SuperAPI AI Directory
Partner中文AI newsFavorites
Skillsutopia

utopia

Agents & models

View repository

Updated 10-29 04:36

About this skill

Utopia

Instructions

  1. Determine whether the user wants an app deployment, local development, data-backup operation, or read-only architecture guidance. Inspect the pinned release/source and deployed config path before acting.
  2. The README documents docker compose --profile app up -d; it exposes the app on localhost:1516 and initializes a public knowledge base on first account registration. This is an example only. Do not deploy, start services, register accounts, or change database/config state without explicit approval.
  3. Treat ingestion, extraction, model calls, database mounts, MCP writes, migrations, backup/restore, and public deployment as separate approval gates. Confirm data classification, target base/database, provider, cost, and destination before any external transfer or mutation.
  4. Preserve the app's review boundary: uncertain facts and agent write proposals require human adjudication. Do not claim an MCP connection can write directly unless source/docs for the exact operation support it.

Examples

  • “Explain Utopia's bitemporal model and what its audit ledger retains.”
  • “Plan a local Docker deployment, but do not start it.”
  • “Configure a read-only MCP client for one knowledge base.”

Best practices

  • The first registered account becomes system administrator, and the initial knowledge base is public in the documented quick start; flag this before signup and verify current deployment settings.
  • Keep secrets out of chat and reports. Utopia reads UTOPIA_ environment configuration; secret.key protects stored credentials and must not be bundled casually with data backups.
  • Read references/upstream.md for backup/restore caveats, exact commands and deployment scope. Example down -v is destructive and must never be run without explicit approval.

References