#Admin console overview
admin.velgent.com is where administrators configure their tenant: mint
API keys, author policies and prompts, tune the engine's PII and
intent-guard settings, and review what the AI is actually doing in
production.
You'll be invited the first time someone provisions your organisation — clicking the Magic Link in your invitation email signs you in and lands you on the dashboard. After that, sign in any time at admin.velgent.com.
#Surfaces at a glance
The sidebar groups everything by what it does:
| Surface | What it's for |
|---|---|
Dashboard (/) | At-a-glance counts (active policies, prompt templates) + recent activity feed. Default landing page. |
Prompts (/prompts) | Author and version the system + user prompts the engine uses at runtime. Edits create a new immutable version; publishing flips the active pointer atomically. |
Policies (/policies) | Author policy-engine policies in plain English. The compiler turns them into a DSL; publishing starts gating runtime decisions. |
Engine settings (/pii-rules) | Per-tenant PII profile (Strict / Standard / Permissive / Custom), intent guard, custom regex recognisers, LLM provider routing. |
Prompt logs (/prompt-logs) | Audit feed of every authenticated call to the AI engine. Inputs and outputs are previews — already redacted by Presidio before storage. |
Security events (/security-events) | Intent-guard violations, prompt-injection detections, schema-drift notes. Surfaces what your guardrails actually caught. |
API keys (/api-keys) | Self-service mint and rotate keys for the tenant in scope. See Setting up API keys for the step-by-step. |
Extract templates (/extract-templates) | Publish, version, and AI-generate ExtractionSchemas for /api/v1/extract. Paste a sample document and let Velgent propose the schema, run the playground, then save. See Data Extractor → Templates. |
Organisations (/orgs) | Root-only — manage tenants and their first admin. Org_admins won't see this in their sidebar. |
Velgent has two admin roles:
- org_admin — configures one tenant. The role you get when you accept an invitation. Can do everything except cross-tenant operations.
- root — the platform owner (Velgent staff). Manages every tenant from one console.
The sidebar surfaces are the same; root sees a few extra entries (and an org picker at the top so they can scope to a specific tenant).
#Roles in practice
- First-time setup: you accept the Magic Link invite, land on the dashboard, mint an API key, and start integrating. The API keys page walks through this end to end.
- Day-to-day: you adjust prompts, policies, and engine settings through their respective sidebar entries. Every edit creates an immutable new version; the active version pointer flips on publish, so rollbacks are atomic.
- Observability: Prompt logs answer "what is the engine doing right now?". Security events answer "what is it rejecting?". Both are tenant-scoped and update in near-real-time.
#What's authoritative
The admin console is a thin shell over the engine API at
aiengine.velgent.com/api/admin/*. Anything you do in the UI you can
also do via that surface programmatically — useful for terraform-style
GitOps once your prompt + policy collection gets large. The admin API
isn't part of this public docs site today; reach out if you want access.
#Next steps
- Set up your first API key: Setting up API keys.
- Make your first request: Quickstart.
- Authentication header format: Authentication.