How to enable AI in Omniscope

Omniscope integrates AI capabilities in various places, but they are not enabled by default and need some configuration steps, as follows.

Enabling labs features

Some AI features require particular "labs" options to be enabled.  Go to "Admin > Labs" and tick the desired feature to enable. Save, then reload the page. 

Finding AI settings

Omniscope AI settings are configurable and customisable at the folder level, currently within the Create/Edit Permissions dialog, available from the corner in the folder list page:

For convenience, you can also access the root folder's AI settings (which will apply server-wide if no subfolders override them) from "Admin > AI Settings".

AI Settings

The AI settings dialog/section will appear somewhat like this:

AI settings comprise two main concepts:

• Providers are services that supply AI models for Omniscope to use, including generative models and decision models. Examples include OpenAI, Azure OpenAI and Jev. You can also configure an OpenAI-compatible service as a Custom provider, such as a local model running through llama.cpp. Commercial providers typically require an API key and may charge for usage.
• Integrations are aspects of Omniscope that take advantage of AI capabilities. Examples include Report Ninja, Workflow Ninja, Insight Explorer, AI Request Block, AI Insights Block, Custom SQL, Workflow AI Decisions and AI Decision Formulas.

The set of providers, and each integration, can be separately configured in a given folder, or allowed to inherit from the parent folder. You can also restrict subfolders from overriding the configuration defined at a given level, as a way of applying a policy to control AI use on a given server.

Providers

By default there won't be any providers, so you won't be able to use AI features. You will typically need to sign up for an account at https://platform.openai.com/, and go to "Settings > API keys", create an All-permissions key, and then come back to Omniscope. Add an OpenAI provider and enter the API key. 

You don't need to touch any other settings.

Alternatively, if you have a local / on-prem / other OpenAI-compatible provider, configure it by adding a "Custom" provider. This is beyond the scope of this article, and requires an Enterprise licence. 

See Using a local AI model or Setting up an Azure OpenAI provider.

Integrations: Report Ninja

Integrations are enabled by default, but will only come to life when you've configured a provider (above), and, currently, have also chosen a Default model. For example here for Report Ninja, we've chosen gpt-4o-mini, a very cost-effective option and surprisingly capable model.

You can pick other models. Note that the costs shown are a snapshot and may not be correct at a given time. Models that are known to be incompatible are excluded from the list. New models Omniscope does not recognise will show up marked as "unknown". If you choose an unknown model, it may not work correctly.

You will now see Instant Dashboard appear in the "Create project" section, and can use Instant Dashboard report preset in the Workflow "Add block" menu. And for any existing report, you can use the corner menu "Ask Report Ninja" to access it, and get an assistant that helps, suggests, answers questions and edits your dashboard:

Report Ninja works well with gpt-4o-mini, gpt-4o, and o3-mini from OpenAI, and also with some larger local models such as DeepSeek-R1-Distill-Qwen-32B-Q4_K_M.gguf via llama.cpp on a MacBook Pro M2 or better, although there are many possibilities in terms of providers and models. 

Other integrations

Insight Explorer lets editors and viewers ask natural language questions across multiple data sources. It uses Omniscope to query and analyse the data, returning answers supported by diagrams, interim results and explanations.

Report Ninja helps you design and build dashboards by answering questions about your data and suggesting layouts, visualisations and configurations.

Workflow Ninja helps you understand and explain complex workflows, showing how fields flow and how data is transformed. Answers are enhanced with diagrams and block links that highlight the relevant block.
From version 2026.2 it now can author workflows and reports and comes with a rich API to support editing via the MCP server too.

The AI Insights Block integration generates data-driven batches of AI-powered insights and answers by querying the connected datasets, returning results you can use downstream in your workflow.

The AI Python Script integration serves the AI Python script block, which is now a legacy block: it keeps working in workflows that already contain it, but it is no longer offered in the Add Block menu. For new work, use the Custom Python Script block (Workflow Ninja can write and edit its script for you), the AI Request block, or the AI Code Interpreter block.

The AI Request Block integration is the way to send data-driven prompts in batches to an LLM endpoint for chat completions with low-level, explicit control over how each prompt is constructed, while keeping simple use cases easy to configure.

The Custom SQL integration allows you to write a natural language query and automatically generate the SQL query. Find this after enabling this integration in the Database block Options tab upon choosing Custom SQL.

The Workflow AI Decisions integration supplies the AI Decision block and AI decision rules in Record filter. These assess records using questions you define, adding answers as fields or selecting records according to a yes/no probability threshold.

The AI Decision Formulas integration supplies AI_DECISION_YESNO, AI_DECISION_CHOICE and AI_DECISION_SCORE for use in workflow expressions, including Field organiser and Control flow blocks.

Generative AI integration Feature / Provider / Plan matrix

✅ Fully Supported on all plans

☑️ It may require extra support to configure it.
ENT Requires Enterprise plan

Jev supports the Workflow AI Decisions and AI Decision Formulas integrations. Both integrations require the AI Features Pack. When using Jev, configure the provider with your Jev API key.

Settings wrap-up

That's it. To summarise:

1. Enable the feature flag Report Ninja (if required).
2. Add a provider, typically OpenAI, and an API key.
3. In (e.g.) Report Ninja integration settings, choose a default model such as gpt-4o.

Last updated on 4 November 2026