ALLM.TextPart (allm v0.3.0)

A text content part used in multimodal ALLM.Message{:content} lists. See spec §35.6.

Layer A — pure serializable data. :text is the only required field; :metadata defaults to an empty map. The struct is opaque from the validator's point of view (ALLM.Validate.message/1 accepts any %TextPart{} regardless of :text content — empty strings included).

Multimodal callers build %ALLM.Message{content: [...]} by hand:

iex> ALLM.Message.new(role: :user, content: [
...>   %ALLM.TextPart{text: "What is in this image?"},
...>   %ALLM.ImagePart{image: ALLM.Image.from_url("https://example.com/cat.png")}
...> ]).role
:user

Serializability

ETF round-trip via :erlang.term_to_binary/1 is total. JSON round-trip via ALLM.Serializer follows the standard __type__-tagged wire shape used by every Layer A struct.

Summary

Types

t()

Functions

new(text, opts \\ [])

Build a %TextPart{} from a string and optional keyword opts.

Types

t()

@type t() :: %ALLM.TextPart{metadata: map(), text: String.t()}

Functions

new(text, opts \\ [])

@spec new(
  String.t(),
  keyword()
) :: t()

Build a %TextPart{} from a string and optional keyword opts.

Accepts an optional :metadata keyword (a map). Empty text is allowed — the validator does not reject empty TextParts; that is an upstream concern.

Examples

iex> ALLM.TextPart.new("hi")
%ALLM.TextPart{text: "hi", metadata: %{}}

iex> ALLM.TextPart.new("hi", metadata: %{source: :test}).metadata
%{source: :test}