LlmCall.Media.Dialect.Adapter (llm_call v0.1.0)

Copy Markdown View Source

use for a media adapter.

Injects the behaviour, the canonical aliases, the codec's name and modality, and the shared validate_request/2 pipeline:

use LlmCall.Media.Dialect.Adapter, codec: :openai_images, modality: :image

The codec supplies info/0, capabilities/2, encode_submit/2, decode_submit/2, validate_params/1, validate_references/1, and validate_provider_options/2.

namespaces: names the provider_options namespaces this codec reads. It is required: a namespace outside the list is a validation error, so a typo cannot silently drop the caller's options.

model_fun: replaces the model check for protocols that constrain the model id itself (Gemini's predict path rejects bare ids containing / or :).