Context Engine
Curate the AI context, playbooks, semantic definitions, and warehouse metadata Sundial's agents use.
The Context Engine holds everything Sundial's agents need to answer questions the way your team expects: business knowledge, playbooks, semantic definitions, and warehouse metadata with attached descriptions.
When observability shows weak answers (missing metrics, wrong methodology, or low trust), the Context Engine is where you fix the underlying context so the next answer improves.
Where to find it
Open the Context Engine from the product. From there you can work in AI context, playbooks, and semantic / warehouse context. Changes can be made in the UI or managed through Git for version control and collaboration; see the GitHub integration.
AI context
Use AI context for arbitrary knowledge the agent should take into account: how the business works, how the product is structured, and important changes over time. This is free-form context that sits alongside formal semantic definitions.
Playbooks
Playbooks live in the Context Engine so the data team can create and curate them. Many playbooks are built by Sundial and available to every customer; your team can also author its own for company-specific patterns. See Playbooks for what they encode and which analysis types they cover.
Semantic and warehouse context
The Context Engine also surfaces:
- Semantic measures and models: governed definitions the agent prefers for lookup and analysis. For how to model them, see the Semantic Reference.
- Warehouse tables and columns: organized metadata with descriptions the agent can consume. You can edit how tables and columns are described so the agent understands what each field means.
Context can be curated manually and, where helpful, initially drafted with AI assistance, then reviewed and refined by your team.