Getting only 1 nearest neighbor even when n is set to higher values (reopen)

[Lavi101]

Issue #679 mentioned this problem. Several people are still encountering this problem, myself included. Moreover, running the provided example code of the repo reproduces only one neighbor. Changing search_k/number of trees/number of indexed points
/etc. did not help either. Does anyone know what might be the issue?

The issue is with both get_nns_by_item and get_nns_by_vector return only one neighbor (always [1]).

Running on macOS (LTS) with Python 3.13.3

Tks! :)

[EvX57]

I'm experiencing the exact same issue. get_nns_by_item() always returns only one neighbor ([0]), regardless of the input data or hyper-parameter values.

I'm using macOS 15.5, Python 3.11.11, and annoy 1.17.3.

Any troubleshooting advice would be greatly appreciated. Thank you in advance!

[caseybasichis]

Just encountered this...

1 nearest neighbor even with the simplest tests -- this code is definitely broken.

It works fine on a slightly older machine with Python 3.11.6 instead of 3.12.9.

Is this package just abandoned? What alternatives are you all using?

[mvinyard]

@caseybasichis @EvX57 @Lavi101 thank you for reporting your experiences: this was helpful to me. After several years of using and relying on this package, I am coming to find the same experience. My best guess is an unaccounted-for dependency update. I tried in two environments using python 3.11.13 and python 3.13, to no avail. However, it does work in python 3.10.18. am personally planning on migrating to their newer ANN library, Voyager.

[elanou]

May be related to issue with latest clang version here: #686