Context Intelligence
Context is not just memory. It is operational state.
This page explains how Subbasis treats context as execution state, not chat history.
What this means
Context Intelligence in Subbasis is:
- not just RAG
- not just chat memory
- not only prompt stuffing
- context attached to execution path
Operational context is the state that lets a workflow continue coherently across agents, approvals, retries, and time.
Why it matters
Without operational context, multi-step workflows rupture after handoffs or interruptions. Teams lose intent, duplicate work, and make inconsistent decisions.
With context bound to execution, Subbasis can support:
- sophisticated multi-agent workflows
- long-running operations
- local LLM continuity
- human-in-the-loop control
- reviewable execution
- fewer workflow ruptures
How Subbasis handles it
Subbasis keeps context attached to workflow state transitions. Instead of treating context as a free-text cache, it links context to roles, policies, approvals, and evidence events.
Typical context fields include:
- original request
- user intent
- organizational rules
- role/agent responsibility
- decisions made
- approvals granted/refused
- failures
- retries
- resumed state
- system state
- evidence
- relevant company data
- local execution state where applicable
Example scenario
A prospect mentions expansion intent during an ongoing sales workflow.
- The signal is captured as workflow context.
- Research notes and qualification rationale are attached.
- Approval state for external outreach is retained.
- CRM update state and connector outcomes are stored.
- Message history remains linked to the same execution path.
When the workflow resumes later, agents and human reviewers continue with full operational continuity instead of rebuilding context from scratch.
What to configure
Practical context configuration areas:
- workflow-level context schema
- role-level context visibility
- retention and redaction rules
- evidence linkage requirements
- local vs hosted context handling policies
- handoff requirements between agent tasks
Limits and deployment notes
- Available context depth and storage characteristics can depend on plan and deployment profile.
- Local execution patterns depend on your runtime topology (including ARIN-like patterns where applicable).
- Context continuity improves reliability but does not replace human oversight in sensitive operations.