Context Engine
Assemble the right context for each persona, domain and use case.
Lesson 12 of 17 · Find it on the platform map
01 What it is
What this layer does
The context engine assembles the specific data, metrics, rules, policies and permissions that a given person, team or agent needs for a given task. Rather than handing an analyst or an AI model everything, SCIKIQ builds context packs per persona, combining domain, process, entity and metric context with history, policy, live KPIs and the actions available. Packs are served by API and versioned per run, so it is always possible to see what context a decision was based on.
02 Concepts
Four ideas to hold on to
Context assembly
Selecting and combining the relevant information for a task rather than sending everything available. In SCIKIQ this draws on persona, domain, process, entity and metric context, plus current state such as live KPIs.
Context pack
A bounded, packaged bundle of the relevant data, rules, metrics, policies and permissions for a persona and use case. SCIKIQ generates context packs per persona, serves them by API and versions them per run.
Memory
Information carried forward from earlier interactions and past events so each request does not start from nothing. SCIKIQ includes historical context and memory as part of the context it assembles.
Subject 360
A consolidated view of one subject, such as a customer, supplier or asset, drawn from across systems. SCIKIQ places subject 360 views and predictions in context, so a pack about a customer includes that customer’s full picture and relevant forecasts.
03 In SCIKIQ
What the layer contains
04 What good looks like
Signs it is working
- Two people with different roles asking about the same customer receive context shaped to their role and their permissions.
- For any agent run, the team can retrieve the exact versioned context pack it used.
- Context packs include the policies and available actions that apply, not just data.
- Live KPIs in a context pack match the figures shown on the governed dashboard at the same time.
05 Diagnostics
Questions to ask your team
- 1
When an AI assistant gives a recommendation, can we see exactly what information it was given?
- 2
How do we stop context for one persona leaking data that another persona is not permitted to see?
- 3
What does a finance user need in context that a sales user does not, and is that written down?
- 4
How do we keep historical context and memory accurate as underlying data changes?
06 Keep going
Related reading
— Questions
Frequently asked
Is a context engine the same as retrieval for a chatbot?
Retrieval is one part of it. A context engine also adds policies, permissions, live state, history and available actions, and shapes all of this to the persona and use case.
Why version context packs per run?
So a decision or recommendation can be explained and reproduced later. If an outcome is questioned, the team can see exactly what the agent or person was shown at the time.
Who uses context packs — people or agents?
Both. The same pack generator serves context by API, so dashboards, assistants and domain agents can all draw on consistent, permission-aware context.