The chat panel is where you talk to the AI assistant. Open it with Tools > AI > AI Chat, or with the robot button on the Action Bar. It appears as a tab in the panel beside the main window, so you can watch a person or family record while the conversation goes on next to it.
Type a message in the box at the bottom and send it. The reply streams in as it is generated.
The assistant answers most questions about your project by looking things up rather than guessing, and the panel shows each lookup as it happens — the name of the tool it is using and, for changes, what it changed. This running commentary is worth reading rather than skipping. It is how you catch a misunderstanding early: if you asked about one family and it starts reading another, stop it.
The Cancel button stops a request in progress. Any changes already made remain made — cancelling stops the assistant, it does not roll anything back.
There is a limit on how much work the assistant will do for a single message. A very large instruction may stop partway through with the job unfinished. This is a good reason to work in small batches, which is also how you get better results.
Use the attach button to include an image, a PDF, or a text file with your message. JPEG, PNG, GIF, WebP, PDF, and plain text are accepted.
This is one of the assistant’s most useful abilities. Attach a photographed headstone, a certificate, or a census page and ask it to read what it can; attach a page of your own typed notes and ask it to turn the relevant parts into facts on a person.
Treat the results as a first draft. A confident misreading of a difficult hand looks exactly like a correct reading, so check the attachment yourself before accepting anything into your project.
The Clear button starts a fresh conversation. Do this whenever you move to a new topic. Everything said so far is re-sent with each new message, so a long conversation is both slower and more expensive than a short one — and an assistant still carrying twenty minutes of talk about the Wedgwood family is more likely to muddle it into a question about the Darwins.
The conversation is not part of your project. It is not saved in the file and is not there when you next open Pedigree Forge. If a reply contains something you want to keep — a summary, a list of discrepancies to work through — copy it somewhere useful, such as a note or a to-do item, before you clear.
The assistant sees only the project that is currently open, and it sees it as it is now, including edits you have just made.
It can also drive the application: ask it to show you a person, family, source, or note and it will navigate the main window there. "Show me Emma Darwin" is often quicker than searching, and it is a good habit after the assistant has changed something — ask to be taken to the record so you can look at the result.