Make Hangfire's SQLite InvisibilityTimeout configurable (30 min is too short for slow local hardware)

[luizbon]

Problem

Lingarr.Server/Extensions/ServiceCollectionExtensions.cs hardcodes the Hangfire SQLite storage invisibility timeout:

.UseSQLiteStorage(sqliteDbPath, new SQLiteStorageOptions
{
    ...
    InvisibilityTimeout = TimeSpan.FromMinutes(30)
});

On a CPU-only box running a small local model through the localai provider (e.g. Ollama with qwen2.5:3b), a single subtitle file with a few hundred lines can legitimately take longer than 30 minutes to translate line-by-line. Once it does, Hangfire assumes the worker crashed, and the translation request ends up Cancelled/requeued - the job never gets a chance to finish, no matter how long you are willing to let it run.

This makes localai effectively unusable as a fallback tier for larger files on modest hardware, even though the endpoint and model are perfectly capable of finishing the job given enough wall-clock time.

Ask

Make InvisibilityTimeout configurable via an environment variable, following the exact same pattern already used a few lines above it for worker concurrency:

options.WorkerCount =
    int.TryParse(Environment.GetEnvironmentVariable("MAX_CONCURRENT_JOBS"), out var maxConcurrentJobs)
        ? maxConcurrentJobs
        : 1;

Something like:

InvisibilityTimeout =
    int.TryParse(Environment.GetEnvironmentVariable("HANGFIRE_INVISIBILITY_TIMEOUT_MINUTES"), out var minutes)
        ? TimeSpan.FromMinutes(minutes)
        : TimeSpan.FromMinutes(30)

would let people on slower hardware (or using a large local model) opt into a longer window without needing a custom build. Happy to test if useful.

[rowanfuchs]

I Couldn't agree more, nice find!