Thin wrapper over Guava's server-side LLM endpoint (v1/llm/generate).
Used by the higher-level helpers (Guava.IntentRecognizer,
Guava.DatetimeFilter, Guava.DateRangeParser) and available directly.
Summary
Functions
Generate text from prompt, returning {:ok, text} | {:error, %Guava.Error{}}.
When json_schema is given, the server constrains the output to match it and
the returned string is JSON.
Like generate/3, but returns the text or raises Guava.Error.
Functions
@spec generate(Guava.Client.t(), String.t(), map() | nil) :: {:ok, String.t()} | {:error, Guava.Error.t()}
Generate text from prompt, returning {:ok, text} | {:error, %Guava.Error{}}.
When json_schema is given, the server constrains the output to match it and
the returned string is JSON.
@spec generate!(Guava.Client.t(), String.t(), map() | nil) :: String.t()
Like generate/3, but returns the text or raises Guava.Error.