Working with Files in ThreoAI

Overview

ThreoAI lets you attach files straight to a conversation. Once a file is in, the AI can read it, summarize it, pull out details, and answer questions about it. This works with documents, images, and spreadsheets, in any chat - the home chat, a Project, or an Expert or Agent conversation.

Attaching files takes you past text-only prompts and brings your real documents, data, and images into the chat. Summarize a long report, pull figures out of a spreadsheet, compare two contracts, or describe what an image shows - ThreoAI reads the content and answers from it.

Attaching a file

To add a file to the current conversation:

1. Open a chat or start a new one. File attachments work in any chat - a general conversation, a Project, or an Expert or Agent chat.
2. Select the attachment control in the composer.
3. Pick one or more files from your device. You can select several at once.
4. The files appear as thumbnails above the composer, each showing its name.
5. Type what you want done with them - for example, “Summarize this document” or “Extract every line item and its cost from this invoice.”
6. Press Enter to send your message and the files together.

The AI reads the files as part of your message and answers from their content and your instructions.

Supported file types

ThreoAI handles the common formats:

Category	Supported formats
Documents	PDF, DOCX, PPTX, TXT
Spreadsheets	XLSX, CSV
Images	PNG, JPG, JPEG, GIF, WEBP

Each format is handled a little differently:

• PDF is read as text. The AI can summarize it, pull out sections, answer questions, and compare several PDFs.
• DOCX (Word) is read like a PDF. The AI can analyze, summarize, or reformat the text.
• PPTX (PowerPoint) is read for the text across slides. The AI can summarize the deck, pull key points, or describe individual slides.
• TXT (plain text) is read directly - good for raw text, code, or log files.
• XLSX (Excel) is read as structured data. The AI can spot trends, summarize, total things up, and answer questions about the numbers and categories.
• CSV is read as tabular data, much like XLSX, and suits data exports and analysis.
• Images (PNG, JPG, JPEG, GIF, WEBP) are read visually. The AI can describe what an image shows, read text out of a screenshot or photo, identify objects and diagrams, and answer questions about it.

File size limits come from your organization’s configuration. If a file will not upload, split it into smaller parts or ask your administrator about the limit.

Previewing files in chat

After you upload a file and get an answer, you can still review the original in ThreoAI:

• Select the attachment thumbnail or name to open a preview.
• PDFs render inline so you can scroll the pages without leaving the chat - handy for checking the AI’s summary against the source.
• Images show inline at a readable size; select one for a larger view.
• Other types may show a file icon with an option to download the original.

Downloading files

Files in a conversation - ones you attached or ones the AI produced, such as a reformatted document or an exported file - can be downloaded straight from the chat:

1. Find the attachment or AI-generated file in the thread.
2. Select the download icon or the file name.
3. The file saves to your device’s downloads folder.

To turn chat content into a brand-new Word, PDF, PowerPoint, CSV, or Markdown file, ask the model to export it - see document export in Tools and Capabilities.

Tips for working with files

How you frame the request shapes the result. A few practical pointers:

Be specific

Skip “analyze this file” and say what to focus on:

• Instead of “Analyze this file,” try “Summarize the key findings in this quarterly report, focusing on revenue trends and operating expenses.”
• Instead of “Look at this spreadsheet,” try “Total the spending by category and tell me which category jumped the most month over month.”
• Instead of “What does this say?”, try “Pull every action item from these meeting notes and list each with its owner and due date.”

Work with several files

You can attach more than one file to a message. This helps when you want to:

• Compare documents - upload two versions of a contract and ask for the differences.
• Cross-reference - upload a policy and an audit report and ask whether the findings match the policy.
• Consolidate - upload several related documents and ask for one summary that pulls from all of them.

Work with large documents

For very long files, the AI may not cover everything in one pass. To get the best results:

• Point it at specific sections: “What does section 3 say about refund policies?”
• Ask for a high-level summary first, then drill in with follow-ups.
• For an enormous document, split it into smaller files and send them one at a time.

Work with images

• Describe content - “What does this screenshot show?” or “Describe this diagram.”
• Read text - “Transcribe the text in this image” or “What does this error message say?”
• Read charts - “What trend is in this bar chart?” or “Summarize this pie chart.”
• Identify objects - “What products are in this photo?” or “What device is pictured?”

Work with spreadsheets

With XLSX or CSV files you can ask the AI to:

• Summarize the data and surface key trends
• Total things up or compute averages and other aggregates
• Find outliers or anomalies
• Compare columns, rows, or time periods
• Draw out insights or recommendations from the numbers

For the cleanest results, give your spreadsheet clear column headers so the AI can read its structure.

Files in an Expert

When you build an Expert, you can upload training documents that become part of its permanent knowledge base. That is different from attaching a file to one chat:

• Conversation attachments are one-off. The AI uses them for the current chat only.
• Expert training documents are uploaded once and stay part of the Expert’s knowledge, so every conversation with it can draw on them without re-uploading.

This makes training documents ideal for an assistant that answers from a fixed set of material - company policies, product manuals, SOPs, or knowledge-base articles. See Creating an Expert for how to upload them.

Troubleshooting

A file will not upload.

• Check the type is supported (PDF, DOCX, PPTX, TXT, XLSX, CSV, PNG, JPG, JPEG, GIF, WEBP).
• Check it is within your organization’s size limit. Ask your administrator if you are unsure.
• Split a large file into smaller parts and upload them separately.

The AI does not read the file right.

• Make sure the file is not password-protected or encrypted - the AI cannot open protected files.
• For scanned PDFs (images of text), reading can be patchy. Use a higher-quality scan or a text-based PDF.
• Rephrase your prompt to name the section or content you care about.

The summary feels incomplete for a long document.

• A large document can exceed what the model takes in one pass. Ask about specific sections or pages.
• Split the work across several follow-up questions to cover the whole thing.

Related

• Starting a Chat - the composer and attachment control
• Tools and Capabilities - document export and other built-in tools
• Creating an Expert - upload training documents as an Expert’s knowledge base
• Navigating the Interface - the composer and its controls
• Projects - organize file-heavy conversations into a workspace