> ## Documentation Index
> Fetch the complete documentation index at: https://conveo.ai/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Give the Assistant Documents and Images

> Attach research briefs and reference images, understand supported files, and manage conversation context.

Attach background materials when the assistant needs a brief, methodology, reference image, or existing document to answer your question. Say how the attachment should be used: background context, a format example, or material to inspect.

An attachment is not automatically participant evidence. Ask the assistant to distinguish supplied context from findings in the selected interview studies.

## Upload a document

In Talk to Your Data, select the **+** menu and choose **Upload a document**. Direct chat document uploads support **PDF** and **DOCX**. Wait for parsing to finish and the attachment to appear before sending your question.

Use a document with selectable text where possible. The chat's initial PDF reader extracts its text layer, and the DOCX reader extracts document text rather than embedded images. Scanned pages, complex layouts, and image-only diagrams may not be represented accurately. Upload an important visual separately as an image and ask the assistant to check it.

Document text consumes conversation context. A large attachment can bring a conversation closer to its context limit; use the relevant sections rather than assuming an entire document can always be retained verbatim in a long chat.

## Upload an image

Choose **Upload an image**, paste a copied image, or drag a supported image onto the composer where offered.

| Requirement          | Supported value                |
| -------------------- | ------------------------------ |
| Formats              | JPEG, PNG, GIF, WebP           |
| Maximum file size    | 10 MB per image                |
| Maximum upload batch | 5 images in one upload request |

Wait for all image uploads to finish before sending. A failed upload is not part of the assistant's message; review the error and retry or choose a supported file. The five-image limit applies to one upload batch, not a documented lifetime limit for the conversation.

The Talk to Your Data composer does not provide the study-design chat's direct video-upload workflow. Use [interview data](/docs/interview-data/interview-grid-and-transcripts) or the appropriate study workflow for recordings.

## Reuse document-library content

When your organization has an enabled document library with available content, the **+** menu can offer **Add from document library**. Select an accessible document and send the question that should use it. This option is not available on every assistant surface or to read-only conversation viewers.

A library document can use formats beyond the direct chat-upload formats because its text has already been parsed through the library. The presence of a library spreadsheet or presentation does not mean the chat's upload picker accepts XLSX or PPTX files directly.

## Understand persistence and access

An attachment added to a conversation can be saved as reusable context through the applicable study or organization workflow. For a single-study conversation, background persistence of the original document requires study editing access. The chat may still contain its initially parsed text if that separate persistence step fails.

A document appearing in a sent message is therefore not proof that its original file is available in a library. Review the relevant context-library entry when you need persistent reuse.

Removing an unsent attachment from the composer prevents it from being included in that outgoing message. It is not a deletion command for a previously uploaded library document or a previously sent message. For access to shared conversations and their contents, see [conversation sharing](/docs/settings-and-management/permissions-and-sharing#sharing-conversations).

## Troubleshooting

* **No attachment appears:** wait for parsing; check that the document is PDF or DOCX and can be opened locally.
* **The answer misses a diagram:** upload the relevant visual separately and identify what needs inspection.
* **An image upload fails:** check file format, size, and the number selected together.
* **The library option is absent:** confirm library availability, document availability, and conversation editing access.
* **The chat reaches its context limit:** use the offered summary-to-new-chat workflow and restate any essential constraints.

See [Talk to Your Data](/docs/interview-data/talk-to-your-data) for choosing studies and reviewing evidence.