Bitbucket Server code search does not discriminate between repos in projects and repos in user forks, so when searching for text, the following two annoying things can happen:

1. multiple duplicates of the same file are displayed in search results, one per fork of the repo.
2. often the file(s) in the "upstream" repository appear far down in the search results below the acres of user forks

Because Bitbucket Server codesearch does not support passing wildcards to elasticsearch, it is not possible to search for "foo NOT proj:~*"

The old third-party Palantir codesearch add-on (https://github.com/terabyte/stash-codesearch-plugin) provided a simple checkbox to exclude user forks from search results – under the hood it simply passed filter.mustNot(prefixFilter("project", "~") to the elasticsearch java API.