Can server mode handle concurrent requests?

[corneliusroemer]

Reading the server mode part of the README I was left wondering what happens if multiple processes try to use the server at the same time. Will the requests be handled in parallel, or processed (randomly/orderly) sequentially, or does the client get error if server already busy?

[corneliusroemer]

Getting Claude to look at the code seems to be:

Concurrent --use-server filter requests are processed one at a time, in the order the OS accepts them — clients neither run in parallel nor get a "server busy" error; they simply wait their turn.`

[bede]

Server mode removes the index loading overhead but processes multiple requests serially. Requests should queue in a well behaved manner up to some limit (LLM suggests 128) and be processed in order of submission.

The recently added python bindings may be a better fit for what I understand your use case to be.
https://github.com/bede/deacon/blob/main/deacon-py/python/README.md

from deacon import Index
idx = Index("panhuman-1.k31w15.idx")
idx.filter("reads.fq", deplete=True, threads=8, output="reads-clean.fq")

With python bindings you could split M filtering tasks across N multithreading-capable Index() instances to combine multithreading and multiprocessing with bounded memory use.