Instant BI: AI Powered Analytics : Semantic Creation

AI response is as good as the semantic model and a good LLM.

Step 1: Open the Semantic Model Creation Screen

Click the ribbon icon to expand it, then select “Semantic Model” to begin creating a new one.

Step 2: Add Your Metadata

Before a Semantic Model can answer questions, it needs metadata added to it. Once inside the creation screen, click the highlighted icon to open the metadata panel and load the data you want the model to work with.

Step 3: Add Fields

With metadata loaded, fields can be added in two ways:

• Click the “+” icon at the top right (shown as “Add Metric” by default, which can be renamed). It is better to use “Add Metric” when we want to add and create a custom calculation.
• Drag the required field directly from the metadata pane — any field from any table can be dragged in.

Once added, a field can be renamed by clicking it’s pencil icon.

Note: keep only the fields business users will actually need when interacting with the data in plain English.

Step 6: Measure Settings — Format & Aggregation

When a field is set as a Measure, two more options appear:

Format — controls how numbers are displayed. More and more prebuilt formatting options will be added. The formatting :

• 0.00 — two decimal places (e.g. 1234.50)
• 00.00 — minimum two digits before the decimal and two after (e.g. 05.25)

Aggregation — controls how the Measure is summarized when used in reports or natural-language queries:

• Sum — total of all values
• Count — number of records
• Average (Avg) — the mean value
• Minimum (Min) / Maximum (Max) — smallest / largest value
• Distinct Count — number of unique values

Step 7: Formula (Custom Calculated Fields)

In many cases we often need a metric which is a complex caluclation. In those cases “Add Metric” can be useful. We also have an option to add a Formula which lets you build a custom metric within the Semantic Model. Click “+”, give it a name, and write the calculation as close to SQL syntax as possible to avoid ambiguity. Any field in the metadata can be referenced, even if it isn’t part of the model.

Example — Name: Unique Travels

Formula: distinct count(travel_details.travel_id)

Step 8: Description & Semantic Type

Description — a short explanation of what the field represents, helping both users and the model understand its purpose. Example: “Cancellation Rate represents the percentage of meetings that were cancelled out of the total scheduled meetings.”

Semantic Type — classifies the kind of information a field holds, so the model can interpret data and queries more accurately. Available types includes:

• Numeric, Identifier, Person, Geographic, Organization, Product & Business, Internet & Network, Documents & Content, Medias

Step 9: AI Context — Instructions, Synonyms & Examples

AI Context gives the model extra help in interpreting a field correctly:

• Instructions — business rules, calculation logic, or other context for how the field should be used.
• Synonyms — alternate names users might type, separated by commas (e.g. users, clients, customers).
• Examples — sample mappings that help translate natural language into data values, e.g. “cancelled meetings → meeting_cancellation_status = true”.

Step 10: Add a Business Domain

The “+” icon in the Business View lets you add a Business Domain — the functional area your data belongs to, giving the AI business context. Common domains include Travel, Sales, Operations, Retail, HR, Finance, and Marketing.

Once created, a domain appears in the Business View, where it can be renamed and managed. Right-click a domain to add a Description or Delete it.

Step 11: Organize with Topics

A Topic groups related fields within a Business Domain — for example, an “Orders” topic under the Sales domain. Drag and drop individual fields onto a topic, or drag an entire hierarchy to assign all its columns at once. A single field can belong to multiple topics if relevant to more than one business area.

Organizing the Semantic Model into clear Business Domains and Topics improves navigation, helps the AI understand the data better, and leads to more accurate answers. In many cases, the same KPI/metric can have different calculations when part of different topics.

• Domain = The business area
• Topic = A specific subject within that business area

Think of it like this:

• Hospital is a domain.
  • Patients, Doctors, Billing, Pharmacy are topics.

Or,

• School is a domain.
  • Students, Teachers, Attendance, Exams are topics.

Step 12: View and Reuse as JSON

The JSON tab shows the complete configuration of the Semantic Model. Every action performed in the designer — adding fields, creating hierarchies, defining domains, assigning topics, configuring properties — is automatically reflected here. Advanced users can review or edit it directly.

The JSON toolbar offers:

• View Raw JSON — opens the full configuration for inspection or manual editing.
• Copy JSON — copies the configuration for backup, sharing, or migrating to another instance.
• Paste JSON — imports a previously copied configuration to recreate or update the model.

Reusing a Semantic Model this way saves time by avoiding manual recreation — it can be shared with teammates, reused on another server, and kept consistent across environments.

Step 13: Save Your Work

Once everything looks right, save your work directly into the Helical Insight file browser — either as a standalone file or inside a folder — so it’s ready to use whenever needed.

The JSON toolbar offers:

• View Raw JSON — opens the full configuration for inspection or manual editing.
• Copy JSON — copies the configuration for backup, sharing, or migrating to another instance.
• Paste JSON — imports a previously copied configuration to recreate or update the model.

Reusing a Semantic Model this way saves time by avoiding manual recreation — it can be shared with teammates, reused on another server, and kept consistent across environments.