WritingmateWritingmate Documentation

Using Files

Attach files when you want a model to summarize a document, find specific information, compare sources, or use your material as context for a response.

Writingmate can extract text from supported document formats. Images use a model's image-input capability. A file being accepted for upload does not guarantee that every model can understand its content or that text extraction will succeed.

Supported uploads

The file picker accepts these document formats:

  • Text: .txt
  • CSV: .csv
  • JSON: .json
  • Markdown: .md
  • PDF: .pdf
  • Word: .docx
  • Excel: .xls, .xlsx, .xlsm, .xlsb
  • HTML: .html
  • XML: .xml
  • RTF: .rtf

You can also attach an image when the selected model supports image input. Writingmate may resize an image larger than 1,024 pixels on either side before sending it to the model.

Each file must be 25 MB or smaller. This upload limit is separate from the model's context limit: a file below 25 MB can still contain too much text for one request.

File attachment availability depends on your plan. If a feature is not included, Writingmate shows the required plan in the menu. See Compare Subscription Plans.

Upload files from your computer

  1. Start a signed-in chat.
  2. Select the + button beside the message box. Its label is Add content or tools.
  3. Select Upload from computer.
  4. Choose one or more files.
  5. Wait until every upload finishes and the attachment previews appear above the message box.
  6. Enter an instruction, then send the message.

You can remove an attachment with its × button before sending. Writingmate prevents the message from being sent while a file is still uploading.

Supported documents uploaded this way are processed into text and saved to your workspace's file list, so you can reuse them later. Images are attached to the conversation and follow the image-input rules of the selected model.

Add saved files from the knowledge base

Use a saved document without uploading it again:

  1. Select + → Add from knowledge base.
  2. In Knowledge Management, search for a file or select its preview button.
  3. Select up to three files in one selection.
  4. Select Add file to chat or Add files to chat.

Writingmate disables a saved file when it could not extract text from that file. The row shows No text content, and its tooltip asks you to re-upload it.

The Knowledge Management dialog also lets you upload a supported document, preview extracted text, or delete a saved file.

Paste an image or file

Copy an image or file, place the cursor in the message box, then press Ctrl+V on Windows or Cmd+V on macOS. Writingmate adds clipboard files through the same upload process.

When a long block of pasted text exceeds your configured threshold, Writingmate may save it as a .txt file instead of placing all of it directly in the message box. A notification confirms when this happens.

Model support

Document and image support work differently:

  • For supported documents, Writingmate extracts text before the model reads it. This allows many text models to work with documents even when the provider does not accept the original file format directly.
  • Image attachments require image input. If the current model cannot read images, Writingmate can switch to a compatible model and show a notification.
  • Image-generation models are designed to produce images, not analyze documents. If one is selected when you attach a document, Writingmate can switch to a compatible text model.
  • In Compare mode, attaching a file can replace the selected model set when Writingmate needs a compatible fallback.

Extraction quality depends on the file. Complex spreadsheets, damaged documents, unusual formatting, and PDFs without usable text may produce incomplete or empty results. Preview a saved file in Knowledge Management when exact extraction matters.

Multiple files

Upload from computer accepts multiple files in one selection and uploads them together. The Knowledge Management picker allows up to three saved files per selection.

Ask a specific question when attaching several files. For example:

Compare the renewal terms in these contracts and list every difference with the source filename.

Combine the revenue totals from these spreadsheets and flag rows with missing values.

Several large files can exceed the available context or use the daily allowance quickly. Split the task or attach only the relevant files when possible.

How files affect message usage

File content becomes model input, so it contributes to token and message usage. One counted message covers up to 16,000 effective tokens, and a request above that amount can use multiple counted messages.

Writingmate includes the attached document text, relevant chat history, and the model's response when it calculates usage. For very large documents, Writingmate can give the model a file-search tool so it reads relevant chunks instead of placing the entire document into one request. The chunks the model reads still count as input.

The Knowledge Management dialog marks a document as Large file when its extracted text is above the message-counting threshold. File size in MB and extracted token count are different, so a small compressed document can still have a large token count.

See Message Limits for the exact formula and reset rules.

Privacy and deletion

Writingmate stores uploaded documents and their extracted text as private workspace content. When you attach a file, the content needed to answer your request is sent to the selected AI provider. That provider's data-handling rules also apply.

Only attach information that you are allowed to share with the selected provider. You can delete a saved file from Knowledge Management. Deleting an attached preview before sending only removes it from that pending message; it does not delete a document already saved in your workspace.

See Data Collection and the Writingmate Privacy Policy for storage, provider, and deletion details.

Tips for better results

  • Name each file in your prompt when asking for comparisons.
  • Ask for page, section, row, or filename references when you need traceable answers.
  • Preview extracted text before relying on exact figures or legal language.
  • Start a new chat when earlier file context is no longer needed.
  • Split very large documents or attach only the relevant sections.