PhoenixKit.Modules.AI (phoenix_kit v1.7.117)

Core-side conveniences for talking to the optional PhoenixKitAI plugin. The plugin is shipped as a separate Hex package and may or may not be installed in a given host application; everything in this namespace is safe to call regardless.

Each function here uses Code.ensure_loaded?/1 to detect the plugin at runtime and returns {:error, :ai_not_installed} (or a documented fallback) when the plugin is absent. That lets feature modules in core (and sibling plugins) reference AI workflows unconditionally without depending on :phoenix_kit_ai at compile time.

See PhoenixKit.Modules.AI.Translation for the higher-level translation orchestration (prompt rendering, structured-response parsing, error normalization).

Summary

Functions

available?()

Returns true when the PhoenixKitAI plugin is loaded and exports the ask_with_prompt/4 entry point this namespace relies on.

Functions

available?()

@spec available?() :: boolean()

Returns true when the PhoenixKitAI plugin is loaded and exports the ask_with_prompt/4 entry point this namespace relies on.

Note on the semantics: this is a module-loadability check — it does NOT verify that the host has a configured endpoint, a valid prompt template, or working credentials. Use it to decide whether to surface AI-driven UI at all; the actual request is the only place that can confirm runtime readiness.

<.language_switcher_dropdown
  ai_translate={
    if PhoenixKit.Modules.AI.available?(),
      do: %{enabled: true, event: "translate_lang", ...},
      else: nil
  }
  ...
/>

Hosts that need a stricter "endpoint configured and reachable" check should call into PhoenixKitAI directly (e.g. validating that PhoenixKitAI.list_endpoints/0 returns at least one connected row) before showing the affordance.

Cheap — a single atom lookup against the code server. No DB hit.