Behaviour for input guard detection strategies.

Implement this behaviour to create custom strategies for detecting malicious or unwanted input. Each strategy receives the user input text, its own configuration, and the full agent context.

Creating Your Own Strategy

Here's a complete example of a custom blocklist strategy:

defmodule MyApp.InputGuard.Blocklist do
  @behaviour Nous.Plugins.InputGuard.Strategy
  alias Nous.Plugins.InputGuard.Result

  @impl true
  def check(input, config, _ctx) do
    blocklist = Keyword.get(config, :words, [])
    downcased = String.downcase(input)

    case Enum.find(blocklist, &String.contains?(downcased, &1)) do
      nil ->
        {:ok, %Result{severity: :safe}}

      word ->
        {:ok, %Result{
          severity: :blocked,
          reason: "Blocklisted word: #{word}",
          strategy: __MODULE__
        }}
    end
  end
end

Then use it in your agent configuration:

agent = Nous.new("openai:gpt-4",
  plugins: [Nous.Plugins.InputGuard]
)

{:ok, result} = Nous.run(agent, "Hello",
  deps: %{
    input_guard_config: %{
      strategies: [
        {MyApp.InputGuard.Blocklist, words: ["hack", "exploit"]}
      ]
    }
  }
)