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

# Files, Images and PDFs

> Send images, PDFs and audio to your Enterprise LLM Endpoint in the chat messages, as a URL or as base64.

You send files **inside the chat messages**, in the standard OpenAI content-part format. There is no separate upload step. Put each file in a message's `content` list, next to your text.

<Info>
  The model must accept the file type. Images and audio go to the model directly. PDFs are converted to text first, so every model can read them.
</Info>

## Supported file types

| File                                      | Content part  | Send it as                         |
| ----------------------------------------- | ------------- | ---------------------------------- |
| Image (PNG, JPEG, WebP, GIF)              | `image_url`   | A public URL or a base64 data URL  |
| PDF                                       | `file`        | A public URL or a base64 data URL  |
| Audio (WAV, MP3, FLAC, OGG, M4A and more) | `input_audio` | Base64 only. URLs are not accepted |

## Images

<CodeGroup>
  ```python Python theme={"theme":{"light":"github-light","dark":"github-dark"}}
  import base64
  from openai import OpenAI

  client = OpenAI(
      base_url="https://modelslab.com/api/v1/enterprise/proxy/v1",
      api_key="YOUR_ENTERPRISE_API_KEY",
  )

  with open("chart.png", "rb") as f:
      image = base64.b64encode(f.read()).decode()

  response = client.chat.completions.create(
      model="default",
      messages=[{
          "role": "user",
          "content": [
              {"type": "text", "text": "What does this chart show?"},
              {"type": "image_url", "image_url": {"url": f"data:image/png;base64,{image}"}},
          ],
      }],
  )
  print(response.choices[0].message.content)
  ```

  ```javascript Node.js theme={"theme":{"light":"github-light","dark":"github-dark"}}
  import fs from "node:fs";
  import OpenAI from "openai";

  const client = new OpenAI({
    baseURL: "https://modelslab.com/api/v1/enterprise/proxy/v1",
    apiKey: "YOUR_ENTERPRISE_API_KEY",
  });

  const image = fs.readFileSync("chart.png").toString("base64");

  const response = await client.chat.completions.create({
    model: "default",
    messages: [{
      role: "user",
      content: [
        { type: "text", text: "What does this chart show?" },
        { type: "image_url", image_url: { url: `data:image/png;base64,${image}` } },
      ],
    }],
  });
  console.log(response.choices[0].message.content);
  ```

  ```bash cURL theme={"theme":{"light":"github-light","dark":"github-dark"}}
  curl https://modelslab.com/api/v1/enterprise/proxy/v1/chat/completions \
    -H "Authorization: Bearer $ENTERPRISE_API_KEY" \
    -H "Content-Type: application/json" \
    -d '{
      "model": "default",
      "messages": [{
        "role": "user",
        "content": [
          {"type": "text", "text": "What is in this image?"},
          {"type": "image_url", "image_url": {"url": "https://example.com/photo.jpg"}}
        ]
      }]
    }'
  ```
</CodeGroup>

To send more than one image, add more `image_url` parts to the same message.

## PDFs

Send a PDF as a `file` part with a `filename` and the `file_data`. `file_data` is a public URL or a base64 data URL that starts with `data:application/pdf;base64,`.

<CodeGroup>
  ```python Python theme={"theme":{"light":"github-light","dark":"github-dark"}}
  import base64
  from openai import OpenAI

  client = OpenAI(
      base_url="https://modelslab.com/api/v1/enterprise/proxy/v1",
      api_key="YOUR_ENTERPRISE_API_KEY",
  )

  with open("contract.pdf", "rb") as f:
      pdf = base64.b64encode(f.read()).decode()

  response = client.chat.completions.create(
      model="default",
      messages=[{
          "role": "user",
          "content": [
              {"type": "text", "text": "List the payment terms in this contract."},
              {"type": "file", "file": {
                  "filename": "contract.pdf",
                  "file_data": f"data:application/pdf;base64,{pdf}",
              }},
          ],
      }],
  )
  print(response.choices[0].message.content)
  ```

  ```bash cURL theme={"theme":{"light":"github-light","dark":"github-dark"}}
  curl https://modelslab.com/api/v1/enterprise/proxy/v1/chat/completions \
    -H "Authorization: Bearer $ENTERPRISE_API_KEY" \
    -H "Content-Type: application/json" \
    -d '{
      "model": "default",
      "messages": [{
        "role": "user",
        "content": [
          {"type": "text", "text": "Summarise this report."},
          {"type": "file", "file": {"filename": "report.pdf", "file_data": "https://example.com/report.pdf"}}
        ]
      }]
    }'
  ```
</CodeGroup>

**How PDFs are read.** The endpoint converts every PDF to text (markdown) before the model reads it. This works well for PDFs that contain text, such as reports, contracts and exported documents. A **scanned PDF** (a photo of each page, with no text layer) gives little or no text. For scanned pages, send each page as an **image** instead.

The endpoint always uses this conversion. If you send a `plugins` setting that asks for another PDF engine, the endpoint replaces it.

## Audio

Send audio as an `input_audio` part with base64 `data` and its `format`. Audio URLs are not accepted.

```python Python theme={"theme":{"light":"github-light","dark":"github-dark"}}
import base64

with open("call.mp3", "rb") as f:
    audio = base64.b64encode(f.read()).decode()

response = client.chat.completions.create(
    model="default",
    messages=[{
        "role": "user",
        "content": [
            {"type": "text", "text": "Transcribe this call and list the action items."},
            {"type": "input_audio", "input_audio": {"data": audio, "format": "mp3"}},
        ],
    }],
)
```

## Size limit

A request can be at most **30 MB**, including all text and files. Base64 makes a file about 33% larger, so the largest file you can send inline is about **22 MB**.

* For a larger image or PDF, send a **URL** instead of base64. A URL adds almost nothing to the request size.
* A request over the limit returns `413`:

```json json theme={"theme":{"light":"github-light","dark":"github-dark"}}
{
  "error": {
    "message": "The request body is larger than 30 MB. Send large files as a URL instead of base64.",
    "type": "modelslab_proxy_error",
    "code": 413
  }
}
```

## Privacy and security

* **Your files are not stored.** The endpoint forwards each request and does not save your messages or files. If a request fails, our error logs can keep a short excerpt of its text to help us fix the problem. Files are never logged.
* **Only providers that do not collect data.** The endpoint sends every request only to model providers that do not store or train on your data. You cannot turn this off.
* **PDFs are converted by a document parser** before the model reads them. The parser gets the PDF only to convert it.
* **Use short-lived URLs.** The model provider downloads a file URL directly, so the URL must be public. For private files, use a signed URL that expires after a few minutes, such as an S3 or R2 presigned URL.
* **Base64 for confidential files.** A base64 file travels only inside your HTTPS request, and no public URL exists.

## Tips

* Put the text instruction first and the files after it in the same message.
* Every file adds input tokens. For a long PDF, ask about specific sections, or send only the pages you need.
* When you ask several questions about one file, send the file again with each request. The endpoint does not keep files between requests.