Currently swarming_server and isolate_server are specified in the tricium service config (https://chromium.googlesource.com/infra/infra/+/d48fe43e5d7047b7a316984184bbb77477a063d4/go/src/infra/tricium/api/v1/config.proto#17). This means that if someone wanted to set up the public tricium instance to analyze a private repo, it would use the public swarming server, which isn't right.

The right answer might involve either making an internal tricium instance, or supporting internal (private) projects to be analyzed by the public instance. In this case, we'll have to make sure that nobody can access information about analysis runs (including swarming logs) for the internal projects. This might involve supporting specifying an internal swarming instance the project repo.