Skip to content
M↓ View as Markdown ↗

Getting started in Studio

Studio is where data teams build and configure AI Analysts. This guide takes you from sign-in to a deployed AI Analyst.

Availability

  • Who can do this: members with the Edit capability in the active Workspace (for example from the Editor or Administrator preset; see Members).
  • Deployment: Actian SaaS and self-managed deployments. Slack and Teams are not available on air-gapped self-managed deployments.

Setup flow

flowchart TD
    A[Sign in and pick a Workspace] --> B[Connect your data]
    B --> C[Build your semantic layer]
    C --> D[Create an AI Analyst]
    D --> E[Test your AI Analyst]
    E --> F[Deploy and invite users]

Before you start

  • Edit capability: building in Studio requires Edit in the Workspace you're working in. The Administrator and Editor presets grant it in the organization's default Workspace only. Inviting people to the organization requires Manage members and access.
  • A language model: AI Analysts need an LLM endpoint to answer questions. If none is connected for your organization, conversations (including tests in Studio) are blocked until someone with Manage inference adds one under Settings → Inference.

Step 1: Sign in and pick a Workspace

  1. Go to your AI Analyst address: the URL Actian gave you for Actian SaaS, or the one your administrator gives you for a self-managed deployment.
  2. Enter your Email, click Continue, and complete sign-in with the method offered (for example Continue with password, Continue with Google, Continue with Microsoft, or your company's single sign-on).

If you're new, enter your Email on the register page (/auth/register at your AI Analyst address), click Continue and choose a sign-in method. You can set up a new organization only where self-serve sign-up is enabled. Otherwise, ask your administrator for an invitation. On a self-managed deployment, the first organization and its administrator are set up during installation, not through sign-up. See Access Actian AI Analyst.

Models and AI Analysts belong to the active Workspace, and connections are attached to Workspaces. A new organization has one default Workspace. If you have access to more than one, open the organization switcher at the top of the sidebar, point to your organization, and choose the Workspace under Workspaces before you start building. See Workspaces.

Step 2: Connect your data

  1. In Studio, open Connections in the left sidebar.
  2. Click Add connection and choose what to connect, for example a warehouse such as PostgreSQL, Snowflake or BigQuery, or a CSV file.
  3. Enter your connection details and save.

Before you connect a warehouse:

  • Your database must accept connections from AI Analyst. On Actian SaaS, the setup help next to the form lists the AI Analyst IP addresses to allow through your firewall.
  • Use a read-only database user. AI Analyst only reads from your data source, and for some connectors a read-only user is the only safeguard against writes.

Note

If an administrator has turned on Restrict connection management (on the organization's Security page), members who have Edit in the Workspace can add, edit or delete connections only if they also hold Manage organization. Everyone else can view connections.

See Connect a data source for detailed instructions per database type.

Step 3: Build your semantic layer

The semantic layer maps your raw tables and columns to the business concepts your team actually uses, so your AI Analyst answers in those terms. In the Studio sidebar it is split into Models, Metrics and Glossary.

  1. Go to Models in the sidebar and click New model.

  2. In the Create Model dialog, choose how to start:

    • With Steward (recommended): Steward analyzes your schema, researches business context and drafts a complete model.
    • From Table: pick a Data connection, Schema and Table, then choose which columns to import.
    • From SQL: define a custom SQL query for complex transformations or multi-table joins.

    With Steward opens the Steward panel with a starter prompt. Replace the bracketed text with the table you want to model, send it, and approve the changes Steward proposes. 3. Refine the model's Dimensions (descriptive attributes like region or customer_tier, added with New dimension) and Measures (numeric calculations like total_revenue, added with New measure). 4. Optionally, go to Metrics and click New metric for your most important KPIs, and add business terms under Glossary.

See Creating models, Creating models with Steward, Metrics and Glossary to go deeper.

Step 4: Create an AI Analyst

  1. Go to AI Analysts in the sidebar and click Create Analyst.
  2. Enter a Name (for example "Sales Performance Analyst") and click Create.
  3. On the Overview tab, add a Description that explains what the analyst does and when to use it.
  4. On the Models tab, click Edit on the Access Mode card and choose Choose specific models (then tick the models the analyst may query), Models with any of these tags or All models except these tags (then pick Tags). Click Save changes. Metrics access follows from model access.
  5. On the Instructions tab, describe what the AI Analyst should focus on and how it should answer.

See Instructions for guidance on writing effective instructions, and Models for the access modes.

Step 5: Test your AI Analyst

Your AI Analyst opens with a Test your analyst panel on the right. If you closed it, click the Open test panel icon at the right edge of the page to bring it back.

  1. Ask the questions you expect real users to ask.
  2. Click New chat to start over with a clean conversation.
  3. Turn on Plan if you want the analyst to propose a research plan for you to review before it answers.

When you send a question, Studio saves any pending changes to the analyst first, so the test uses your latest configuration. If the save fails, nothing runs. If results are off, go back to your semantic layer or instructions, refine them, and test again.

Test conversations are labelled Studio test in the analyst's monitoring data. To review them, open Monitoring in the analyst header and choose Tasks, where you can hide them with the Hide test runs filter. They aren't counted in the analyst's Tasks run on the Overview tab.

Try it as a business user

On the AI Analysts list, open a row's actions menu and choose Open in Explorer. Explorer opens with that analyst already selected. Until the analyst is deployed, only members with Edit see it in Explorer, marked Not deployed.

Step 6: Deploy and invite users

New AI Analysts start as drafts. Business users can't see them in Explorer until you deploy them.

  1. In the analyst header, open the deployment dropdown (it reads Not deployed until you deploy) and choose Explorer.
  2. In the Deploy to Explorer dialog, choose who can use the analyst:
    • Studio users only: keep it as a draft, not visible to business users in Explorer.
    • Specific users: pick the members who can use it.
    • Everyone: every member of this Workspace.
  3. Click Save. Members with Edit in the Workspace always have access to every AI Analyst.
  4. Optionally deploy to Slack or Teams from the same dropdown. If the channel isn't installed yet, you're taken to Connections to add it first. See Slack and Microsoft Teams.

Once the analyst is deployed to Explorer, the same dropdown also offers Open in Explorer.

To bring new people into the organization, go to Members in the sidebar's Manage group. Inviting members requires the Manage members and access capability. See Members.

Next: explore Suggestions and Saved Prompts to help your users get started. For a short checklist of these steps, see Creating your first AI Analyst.