Make lab-report OCR scan rate limit configurable via env var

[rkerian]

Problem

The lab-report OCR scan endpoint (/api/labs/ocr/extract) hardcodes its rate limit at 6 scans/hour (EXTRACT_LIMIT_PER_HOUR in
src/app/api/labs/ocr/extract/route.ts, lines 67-68), with no way to raise it. For a self-hoster batch-scanning a stack of lab reports after a doctor
visit, this throttles hard and there's no config path around it — you just wait an hour, repeatedly, or manually clear the labs-ocr: row in the
rate_limits table.

Proposed solution

Make EXTRACT_LIMIT_PER_HOUR operator-tunable via an environment variable, following the exact pattern already used for the sibling document-AI bucket (DOCUMENT_AI_LIMIT_PER_HOUR in src/lib/documents/ai-route-support.ts — default 6, clamped 1-1000, falls back to the default on unset/non-numeric). Something like LABS_OCR_LIMIT_PER_HOUR with the same resolver shape would be a drop-in, low-risk change and consistent with how the rest of the AI egress surfaces are already configured.

Alternatives considered

Manually deleting the rate_limits row for labs-ocr: via psql works as a stopgap but isn't something a typical self-hoster should have to do,
and it doesn't help mid-scanning-session — you have to notice you're throttled, drop into the DB, then resume. This also isn't practical when there are multiple users running into the throttling issue.

Affected surface

Self-hosting / Docker

Additional context

No response

Pre-flight checks

• I confirmed this is not a duplicate of an existing feature request.
• I'm proposing a feature, not asking a usage question (those go to Discussions or the docs).

[MBombeck]

@rkerian thanks, this is a fair ask and you found the right pattern to copy.

I checked it: the lab scan limit is fixed at 6 per hour and both scan modes share it, while the document AI already reads DOCUMENT_AI_LIMIT_PER_HOUR. The lab route will get the same treatment, LABS_OCR_LIMIT_PER_HOUR with the same default, the same 1 to 1000 clamp and the same fallback, and it will be on the compose whitelist and in the docs so it actually reaches the container.

While reading the route I noticed something that made your batch scanning worse than it had to be. The lab route takes a slot before it has even looked at the request, so a file it can't read, or a scan the daily AI budget turns away, still uses up one of your six. The document AI only counts a slot once the request really goes to the provider. The lab route will work the same way.

Cost stays covered by the daily AI budget either way, so raising the hourly number on your own instance is safe. This will ship with the next release.

[MBombeck]

@rkerian v1.38.23 is out: https://github.com/MBombeck/HealthLog/releases/tag/v1.38.23

Set LABS_OCR_LIMIT_PER_HOUR in your .env (anything from 1 to 1000) and recreate the container; the shipped compose file already passes it through. Scans that fail before they reach the AI, like an unreadable file or a day's budget that is used up, no longer count against the hour either.

Thanks for the suggestion and for pointing straight at the pattern to reuse.