`streamObject` with `@ai-sdk/anthropic` only provides a full object once finished via `.toTextStreamResponse()` about ai

[lgrammel]

Currently not supported.

Anthropic now supports https://docs.anthropic.com/en/api/messages-streaming#input-json-delta which could be used.

[lgrammel]

Right now it is. I'll look into supporting input-json-delta (it was not available when the provider was first implemented)

[lgrammel]

Example with delay that shows the tool call is streamed:

import { anthropic } from '@ai-sdk/anthropic';
import { streamObject } from 'ai';
import dotenv from 'dotenv';
import { z } from 'zod';

dotenv.config();

async function main() {
  const result = await streamObject({
    model: anthropic('claude-3-5-sonnet-20240620'),
    maxTokens: 2000,
    schema: z.object({
      characters: z.array(
        z.object({
          name: z.string(),
          class: z
            .string()
            .describe('Character class, e.g. warrior, mage, or thief.'),
          description: z.string(),
        }),
      ),
    }),
    prompt:
      'Generate 3 character descriptions for a fantasy role playing game.',
  });

  for await (const partialObject of result.partialObjectStream) {
    // delay 50 ms to simulate processing time:
    await new Promise(resolve => setTimeout(resolve, 50));

    console.clear();
    console.log(partialObject);
  }
}

main().catch(console.error);

[Evgastap]

Thanks for the fast reply! For extra context, throwing in mode: "json" does trigger an error on the anthropic provider, but leaving out that line, in the end, still returns an object. I guess that's expected bevahiour.

[lgrammel]

I just checked and it is supported in the AI SDK: https://github.com/vercel/ai/blob/main/packages/anthropic/src/anthropic-messages-language-model.ts#L323

However, Anthropic waits and then streams all deltas really quickly, giving the impression that it only provides a full object.