My Agent Can't Handle Large Files — What's the Workaround?

Last updated: March 24, 2026

Gumloop agents have a 32 MB attachment limit per file, and the underlying AI model has its own limits for images (5 MB) and PDFs (32 MB / 100 pages). If your file exceeds these limits, use the agent's code sandbox to chunk, compress, or pre-process it before the model sees it.

Symptoms

  • You attach a file to the agent chat and see File exceeds 32MB limit

  • The agent says something like the file exceeds the 5MB size limit and cannot be processed for an image

  • A PDF is rejected with a message about exceeding the size limit or page count

  • The agent seems to silently ignore a large file attached via Slack or another trigger

What Are the Actual Limits?

Context

Limit

Notes

Agent chat attachment (upload)

32 MB per file

Enforced before the file reaches the agent

Image sent to AI model

5 MB

Images are auto-resized if dimensions are too large, but file size is a hard cap

PDF sent to AI model

32 MB and 100 pages

Encrypted PDFs are also not supported

Sandbox workspace

200 MB total, 200 MB per file

Files written by sandbox code during a session

Workflow file upload

200 MB

Uploads are chunked into 10 MB pieces automatically

Cause

These limits exist for two reasons. First, the upload limit (32 MB) prevents the browser from timing out during transfer. Second, the AI model limits (5 MB for images, 32 MB / 100 pages for PDFs) are set by the model providers themselves — the agent can't send the model more data than the provider accepts.

When a file arrives via Slack or another trigger and exceeds the agent's internal limit, the agent silently skips the file rather than crashing — so you may not see an error at all.

Workarounds

1. Use the agent's code sandbox to pre-process the file

The sandbox can handle files up to 200 MB. Ask the agent to write a Python script that processes the file before analysis. For example:

  • Large PDFs: Ask the agent to extract only the pages you need, or split the PDF into smaller chunks and process them one at a time. Libraries like PyMuPDF and pypdf are pre-installed.

  • Large CSVs / spreadsheets: Ask the agent to read only specific columns or row ranges using pandas, then summarize or filter before analysis.

  • Large images: Ask the agent to resize or compress the image using Pillow (pre-installed) before sending it to the model.

Example prompt: "This PDF is 150 pages. Read it in the sandbox with PyMuPDF, extract text from pages 1–50, and summarize the key findings."

2. Upload to Google Drive / cloud storage and give the agent a link (Best Option)

If the file is too large to attach directly, upload it to Google Drive, Dropbox, or another cloud service. Then give the agent the link and ask it to download and process the file in the sandbox using its integration tools.

3. Split the file before uploading

For very large files, manually split them into smaller chunks before attaching:

  • Split a large CSV into multiple files under 32 MB each

  • Split a PDF into sections (e.g., 50-page batches)

  • Compress images below 5 MB before attaching

Validation

After applying a workaround, confirm that:

  • The agent acknowledges the file and begins processing it (no silent skip)

  • You see output from the sandbox code execution (for sandbox-based workarounds)

  • The agent's response references content from the file, confirming it was read

Still Need Help?

If this didn't resolve your issue, reach out to support at support@gumloop.com.

Related Docs