Skip to main content

Setting up Axiome

You administer an Axiome workspace. Three steps take it from empty to productive:

1. Connect a database​

Open Settings → Admin → Connections, click Add Connection, and configure BigQuery or PostgreSQL. Always Test Connection before saving.

Credentials are stored in Google Cloud Secret Manager — never in Axiome's own database — and Axiome only ever reads from your warehouse. Use a read-only role or service account scoped to the data you want to analyze. Details in Managing connections.

2. Curate the semantic layer​

The moment a connection is saved, Axiome introspects its schema and generates semantic models — one per table, with dimensions and measures inferred from column types. This baseline works immediately, but curation is where Axiome becomes reliable:

  1. Describe each model — "What does one row represent?". Descriptions feed the catalog and ground Axi: they're the single highest-leverage input to AI answer quality.
  2. Declare keys so models join correctly — see Keys and joins.
  3. Define metrics — your organization's official numbers (revenue, conversion rate…). Metrics are never generated; a business definition is a decision, not a guess. See Metrics.
  4. Hide technical columns so field pickers stay clean.

Start with the two or three tables your team asks about most; expand from there. The full reference is in Semantic layer.

3. Invite your team​

Create accounts in Settings → Admin → Users — each new user gets a one-time password to pass on securely. Grant the Admin role only to people who should manage connections, models, and users. See Users and roles.

4. Connect Slack (for alerts and reports)​

Alerts and scheduled reports are delivered to Slack, and users can't create any until a workspace is connected. Go to Settings → Admin → Integrations and click Connect Slack — a one-time OAuth install. See Slack integration.

Once done, alerting is self-service: any member can set up their own alerts and reports without coming back to you.

5. Write the company context​

Settings → Account → Context → Company context is where you tell Axi what your data can't say — your fiscal calendar, what "active customer" means, the periods that look broken but aren't. A handful of lines here measurably improves every user's answers. See Company context and memory.

Then, on an ongoing basis​

  • Watch usage — AI credits, error rates, per-member activity. A rising AI error rate usually points at a curation gap.
  • Refresh the schema after database migrations, and fix orphaned dimensions.
  • Grow the semantic layer as the team's questions evolve — recurring ad-hoc calculations deserve promotion to metrics.
Send your team the user guide

The Using Axiome tab of this documentation is written for end users — no technical background needed.