AACSearch / guide
Console overview
The organization dashboard — who sees what, how it reaches the API, and every screen in the sidebar.
The console is the web dashboard where an organization runs AACSearch:
create collections and index documents, tune relevance, configure AI
assistants, read analytics, publish a search page, and manage the team and
billing. It lives at the console origin (https://app.aacsearch.com)
and is a Next.js application over the same control plane and tenant API this
documentation describes.
Everything in the console is scoped to the active organization. A user may belong to several; the organization switcher in the sidebar changes which one every screen reads and writes.
How the console reaches the API
The dashboard uses two paths, and it is worth knowing which one a screen uses when you go looking for the equivalent API call:
| Path | Used for | Authentication |
|---|---|---|
Control plane (oRPC) | Organizations, members, API keys, billing, wallet, webhooks, audit log, analytics rollups, quality lab, alerts, hosting configuration. | The better-auth session cookie. |
Tenant API (/v1/*) | Everything the search engine owns: collections, documents, search rules, synonyms, curation, presets, connectors, assistants, models. | The same session cookie, resolved by the API to the active organization and the member's role — no API key is stored in the console. |
Owners and admins may write; members are read-only. A control-plane call made without an active organization fails, which is why screens prompt you to pick or create one.
Who sees what
- Customers — owner, admin or member of their own organization — see the product screens: data, search, AI, insights, publishing and organization.
- Organization admins additionally see the destructive Danger zone in Settings, and are the only members the API accepts writes from.
- Platform operators — accounts listed in the API's
PLATFORM_ADMIN_EMAILS— additionally get the Platform group in the sidebar (search cluster, scheduler, raw resources) and the engine-wide operations behind it. The API enforces this independently, so a customer who types a platform URL gets a polite refusal rather than data. - Plan gating — features your plan does not include are rendered with a lock badge in the sidebar and a FeatureGate explanation on the page, pointing at the upgrade screen. The free tier always includes search, synonyms and stopwords; see Plans, usage and billing.
Overview
/dashboard is the landing screen, and the console root / redirects to it.
Three status cards sit at the top, each linking to the screen that acts on it:
- Search service — healthy or down, linking to Alerts & SLOs.
- Requests — used against the month's limit, the plan name, and an Exhausted badge when the quota is spent, linking to Billing.
- Credits — the wallet balance and an Empty badge, linking to Billing.
Below them, product health reports four things: data (document and collection counts, with a hint when collections are empty), search (the 30-day zero-result rate or zero-hit count and CTR, linking to Search issues), recommendations (how many models exist on the primary collection) and AI (the assistant count, with a no-credits badge).
A Suggested actions row offers the next useful steps — ingest documents, connect a source, open the quality lab, look at search issues, start the quick start — and hides the quick-start nudge once it is dismissed (remembered in the browser). A Collections grid lists each collection with its document and field counts and creation date, or a "create your first collection" empty state. An Organizations card lists the organizations you belong to with an active marker and a Switch action, plus New organization; without an active organization it becomes the prompt to create one, and a suspended organization shows an "API access is paused (reason)" banner linking to Billing.
Every section loads and fails independently, so one bad request degrades its own card rather than the screen.
Navigation
The sidebar is ordered the way a customer meets the product: get data in, make search good, add AI, measure, publish — then the organization's own affairs. The command palette (⌘K / Ctrl-K) searches the same screens.
| Group | Screens |
|---|---|
| Start | Overview (/dashboard, and / redirects here), Quick start |
| Data | Collections, Documents, Sources, Integrations, Migrate from Algolia — see Data and indexing |
| Search | Search preview, Playground (Search, Chat), Relevance (Search rules, Synonyms, Synonym suggestions, Curation, Stopwords, Stemming, Presets, Aliases), Semantic & image, Query intelligence, Natural-language search — see Search and relevance |
| AI | Assistants, Conversations, Conversation models, Recommendations, Personalization, Knowledge graph — see AI features |
| Insights | Analytics, Search issues, Analytics rules, Experiments, Campaigns, Quality lab, Autopilot, Observability (Query log, Sessions, Latency), Alerts & SLOs — see Insights and quality |
| Publish | Storefront, Hosted page, Webhooks, Support inbox — see Publishing and support |
| Organization | Team, Usage, Billing, Audit log, Settings (General, API keys, Security, Danger zone) — see Organization and billing |
| Platform | Search cluster, Scheduler, Raw resources — platform operators only, see Platform operations |
Quick start
/dashboard/quick-start is a six-step onboarding checklist with a progress bar
and a Done / Up next / Pending badge on each step. Expanding a step
reveals its body and the links that complete it:
Create a collection
A schema for the documents you index.
Index documents
With a copyable curl snippet, the console's own import, or a connector.
Run your first search
The step completes once a search has run in the search preview or playground, or an assistant exists.
Create an API key
A full key for your server, or a search-scoped key for a browser.
Put it on your site (optional)
Widget, React or SDK snippet.
Publish a hosted page (optional)
A search-and-chat page the platform serves for you.
The checklist probes collections, connectors, assistants, keys and hosting independently each time the window regains focus, so a step ticks itself off when the work is done elsewhere; a probe that fails shows "Could not check every step" with Retry, and a top-level Refresh re-runs them all. A few steps cannot be detected server-side and are remembered locally in the browser. Hide reminder dismisses the Overview nudge.
Readiness matrix
/dashboard/capabilities is a readiness matrix for the whole account. Every
major console surface is a row, checked twice: the plan gate from the
quota's feature flags, and a live probe of the control plane (a collection
with documents, a connected source, a created assistant, a usable key, an
enabled hosted page, an answering wallet).
Four statuses are possible: Ready (plan allows it and the probe answered positively), Not configured (allowed but not set up), Plan gated (the plan does not include the feature — the row shows a "Plan feature" badge and its button becomes View plans) and Unknown (a probe failed; its own rows degrade rather than the page). A header line counts them — "plan plan · N ready · N not configured · N plan-gated · N unknown" — and Refresh re-runs every probe. Without an active organization the screen asks you to pick one on Overview.
The screen is not linked from the sidebar — open /dashboard/capabilities
directly. A second unlisted screen, /dashboard/ranking, is the
Ranking Studio.