Supported Formats and Limitations
Writingmate can extract text from supported documents and send that text to an AI model. An accepted upload is not a guarantee that every page, cell, or visual element will be preserved. File size, extracted text, the selected model, and the available conversation context all affect the result.
Supported document formats
The document picker accepts these file extensions:
| Content | Extensions |
|---|---|
| Plain text | .txt |
| Structured text | .csv, .json |
| Markdown | .md |
.pdf | |
| Microsoft Word | .docx |
| Microsoft Excel | .xls, .xlsx, .xlsm, .xlsb |
| Web and markup | .html, .xml |
| Rich Text Format | .rtf |
Older Word .doc files, PowerPoint files, archives such as .zip, and source or shell files such as .sh are not accepted as documents. Export them to a supported format first. Excel macros are not run.
Writingmate checks both the filename extension and the file's media type. A file can still be rejected if its extension does not match its content, or if the file is damaged, encrypted, or password-protected.
Size and selection limits
Limits depend on where you add the file:
| Upload location | Limit |
|---|---|
| Upload from computer in a chat | 25 MB per file; you can choose multiple files |
| Knowledge Management upload | 25 MB per document |
| Add saved files from Knowledge Management to a chat | Up to 3 files per selection |
| An Agent's Knowledge Files | 10 MB per file |
| A Project's Sources | 10 MB per file; you can choose multiple files |
These are upload limits, not model context limits. A compressed 5 MB PDF can contain more text than a 20 MB image-heavy PDF. Several valid files can also exceed the space available in one AI request.
Images and other media
Images are chat attachments, not reusable text documents in Knowledge Management. The selected model must support image input. If it does not, Writingmate can switch to a compatible model and show a notification.
Chat images use the same 25 MB upload boundary. Writingmate resizes an image when either side is larger than 1,024 pixels. Common browser-readable image formats such as JPEG, PNG, and WebP work best. Support for other image formats can vary by browser and AI provider.
Audio and video files are not document attachments in chat. Image and video generation pages use separate reference-media controls. Their accepted formats and limits appear in those controls and can vary by generation model. For example, the standard image and video reference pickers accept JPEG, PNG, or WebP images, with a 10 MB server limit.
Text extraction limitations
Writingmate reads plain-text formats directly. It converts PDF, Word, and Excel files to extracted Markdown before a model reads them. Markdown is plain text that keeps useful structure such as headings, lists, and tables.
Conversion can change or omit:
- page layout, columns, headers, and footers
- charts, diagrams, images, and embedded media
- spreadsheet formulas, formatting, hidden sheets, and macros
- handwriting or text inside low-quality scans
- unusual fonts, symbols, or complex tables
Preview a saved document in Knowledge Management when exact extraction matters. A file marked No text content cannot be added to a chat.
Model, context, and plan limits
Document attachments work with text models after Writingmate extracts the content. Image-generation models are designed to create images, not analyze documents, so attaching a document can switch the selected model. Images require a model with image input.
Extracted text uses tokens. A token is a small piece of text that an AI model reads. Writingmate marks a saved file as Large file when its extracted content is above the message-counting threshold, which is 16,000 tokens by default. A large file can count as multiple messages. Very large documents can use file search so a tool-capable model reads relevant sections instead of receiving the entire document at once.
The available context also includes your prompt, chat history, Agent instructions, and the model's response. Its practical limit can therefore be lower than the upload limit and can vary by model. If the latest request cannot fit, Writingmate asks you to use a smaller file or split it into parts.
File upload and saved knowledge access depend on the workspace plan. Current access rules require Premium or higher for Upload from computer and Pro or higher for Add from knowledge base. Agent and Project availability also depends on the workspace plan. Legacy and promotional plans can differ; use the plan badge in the menu or the live pricing page as the source for your workspace.
Fix a failed or incomplete upload
- Unsupported format: Export the file as PDF, DOCX, TXT, CSV, or another supported format.
- File too large: Compress it or split it into files below the limit for that upload location.
- No text content: Remove encryption, run OCR on a scanned document, or export a text-readable copy, then upload it again.
- Context error: Attach only the relevant pages or sheets, split the task into several messages, or start a new chat.
- Incorrect or missing details: Preview the extracted text, simplify complex tables, and verify the answer against the original file.
- Image rejected: Convert it to JPEG, PNG, or WebP and use a model that supports image input.
- Upload interrupted: Check your connection and retry failed files individually. Successfully uploaded files do not need to be uploaded again.
See Uploading and Managing Files, Using Files, and Message Limits for the full workflow.