Skip to content

"Input too large: couldn't fit with truncation" when attaching base64-files to conversation #212

Description

@jorgtho

Hi!

When using conversations-api and posting a prompt like this:

{
  "model": "mistral-medium-latest", // also happens with mistral-large-latest
  "instructions": "",
  "inputs": [
    {
      "type": "message.input",
      "role": "user",
      "content": [
        {
          "type": "document_url",
          "documentUrl": "data:application/vnd.openxmlformats-officedocument.wordprocessingml.document;base64,UEsDBBQABgAIAAAAI...", // also happens with pdfs when they exceed approximately 400 kb
          "documentName": "A file.docx"
        },
        {
          "type": "text",
          "text": "give me a summary of the document"
        }
      ]
    }
  ],
  "store": false,
  "tools": []
}

We get response: {"object":"Error","message":"Input too large: couldn't fit with truncation","type":"invalid_request_error","code":3051}

But this did not happen before, we could attach base64 files of 30 MB and more. Is this now expected behaviour or a bug?

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Fields

    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions