Deploy a PromptPack to Omnia

This tutorial shows how to deploy a PromptPack to an Omnia workspace using the promptarena-deploy-omnia deploy adapter. The fastest path is to export a deploy profile from the dashboard, which bootstraps the adapter’s config: block for you.

Bootstrap your deploy config (recommended)

In the dashboard, open your workspace → Settings → Deploy → Export deploy profile.

This generates a ready-to-paste config: block containing:

• api_endpoint — your dashboard’s external URL.
• workspace — the target workspace name.
• api_token — a show-once omnia_sk_ token scoped to this workspace: it can only deploy here (not to any other workspace you can access), and it acts with your own role in this workspace. The first export mints the deploy-<workspace> key. On a later export, if that key still exists the dialog asks whether to regenerate it (revokes the old key and mints a fresh token — any previously downloaded profile for this workspace stops working) or reuse the token you saved, so re-exporting doesn’t pile up duplicate keys.
• a configure step listing the Ready Providers and SkillSources in the workspace. Only llm-role Providers are exported — embedding, TTS, STT and image Providers are workspace-level services (consumed by memory-api and the like), not per-agent extras, and bundling them into a deployment breaks the pack at first request. Non-Ready resources (Unavailable, Error, still syncing) are also excluded, since a deployment that references one fails. You check which to include and pick one LLM as the default provider; the runtime requires a provider bound under default as its primary, so the export marks your choice as name: default (with ref pointing at the real Provider).

Paste the generated block into your arena deploy config:

config:
  api_endpoint: https://omnia.example.com
  workspace: team-acme
  api_token: omnia_sk_...        # freshly minted, show-once, revocable
  # --- discovered in this workspace (pick what you need) ---
  providers:
    - { name: default, ref: claude-sonnet, role: llm }
  skills:
    - docs-search

The token is a real credential

The minted api_token is shown once. Store it securely and revoke it from Settings → API keys when you no longer need it. It is confined to this workspace, but anyone holding it can act with your role here.

If your deployment uses a read-only API-key store, the dashboard still exports the discovery menu but leaves the api_token as a placeholder — mint a token manually under Settings → API keys and paste it in.

Autoconfigure via browser login

When the dashboard runs in OAuth mode, the deploy CLI can skip the manual copy-paste and mint everything through a browser-login flow, the same shape as gcloud auth login or gh auth login. This is the fastest path for a local workstation:

1. The CLI opens the dashboard’s browser-login entry point (GET /api/cli/authorize), passing a loopback callback URL and a random state.
2. Your browser goes through the normal OIDC login (if you aren’t already signed in), then lands on a workspace picker (/cli/select). You choose the target workspace.
3. Granting (POST /api/cli/grant) re-checks your editor access to that workspace, mints a one-time exchange code, and redirects the browser back to the CLI’s loopback callback with ?code=&state=. No token crosses the browser — only the short-lived code.
4. The CLI exchanges the code on a back channel (POST /api/cli/token). That endpoint mints a workspace-scoped, short-lived omnia_sk_ token (default TTL 1 hour, OMNIA_AUTH_CLI_TOKEN_TTL_SECONDS) and returns it together with the deploy profile — the same api_endpoint / workspace / providers / skills discovery menu the dashboard export produces. The CLI writes the ready-to-use config: block for you.

The exact CLI command that drives this lives in the promptarena deploy tooling; the endpoints above are the Omnia-side contract. Browser login requires the dashboard to be in OAuth mode and API-key creation to be enabled — otherwise fall back to the dashboard export or manual setup.

Manual setup (fallback)

If you prefer to assemble the config by hand:

1. Mint an API key — go to Settings → API keys, create a key, and copy the omnia_sk_ value (shown once).
2. Copy the connection details — your dashboard URL is the api_endpoint, and the workspace name is the workspace.
3. Reference your resources — list the Providers and SkillSources in the workspace (Settings → Services) and fill in providers / skills with their exact names. A typo in a Provider name fails the deploy at plan time.

Deploy

With the config: block in place, run the deploy adapter as usual:

promptarena deploy --provider omnia

The adapter validates the config (validate_config), plans the changes, and applies them to your workspace.