> ## Documentation Index
> Fetch the complete documentation index at: https://docs.tokenlab.sh/llms.txt
> Use this file to discover all available pages before exploring further.

# Upload File

> Uploads a JSONL batch input file for later batch creation.

## Overview

This endpoint is part of TokenLab's OpenAI-compatible Batch API. Batch jobs use uploaded JSONL files, run asynchronously, and can return output/error files later.

## Supported File Modes

TokenLab treats `/v1/files` as three explicit modes:

* Batch files: use `purpose=batch` for `/v1/batches`.
* Image/reference files: use `purpose=user_data` or `purpose=vision` for file-backed image requests.
* Anthropic Files API mode: add `anthropic-beta: files-api-2025-04-14` for Anthropic-native file storage.

## Notes

* Batch input files must use `purpose=batch`.
* `completion_window` is currently `24h`.
* Output ordering is not guaranteed; always match by `custom_id`.
* Streaming is not supported inside batch items.

## Example

<RequestExample>
  ```bash cURL theme={null}
  curl -X POST "https://api.tokenlab.sh/v1/files" \
    -H "Authorization: Bearer sk-your-api-key" \
    -F "purpose=batch" \
    -F "file=@batch-input.jsonl;type=application/jsonl"
  ```
</RequestExample>

## Request / Response

Use the interactive OpenAPI panel above for the exact schema.

## Operational Tips

* Use `custom_id` for idempotent downstream reconciliation.
* Expect `output_file_id` and `error_file_id` only after the worker finalizes the batch.
* Batch pricing may differ from synchronous pricing because `isBatchRequest=true` discount rules apply.

## Image File Inputs

This same `/v1/files` route can upload image reference files for `/v1/images/edits`.

* Use `purpose=user_data` or `purpose=vision`.
* Optionally include `model` to select the image-edit model; it defaults to `gpt-image-2`.
* The returned `file_id` is bound to the selected image-edit configuration and can be reused in `images[].file_id`.

## Anthropic Files API Mode

This same `/v1/files` route also supports the Anthropic Files API path.

* Add `anthropic-beta: files-api-2025-04-14` to switch the route into Anthropic mode.
* In Anthropic mode, upload the original file directly and do not send `purpose=batch`.
* The returned `file_id` is bound to the same official Anthropic channel/workspace and should be reused from `/v1/messages` requests.
* If you plan to reference the uploaded file from Anthropic message content, keep using Anthropic-native headers (`x-api-key`, `anthropic-version`, and the beta header when required).

## Response example

<ResponseExample>
  ```json 200 OK theme={null}
  {
    "id": "file_abc123",
    "object": "file",
    "bytes": 32768,
    "created_at": 1706000000,
    "filename": "batch-input.jsonl",
    "purpose": "batch"
  }
  ```
</ResponseExample>

## Important fields

<ResponseField name="id" type="string">File identifier used by related APIs.</ResponseField>
<ResponseField name="purpose" type="string">Purpose assigned to the file. Uploads accept `batch`, `user_data`, or `vision`.</ResponseField>
<ResponseField name="bytes" type="integer">File size in bytes.</ResponseField>