Bug when using Extract API

[bwprado]

Bug Report: Body is unusable: Body has already been read

Describe the Bug

When attempting to extract data from a PDF file using the extract method of the llama-cloud-services Typescript library, the API returns the error:
TypeError: Body is unusable: Body has already been read.

This suggests that the body stream is being consumed multiple times during the request, which prevents successful processing of the file.

---

Files

Unfortunately, I cannot attach the specific PDF file here, but the issue occurs consistently with any valid PDF file uploaded.

---

Job ID

I do not have the specific Job ID, as the request fails before completion. However, the issue can be reproduced by running any extraction job with the provided code.

---

Client

• Typescript Library

---

Additional Context

Options Used:

• Extraction Target: PER_DOC
• Extraction Mode: BALANCED
• Extract Model: openai-gpt-4-1
• Multimodal Fast Mode: false
• High Resolution Mode: true
• System Prompt: Custom instructions to extract transactions from the document.

---

Code Snippet:

Here is the relevant code used to call the extract method:

const arrayBuffer = await file.arrayBuffer();
const uint8Array = new Uint8Array(arrayBuffer);
const result = await extractor.extract(statementSchema, config, undefined, uint8Array, file.name);

---

Error Message:

TypeError: Body is unusable: Body has already been read
    at uploadFile (C:\wip\cookies\.next\dev\server\chunks\ssr\82481_llama-cloud-services_24c288ef._.js:5254:57)
    at async Module.extractStateless (C:\wip\cookies\.next\dev\server\chunks\ssr\82481_llama-cloud-services_24c288ef._.js:5524:20)
    at async LlamaExtract.extract (C:\wip\cookies\.next\dev\server\chunks\ssr\82481_llama-cloud-services_24c288ef._.js:6988:16)

This issue appears to be a bug in the API or the Typescript library, as the body stream should not be consumed multiple times during the request.