omnirt.engine¶
The async engine underlying the HTTP server and batched dispatch. Auto-rendered reference: → API Reference / omnirt.engine.
Summary:
OmniEngine— constructor takesmax_concurrency,pipeline_cache_size,batch_window_ms,max_batch_size.submit(request)returns aJob;wait(job_id)blocks for itsGenerateResult.ResultCache— LRU cache ofGenerateResultobjects keyed by request signature (enabled by the HTTP server).get_default_engine()— lazily constructs a process-wideOmniEnginesingleton.
For concurrency / batching tuning, see Dispatch & Queue.