What Are Artifacts?

Artifacts are the platform's way of saving discrete pieces of content — code, documents, lists, data — separately from the conversations that produced them. Think of them as a personal vault of named, searchable, reusable building blocks.

What you can save as an artifact

When you save, the platform auto-detects the type from content shape (JSON-shaped → JSON, has function definitions → code, starts with # → markdown, otherwise text). You can override the detection.

How saving works

Two main paths to save:

1. Save from chat reply. Most chassis show a small "🔖 Save" button on assistant messages. Click it to open the save dialog.
2. Save from anywhere. Some chassis surface a "+ New artifact" button in the artifact panel for free-form creation.

The save dialog asks for:

• Title (auto-suggested from content)
• Type (auto-detected, editable)
• Tags (comma-separated, optional)

Saved artifacts appear instantly in your artifact vault.

The privacy story (important)

Artifacts are stored locally in your browser's IndexedDB, not on Nyquest's servers. The server only sees:

• Auth checks (verifying you're signed in)
• Embedding computation (when you save, an embedding is computed for semantic search; this is sent to the embedding model and discarded)

What this means:

• ✅ Artifacts are on your device, not in our database
• ✅ Multiple devices = multiple separate vaults (no cloud sync currently)
• ✅ Sign out = vault stays on the device; sign in elsewhere = empty vault on that device
• ⚠ Clearing browser data wipes the vault — back up via Export
• ⚠ No vault sync between devices yet (roadmap)

If you want cloud-stored artifacts that sync across devices, that's not built. The local-first design is intentional for privacy. A future opt-in cloud sync is on the roadmap.

What artifacts are NOT

• Not file uploads. Currently you can't drag-drop a PDF and have it become an artifact. The vault is for content text/code/etc., not arbitrary files. (File-attachment-as-artifact is a separate roadmap item.)
• Not chat messages. A chat message is conversational; an artifact is reusable. You explicitly choose to extract things.
• Not memories. Memories (facts, summaries) are derived automatically from conversations. Artifacts are explicit user-saved units.
• Not shared. Each user's vault is private to that user, on that device. No multi-user sharing yet.

Artifacts and memory are separate

Memories are recalled automatically; artifacts never are. The vault lives in your browser's local storage, so the server-side recall layer cannot see it — a saved artifact reaches a conversation only when you inject it yourself. See Using Artifacts in Chat.

Typical workflows

• "Code library" — save useful functions as Code artifacts with tags like python, auth, parser. Inject one when you are working on something related.
• "Drafts and snippets" — save reply drafts you might want to refine later. Markdown type, tag with the project.
• "Reference material" — save key passages from articles or research. Text/markdown type.
• "API catalog" — save useful API responses as JSON artifacts when iterating against an external service.
• "Task lists" — list-type artifacts for ongoing checklists, refined as you progress.

Where to next

• Vault Overview — the vault UI, search, tags
• Generated vs Saved — model output vs user-saved
• Using Artifacts in Chat — explicit injection
• Search and Tags — finding things later
• Export and Import — portability + backup