Capturing Artifacts

The two-button flow

On a supported AI chat page, click the Remoria button in the bottom-right corner. The panel opens with two actions:

1. Summarize — Pastes a specialized prompt into the chat composer. You send it. The AI extracts a structured artifact from the conversation.
2. Capture — After the AI finishes, saves that structured reply to your dashboard.

Remoria does not send the prompt or wait for the answer for you — that stays under your control. If the structured reply names no collection, the artifact goes to Inbox.

What gets captured

Each artifact includes the following fields extracted by the AI:

• Title — A descriptive title for the artifact.
• Summary — A concise, human-readable overview of the conversation's key content.
• Key Points — Bullet-point takeaways and actionable insights.
• AI Context — A structured continuation block designed for AI consumption (see below).
• Collection — A suggested collection to organize the artifact into.
• Tags — Relevant tags for filtering and discovery.

AI Context for continuation

In addition to the human-readable summary, every capture includes a structured AI Context block. This is not meant for you to read — it's optimized for another AI to pick up exactly where the conversation left off.

It contains three sections:

• PROBLEM — The original question or goal (1–2 sentences).
• KEY EXCHANGES — 3–8 condensed Q&A pairs with concrete specifics (decisions, numbers, code).
• STATE — Decisions made, action items, open questions, and unresolved threads.

When you inject an artifact into a new chat via @rem, Remoria uses this AI Context instead of the summary — giving the new AI a precise, actionable starting point rather than a vague overview.

You can view, edit, and copy the AI Context from the artifact detail drawer in your Library. Click any artifact, then expand the “AI Context” section to see or modify what gets injected.

Incremental capture

You do not have to wait until a conversation is finished to capture. The extension tracks your position in the conversation (message index), so you can capture at any point and then capture again later to pick up new content. Each incremental capture is linked to the same conversation thread.

Deduplication

Remoria automatically deduplicates artifacts based on a SHA-256 hash of the raw structured block. If you accidentally capture the same content twice, only one copy is stored. This keeps your library clean without any extra effort.