Files

The Files endpoints store files for two uses: model input such as PDFs and images (purpose user_data), and batch input files (purpose batch). They are OpenAI-compatible. You only ever see files uploaded with your Glexir organization, plus the output files of your own batches.

Upload a file

import { readFile } from "node:fs/promises";
import Glexir, { toFile } from "@glexir/sdk";

const glexir = new Glexir({ apiKey: process.env.GLEXIR_API_KEY });

const file = await glexir.files.create({
  file: await toFile(await readFile("report.pdf"), "report.pdf"),
  purpose: "user_data",
});

console.log(file.id); // file-...

Response:

{
  "id": "file-abc123",
  "object": "file",
  "bytes": 182044,
  "created_at": 1761350400,
  "filename": "report.pdf",
  "purpose": "user_data"
}

Then pass the ID to a model, for example as an input_file part with file_id in Responses. Uploading, listing and downloading files does not use credit.

Purposes and limits

Glexir accepts two purposes. Any other value returns 400 with param purpose, even though the FilePurpose type lists more of OpenAI's values.

A file over its limit returns 413 with code file_too_large.

file is an Uploadable (File | Blob | Uint8Array | ArrayBuffer | ReadableStream<Uint8Array> | Response). Give it a filename with an extension: pass a File, or use toFile(data, "name.ext"). See Audio for toFile().

Batch input files

Glexir checks a batch file when you upload it, not when you start the batch. Every line must:

• be valid JSON with a string custom_id that is unique in the file,
• use "method": "POST" and a url of /v1/responses, /v1/chat/completions or /v1/embeddings,
• use the same url and the same body.model as every other line,
• have an object body without stream: true,
• pass the same model and tool rules as a live request to that endpoint.

A batch file can have at most 50,000 lines. Lines can't reference stored responses, conversations or reasoning items (previous_response_id, conversation). A failing line returns 400 with param file and a message such as Batch line 3 needs a unique string custom_id.

Glexir rewrites each line before storing it, the same way it rewrites live requests (for example, the end-user identifier). toBatchJSONL() builds a valid file for you. See Batches.

List files

import Glexir from "@glexir/sdk";

const glexir = new Glexir();

for await (const file of glexir.files.list({ purpose: "batch", limit: 50 })) {
  console.log(file.id, file.filename, file.bytes);
}

for await walks every page. To page by hand, await the call for a CursorPage<FileObject> and use hasNextPage() and getNextPage(). See Pagination.

The list comes from Glexir's own records. Batch output and error files are added when Glexir sees the batch finish, for example when you call batches.retrieve().

Retrieve, download and delete

import Glexir from "@glexir/sdk";

const glexir = new Glexir();

const info = await glexir.files.retrieve("file-abc123");

const response = await glexir.files.content("file-abc123");
const text = await response.text();

const deleted = await glexir.files.delete("file-abc123");
console.log(deleted.deleted); // true

files.content() returns the raw Response. Read it with .text(), .arrayBuffer() or .body. The timeout stops applying once headers arrive, so large downloads aren't cut off.

A file ID that doesn't belong to your organization returns 404.

Methods

All methods take options?: RequestOptions as the last argument. See Client configuration.

import type { APIPromise, FileCreateParams, FileDeleted, FileListParams, FileObject, PagePromise, RequestOptions } from "@glexir/sdk";

// The methods on glexir.files
interface Files {
  create(body: FileCreateParams, options?: RequestOptions): APIPromise<FileObject>;
  retrieve(fileId: string, options?: RequestOptions): APIPromise<FileObject>;
  list(query?: FileListParams, options?: RequestOptions): PagePromise<FileObject>;
  delete(fileId: string, options?: RequestOptions): APIPromise<FileDeleted>;
  content(fileId: string, options?: RequestOptions): APIPromise<Response>;
}

glexir.files.create()

POST /files, sent as multipart.

glexir.files.retrieve()

GET /files/{file_id}.

glexir.files.list()

GET /files.

glexir.files.delete()

DELETE /files/{file_id}. Returns { id, deleted, object: "file" }.

glexir.files.content()

GET /files/{file_id}/content. Returns the raw Response.

File object

Errors

See HTTP error codes for the full list.