Route LLM calls through my backend

[RafaelRMaier-CA]

As the title says, I have my custom REST API which handles LLM calls. They are all running inside EC2 containers because we dont wanna expose company data .

I want to connect the BlockNote AI component to my endpoint, so that I can route the calls through my backend, to our internally-hosted LLMs. The AI SDK has no clear way of doing this. I tried using createBlockNoteAIClient, but how can I plug this AI client to the editor?

export const customBlocknoteAIClient = createBlockNoteAIClient({baseURL: ${API_URL}/canvas/block_level_call, apiKey: ""});

const editor = useCreateBlockNote({
    initialContent: placeholderText,
    dropCursor: ({editor}) => createCustomMDImagePlugin(editor, chatSessionId!),
    schema: customSchema,
    codeBlock,
    dictionary: {
      ...en,
      ai: aiEn, // add default translations for the AI extension
    },
    // // Register the AI extension
    extensions: [
      createAIExtension({
        executor: async (opts) => {
          console.log({opts})
          // You may need to extract blockId, sessionId, etc. from options or context
          const { blockId, sessionId, blockType, user } = opts;

          // Call your backend
          const response = await apiService.blockLevelCanvasCall(sessionId, prompt);

          // BlockNote expects a Block or array of Blocks as the result
          return response;
      }})
    ]
  });