acme-directory 0.4.2
python server.py
openrouter/example/tiny:free   16 cases   16 from cache   free

Selection accuracy     43%   6/14
  positives            50%   5/10
  siblings             25%    1/4
Abstention             50%    1/2
Argument validity      79%  11/14

tool           hit       went instead to
search_orgs    17%  1/6  search_users 83%
doStuff        50%  1/2  (nothing) 50%
search_users   67%  4/6  search_orgs 33%

search_users captures 83% of the prompts meant for search_orgs.
search_orgs captures 33% of the prompts meant for search_users.

9 failing cases, -v to see them
