omni-model-explorer

Omni Model Explorer

Explore and understand an Omni semantic model through the REST API. This is the starting point — understand what exists before building, querying, or modifying anything.

Tip: Start with the Shared model — it contains the curated analytics layer.

Prerequisites

Set environment variables:

export OMNI_BASE_URL="https://yourorg.omniapp.co"
export OMNI_API_KEY="your-api-key"

API keys: Settings > API Keys (Organization Admin) or User Profile > Manage Account > Generate Token (Personal Access Token).

API Discovery

When unsure whether an endpoint or parameter exists, fetch the OpenAPI spec:

curl -L "$OMNI_BASE_URL/openapi.json"

Use this to verify endpoints, available parameters, and request/response schemas before making calls.

Core Workflow

Explore top-down: List models → Pick a model → List topics → Inspect a topic → Explore views and fields.

Step 1: List Available Models

Returns models with id, name, connectionId, and modelKind (SCHEMA or SHARED). Use the SHARED model — it contains the curated semantic layer.

To also see active branches on each model:

Each model in the response will include a branches array. Each branch has an id (UUID) and name — use the id as the branchId parameter in other API calls.

Step 2: List Topics in a Model

Topics are entry points for querying. Each topic defines a base view and the set of joined views available.

Returns topic names, base views, labels, and descriptions.

Step 3: Inspect a Topic

Get full detail including all views, dimensions, measures, relationships, and AI context:

The response includes:

base_view_name — the primary table
views[] — all accessible views, each with dimensions[] and measures[]
relationships[] — how views join together
default_filters — filters applied by default
ai_context — instructions for Blobby (Omni's AI)

Step 4: Read the Model YAML

For the full semantic model definition:

The mode parameter: combined (default) merges schema + shared model; extension shows only shared model customizations.

Model Architecture

Omni has three layers:

Schema Model — auto-generated from your database (read-only)
Shared Model — analytics engineer customizations (dimensions, measures, labels, topics, AI context)
Workbook Model — per-dashboard customizations (ad-hoc, not shared)

When exploring, use the combined view to see everything available.

Key Concepts

Views correspond to database tables. Each has dimensions (groupable fields) and measures (aggregations).

Topics join views together into queryable units — curated starting points for analysis. A topic has a base view, joined views, default filters, and AI context.

Relationships define joins: join_from_view, join_to_view, on_sql, relationship_type (one_to_one, many_to_one, one_to_many, many_to_many), and join_type (always_left, inner, full_outer).

Field naming: view_name.field_name with bracket notation for date granularity: orders.created_at[week].

Exploration Patterns

"What data do we have about X?" — List topics → inspect the most relevant one → review views and fields.

"How do these tables relate?" — Inspect the topic's relationships[] — check join_from_view, join_to_view, on_sql, and relationship_type.

"What measures are available for Y?" — Inspect the topic containing view Y → review the measures[] array with aggregate_type and sql definitions.

Calculation Fields

Calculation fields in the model use a different format than regular dimensions/measures. The field key is calc_name and the expression property is sql_expression — not name/sql.

Field Impact Analysis

Assess the blast radius of a field migration or removal before pushing changes to dbt:

Create a model branch with omni-model-builder where the field is removed or renamed
Run the content validator against that branch:

This returns all dashboards and tiles with broken references to the removed field.

Search model YAML for additional references (run in parallel with step 2):

Search the response for the field name to find references in other views, topics, and calculated fields.

Report: Combine content-validator results (broken dashboards/tiles) with YAML search results (model references) into a structured blast-radius report.

Do NOT paginate documents and check queries individually — the content validator does this for you in one call.

Docs Reference

Models API · Topics API · Modeling Overview · Views · Topics · Dimensions · Measures

omni-model-builder — create or modify views, topics, and fields
omni-query — run queries against discovered fields
omni-ai-optimizer — inspect and improve AI context on topics