 s delays, then either succeeds or reports an error.
  • After any error, the parallelism stays at 1 for the rest of the run.

🤖 Generated with Claude Code

[konard]

Working session summary

PR is ready for review.

Summary

Implemented the three requirements from issue #15 in templates/bill.html:

1. «Остановить расчет» button — appears while processArtists runs, sets a stopRequested flag that halts the parallel loop, the per-artist task, and the retry sleep.
2. maxConcurrent = 1 while errors persist — first error drops concurrency to 1 and keeps it there for the rest of the run (the post-retry restoration from Fix: Issue #9 - Make maxConcurrent variable and restore after retry #10 is removed). processWithLimit now reads the current maxConcurrent per iteration so the drop applies immediately to the next artist.
3. 3 retry attempts — processArtist(artist, retryCount = 0) recurses up to maxRetries = 3 with the existing 5 s delay; only after exhausting retries does it throw.

Also added experiments/test_issue15_logic.js covering the retry count, the non-restoration of maxConcurrent, and the stop flag (all 4 cases pass).

PR is now ready for review: #16

---

This summary was automatically extracted from the AI working session output.

[konard]

🤖 Solution Draft Log

This log file contains the complete execution trace of the AI solution draft process.

💰 Cost: $1.885220

📊 Context and tokens usage:

• 79.0K / 1M (8%) input tokens, 16.2K / 128K (13%) output tokens

Total: (4.3K new + 87.8K cache writes + 1.8M cache reads) input tokens, 16.2K output tokens, $1.885220 cost

🤖 Models used:

• Tool: Anthropic Claude Code
• Requested: opus
• Model: Claude Opus 4.7 (claude-opus-4-7)

📎 Log file uploaded as Gist (1164KB)

• View complete solution draft log

---

Now working session is ended, feel free to review and add any feedback on the solution draft.