Attachments and file uploads

Attachments let you add images, documents, spreadsheets, audio, and plain-text files directly to a chat message, so the assistant can read them and answer questions about their contents in that conversation. This page explains what the composer shows, the file kinds and limits you can send, how to read each attachment's status, and how the assistant uses what you upload. Attachments are part of the AI Console and are available on the Free, Pro Lite, and Pro plans.

Before you begin

Sign in to your account. Attachments are sent from a chat in the AI Console.
Have a file ready in a supported format. The accepted kinds and the size limit are listed under Inputs.
To record a voice note, use a secure (https) connection and grant microphone permission when prompted.

Quickstart

Open a chat
Go to the AI Console and start a new chat or select a saved one.
Attach one or more files
Select the paperclip in the composer to open the file picker, or drag and drop files onto the composer. You can select several files at once.
Check the staged chips
Each file appears as a chip above the text box with its type icon, name, and size. Wait for the reading spinner to clear so the assistant has the contents.
Type a question and send
Write what you want the assistant to do with the files, then send the message. The files render above your message bubble.

What it shows

The composer holds the attachment controls and the staged files. The parts are described in the following list.

Paperclip button: opens the file picker. You can also drag and drop, or select several files at once.
At-sign button: mentions a saved subagent for the next message. See Subagents.
Microphone button: records a short voice note. The note is uploaded as an audio attachment and transcribed.
Attachment chips: appear above the text box once files are staged. Each chip shows a type icon (an image thumbnail, a document icon, or a microphone), the file name, and the file size.
Per-chip status: a spinner while the file is being read, a warning icon if reading failed, and a lock icon if a file has refused to be read several times. Select a failed chip to expand it inline, read the reason, and choose Retry or Remove.
Sent attachments: after you send, the files render above your message bubble. Images become clickable thumbnails that open full size. Other files become click-to-open chips labeled with their type (PDF, Word, Excel, PowerPoint, CSV, Text, JSON, XML, YAML, Audio) and size.
Save to knowledge: an eligible document carries a bookmark action so you can save it into your knowledge library in one selection. See Your knowledge and retrieval.

Inputs

File picker or drag-and-drop: accepted kinds are images, audio, PDF, Word (.doc and .docx), Excel (.xls and .xlsx), PowerPoint (.ppt and .pptx), CSV, and plain-text formats (.txt, .md, .json, .csv, .xml, and .yaml or .yml). An unsupported type is rejected with a message listing what is allowed.
Size limit: 50 MB per file. A larger file is rejected.
Count caps per message: up to 5 images and up to 5 non-image files in the same message. The two limits are separate, so you can send 5 images and 5 documents together. Going over a cap shows "Max 5 images per message" or "Max 5 files per message."
Remove control: the X on each chip removes a staged file before you send.
Retry: a failed file offers an inline retry.
Voice note: select the microphone to start, then select stop to finish. Recording needs a secure (https) connection and microphone permission.

Attachments have no other settings or defaults. They are scoped to the single message you send them with.

Outputs and how to read them

Staged chips (before send): name, size, and a live status. A spinner means the assistant is reading the file. A warning means it could not be read. A lock means the file failed too many times, so try a different format.
Sent attachments (after send): image thumbnails open the original in a new tab. File chips open or download the original. A chip can read "failed" if reading did not complete, or "in memory" if you have already saved the file to your knowledge library.
The assistant's reply: reflects what it could read. For an image it sees the picture directly. For a document or spreadsheet it reads the extracted text. If a file failed to read, the assistant does not have its contents, which is why the retry control exists.

How it is calculated

Not applicable. There is no score or computation in the attachment mechanic. The assistant reads the file contents and uses them as context for its answer.

Data and timing

The data is whatever you upload. A file is read on upload and is available to the assistant in that chat within the few seconds the reading spinner is shown. Attachments are not market data and do not refresh. Each is a snapshot of the file at the moment you sent it. No AlphaVantage, EDGAR, or Nasdaq-via-AlphaVantage data is involved unless your own file happens to contain it.

Use cases

Drop in an earnings deck, a 10-K excerpt, or a research PDF and ask the assistant to summarize it or pull specific figures.
Upload a spreadsheet of holdings or transactions and ask for a plain-English read of what it contains.
Paste a chart or a screenshot as an image and ask the assistant to interpret it.
Record a short voice note instead of typing a long question.

Limitations and disclosures

An attached file is read only inside the chat you send it with. To reuse a document across chats, save it to your knowledge library. Any analysis the assistant produces from an attached file is research, not a recommendation. Upload only material you have the right to use.

GNG Research provides equity research and educational tools, not investment advice. Nothing on the platform is a recommendation to buy or sell any security. Do your own research and consider your circumstances before making any investment decision.