Class LlmChatBinding

java.lang.Object
com.codename1.ai.LlmChatBinding
public final class LlmChatBinding extends Object

Convenience wiring that turns a ChatView into an LLM-driven chat surface in one call. Pulls the user's input out of the view's ChatInput, appends it as a USER message, opens a streaming chat call against the supplied LlmClient, and pipes deltas into a freshly-appended assistant bubble.

ChatView itself has no dependency on the AI package, so apps that want a peer-to-peer messaging UI (a WhatsApp clone, for example) can keep using ChatView without pulling LlmClient onto their classpath -- it's only when you call bind(ChatView, LlmClient, ChatRequest) that the binding is established.

Example
LlmClient client = LlmClient.openai(SecureStorage.getInstance().get("openai_key"));
ChatRequest base = ChatRequest.builder()
        .model("gpt-4o-mini")
        .addMessage(ChatMessage.system("You are a terse assistant."))
        .build();
ChatView view = new ChatView();
LlmChatBinding.bind(view, client, base);
// ...add view to a Form and that's it.

The base request's messages are a fixed prefix: they lead every request, ahead of the conversation held by the view. That is what makes it the place for a system prompt, and it means anything else put there -- a seed user or assistant turn -- is repeated on every turn too, so put those in the view instead.

The view's accumulated history is replayed on every turn so the model has full conversation context. The original baseRequest is treated as a template -- its model, tools, temperature, etc. are preserved across turns, and its messages lead every request as described above.

  • Method Details

    • bind

      public static void bind(ChatView view, LlmClient client, ChatRequest baseRequest)
      Connects a chat view to a streaming client. Each submitted user message replays the current view history while retaining the base request's model, sampling, tool, metadata, and safety settings.
      Parameters:
      view - chat UI to drive
      client - provider client used for every turn
      baseRequest - request template and optional initial messages