adrian-init

Walk the user through Adrian setup in a SINGLE run - concisely, step by step.

CRITICAL - do the WHOLE flow in ONE invocation. NEVER finish by telling the user to run /adrian-init again. If a check fails, that is your cue to CONTINUE with the setup steps below yourself. In particular: if you run verify first and it FAILS, do NOT stop and re-prompt - go straight to step 1 and continue the setup in the same run.

SECURITY - non-negotiable: NEVER ask the user to paste their API key into the chat, and NEVER print, cat, or echo the key value or the contents of the .env file. The key stays out of this conversation. You only ever run the verify command (which reports OK/FAIL, not the key).

0. (Optional) Short-circuit if already configured

You MAY first run adrian-python -m adrian_cc.agent verify:

• OK → Adrian is already set up; say so and stop. Nothing else to do.
• FAIL (or not configured) → this is expected for first-time setup. Do NOT stop, and do NOT ask the user to re-run the command - immediately continue to step 1 in this same run.

1. Choose the backend

Use the AskUserQuestion tool (header "Backend"). The tool auto-adds an "Other" free-text choice = the custom path; word it so the user knows Other = a custom self-hosted URL:

• "Adrian Cloud (hosted)" - managed backend. (Recommended)
• "Self-hosted (local)" - a local OSS backend on the default host/port.
• (built-in "Other") - the user types a custom self-hosted host/port or ws:// URL.

Resolve the answer to a WebSocket URL:

• Adrian Cloud (hosted) → wss://adrian.secureagentics.ai/ws
• Self-hosted (local) → ws://localhost:8080/ws
• "Other" free text → normalize whatever they typed:
  • a full ws:// or wss:// URL → use as-is
  • host:port → ws://host:port/ws
  • host only → ws://host:8080/ws (Defaults: port 8080, path /ws, scheme ws:// unless they gave wss://. If their input is ambiguous, confirm the resolved URL before writing.)

2. Write the config (commented, with an API-key PLACEHOLDER)

Create ~/.adrian/ if needed and write ~/.adrian/.env with ADRIAN_WS_URL set to the resolved URL and a short # comment above each variable explaining what it's for, so the user can understand and hand-tune the settings they weren't asked about:

# Adrian backend WebSocket URL (set from your backend choice above).
ADRIAN_WS_URL=<resolved>

# Your Adrian API key - replace this placeholder with your real key:
# adr_live_... (Adrian Cloud, from your dashboard) or adr_local_... (self-hosted).
ADRIAN_API_KEY=REPLACE_WITH_YOUR_ADRIAN_KEY

# Seconds to wait for a verdict before giving up on a tool call.
ADRIAN_CC_VERDICT_TIMEOUT=15

# What happens if Adrian can't reach the backend or a verdict times out:
#   true  = fail open  - allow the tool (best-effort monitoring, never blocks work)
#   false = fail closed - block the tool (max security, but blocks work when the backend is down)
ADRIAN_CC_FAIL_OPEN=true

The ADRIAN_API_KEY line is a PLACEHOLDER written for the user, so all they have to do is swap the value. Preserve an existing real key: if ~/.adrian/.env already has an ADRIAN_API_KEY line, do NOT overwrite it with the placeholder - detect this silently with grep -q '^ADRIAN_API_KEY=' ~/.adrian/.env (exit status only; it never prints the value). Never print the file contents or the key. Tell the user the exact file path you wrote.