Process-scoped attributes used to evaluate targeting rules.

Set it once where you already have the current user — a plug, or a LiveView on_mount hook — and every read in that process can then be targeted without threading a context through every call site:

# lib/my_app_web/plugs/flag_context.ex
def call(conn, _opts) do
  user = conn.assigns.current_user
  PhoenixFlags.Context.put(user_id: user.id, company_id: user.company_id)
  conn
end

# anywhere in the request
MyApp.SystemConfig.get("enable_benefits", false)

A context passed explicitly to a read wins over whatever is in the process:

MyApp.SystemConfig.get("enable_benefits", false, context: %{company_id: 999})

Scope

This is the process dictionary, so it is scoped to one process and is not inherited by processes you spawn. A Task.async/1, a GenServer.cast/2, or a spawn/1 starts with an empty context. Either set it again inside the process, or pass :context explicitly to those reads:

context = PhoenixFlags.Context.get()

Task.async(fn ->
  MyApp.SystemConfig.get("enable_benefits", false, context: context)
end)

Attribute keys and values

Keys may be atoms or strings; :company_id and "company_id" are the same attribute. Values are compared as strings, so 123 matches a rule value of "123" — consistent with every flag value being stored as a string. See PhoenixFlags.Target for the matching rules.