# `Planck.AI.Adapter`
[🔗](https://github.com/alexdesousa/planck/blob/v0.1.0/lib/planck/ai/adapter.ex#L1)

Translates between Planck's types and `req_llm`'s call interface.

This is the only module that knows `req_llm`'s input and output shapes.
Everything above this layer works exclusively with `Planck.AI` structs.

# `to_req_llm`

```elixir
@spec to_req_llm(Planck.AI.Model.t(), Planck.AI.Context.t(), keyword()) ::
  {model_spec :: String.t() | map(), req_llm_context :: ReqLLM.Context.t(),
   opts :: keyword()}
```

Converts a `Planck.AI.Model`, `Planck.AI.Context`, and call-site opts into
the three arguments expected by `req_llm`'s `stream_text/3`:
`{model_spec_string, req_llm_context, opts}`.

Tools from the context are added to opts as `%ReqLLM.Tool{}` structs.
Inference params (e.g. `temperature:`, `max_tokens:`) in opts are forwarded
directly to `req_llm`, which handles per-provider translation.

## Examples

    iex> model = %Planck.AI.Model{id: "claude-sonnet-4-6", provider: :anthropic, context_window: 200_000, max_tokens: 8_096}
    iex> context = %Planck.AI.Context{messages: []}
    iex> {model_spec, _ctx, _opts} = Planck.AI.Adapter.to_req_llm(model, context, [])
    iex> model_spec
    "anthropic:claude-sonnet-4-6"

---

*Consult [api-reference.md](api-reference.md) for complete listing*