Gary Chat Features
Sources Available in Gary Conversations
Gary can draw from three categories of files during a conversation, in addition to your SmartList data:
Conversation Documents — files you upload directly into the Gary chat window for that session
System Documents — files attached at the agent or prompt level by your administrator
File Library Files — files stored in your organization's CORAS File Library
Linked Sources Load Automatically
Sources linked to a Smart Agent or Prompt are available automatically when you start that conversation. You don't need to re-attach them.
If Gary references information you didn't provide in this conversation, it's drawing from a source your administrator has attached to the agent or prompt you're using. This is intentional — your team's reference documents are pre-loaded so you don't have to upload them manually. The file names and contents are not visible in your chat.
Some Smart Agents and Prompts are locked by your administrator and carry confidential global sources. These are supplied to Gary's context automatically and are never shown in your chat — you will not see them in the Sources panel. Gary uses them to improve answer quality, but they are not listed because they are not yours to manage.
Upload a file for a single conversation:
In the Gary chat window, click the + icon at the bottom left
Select a PDF, Word (.docx), PowerPoint (.pptx), plain text (.txt), Markdown (.md), or JSON file
Wait a few seconds for Gary to process the file
Ask Gary questions about the content
You can also drag and drop a file directly into the text box. Files uploaded this way are available for the current conversation only.
Uploading Files as Context
You can upload files directly into a Gary conversation so Gary can read and reference them in its responses.
Supported file types: PDF, Word (.docx), PowerPoint (.pptx), Excel (.xlsx), plain text (.txt), Markdown (.md), and JSON.
Upload a file for a single conversation:
In the Gary chat window, click the + icon at the bottom left.
Select your file and wait a few seconds for Gary to process it.
Ask Gary questions about the content.
You can also drag and drop a file directly into the text box. Files uploaded this way are available for the current conversation only.
Oversized Excel files: If an .xlsx file exceeds the size limit, CORAS will not upload it as a direct AI source. In Nodes, you'll be directed to use Central Ingest — see Getting Data into CORAS. In Coras Light, the system redirects you to create a dynamic source instead.
To use a File Library file:
Click the source icon in the chat window, select File Library, choose your file, and ask your question. File Library files persist independently and stay available across conversations.
Note: When you upload a Word, PowerPoint, or Excel file, CORAS converts it to PDF to enable AI processing. Gary will refer to the file as a PDF in its responses — this is expected. You'll see a notice when the conversion happens: "Please note that the Agent will refer to your file as PDF because it will be converted into PDF to enable AI support."
Switch Agents Mid-Conversation
You can switch to a different Smart Agent without starting your conversation over. When you start a new Ask Gary session from within an existing conversation, CORAS gives you the option to carry your current context — the conversation history, data references, and findings — into the new session with the new agent.
How to switch agents:
In your current Gary conversation, click "+ New Chat."
Select the agent you want to switch to.
Choose whether to carry context from the current conversation into the new session.
Click Start — Gary opens with the new agent and your selected context.
Carry context forward when your new agent needs to build on prior analysis — for example, moving from a Data Scientist agent to an Executive Briefing agent to turn findings into a leadership summary.
Start fresh when you want the new agent to approach your data without influence from the previous conversation.
Speak to Gary (Voice Input)
You can speak to Gary instead of typing. Gary transcribes your speech and places the text in the prompt box for review before you send it.
1. In the Gary chat window, click the microphone icon in the prompt box.
2. Allow microphone access if your browser prompts you, then speak your request clearly.
3. Gary transcribes your speech and places the text in the composer.
4. Review and edit the transcribed text as needed, then send it like any typed message.
Gary doesn't receive your audio directly — only the transcribed text. Review before sending to catch any transcription errors, especially for technical terms, acronyms, or proper nouns. If a chat returns an error, you can delete that conversation or select it to retry without losing your session context.
Stopping Gary mid-response
If Gary is generating a response and you want to stop it, select the stop control in the chat window or press ESC. Gary will ask you to confirm before stopping.
The confirmation reads: "Do you want AI to stop generating the response?"
Two options appear:
Stop generating — ends the current response immediately. Gary's partial output stays visible.
Continue generating — dismisses the prompt and Gary keeps going.
Keyboard shortcuts: Enter = Stop generating. ESC = Continue generating.
Staging Your Next Prompt while Gary responds
You don't have to wait for Gary to finish before composing your next message. While a response is generating, you can type in the composer and attach files — Gary holds your input ready until the current response completes.
The send button is disabled while Gary is responding. A notice at the top of the composer confirms Gary is still working. When the response finishes, the notice clears and your staged message is ready to send.
Your Draft Is Saved Automatically
If you start typing a message or stage sources in the composer and then leave — switching conversations, closing the tab, or coming back on a different device — your unsent draft is saved. When you return, the prompt text and any staged sources are restored so you can pick up where you left off.
• Drafts are saved per conversation — a draft in one conversation never affects another.
• Only sources already stored in the system are restored. Unavailable sources are skipped; the rest of the draft still loads.
• A draft that has not been updated for 7 days is deleted automatically.
• Attachments that have not finished uploading are not saved as part of a draft.
Export a Conversation
You can export any Gary conversation for records, briefings, or future use.
Export as PDF — produces a clean, readable document with code-execution blocks and system content masked. Suitable for briefings and formal review.
Export as JSON — preserves the full conversation structure. Use for records or to import as context into a future Gary session.
To export, click the Export option in the Gary conversation window.
To save a single response to your personal collection without exporting the full conversation, see Save Responses with Memos.
Submit Feedback on a Gary Response
You can submit anonymous feedback on any Gary AI response directly from the chat window. Feedback goes to your organization's support team — not to the AI model — and helps improve future answers.
Important: Do not include personal, confidential, or classified information in your feedback.
How to submit feedback
In a Gary chat, locate the response you want to comment on.
Click the feedback icon on the response.
A feedback form opens. Read the disclaimer, then type your note.
Click Submit.
Your feedback is submitted anonymously — your name is not shown to the reviewing team.