NNyquest docs

Conversations and History

Signed-in chat saves exchanges as conversations so you can return to them. Saving is automatic; you do not need a Save button.

Open an existing conversation

On desktop, open the Conversations / History control in the top bar. On mobile, open the menu and choose a conversation from the list. Names can vary with your chassis.

The list shows recent conversations, with projects above them. Choosing a project filters its conversations; choose All conversations (or the equivalent chassis label) to leave that filter. An older conversation missing from the recent list has not necessarily been deleted.

What the model can see

The saved transcript and the model's working context are different. Nyquest supplies recent messages, an available summary, and selected eligible memory within a context budget. A long conversation is not necessarily sent in full on every request.

Starting a new conversation begins a new transcript. Scoped memory can still contribute across conversations, but it does not guarantee that every old detail will appear. Reopen the original conversation or provide exact context when it matters. See How Memory Works.

Interrupted answers

If a reply stops before completion is confirmed, the text received so far stays visible with an Incomplete answer notice. A request with no usable completed answer can show Answer unavailable. The notice explains a recorded cause when available, such as a connection interruption, timeout, provider error, or model output limit.

Use Reload saved answer, when offered, to check the server's saved result without sending another model request. It can recover a stored answer after the browser lost its connection; it cannot complete text the model never produced. A saved result can still be incomplete. If the result is not ready, wait briefly and check again. A failed reload keeps the output already visible.

Older replies may have no completion status, so the absence of a notice on an old message does not prove it finished. Asking the model to continue is a new request and is separate from reloading the saved answer.

Attachments in a conversation

Use the attachment control in regular chat, select the intended files, and wait until they are ready before sending. An upload error must be retried or the attachment removed before it can be sent.

Sent attachments stay associated with their message when you reopen the conversation. Supported photos can be used in follow-up questions in that conversation; the model must support image input. Attaching a photo does not make its contents a permanent memory fact.

For photos, use JPEG, PNG, or WebP. Export HEIC/HEIF to a supported format first. Limits are 25 MiB per file, 40 megapixels decoded, and five attachments per message. The attachment workflow is not available in Splicer or stateless API-key chat. See Using Artifacts in Chat for file guidance.

Rename or delete

Use a conversation row's Rename control to change its title, then save. Renaming changes its label, not its messages or context.

Use Delete and confirm to remove a conversation. Deletion is permanent in the app and removes associated source data. Other memory entries can remain, so deleting one conversation is not a full memory reset. See Memory Privacy and Deletion.

Keep your own copy

Use Settings → Download my data for a JSON download in your browser. It includes conversation text and saved facts along with other account data. The current export is capped at 1,000 conversations, 100 messages per conversation, and 500 facts. Keep separate copies of important material and the original attachment files.

To share a transcript, copy the relevant text yourself. There is currently no read-only conversation sharing link.

Where to next