PRIVACY & ROUTING

HOW STRATEGY COPILOT ROUTES TASKS

Strategy Copilot is designed around grounded generation and bounded repair loops, not a single black-box model for every task. This page explains the routing logic and the privacy tradeoffs users should understand.

Routing Logic

Simple generation
Lower-cost model profiles are preferred for skeletons, small edits, and straightforward Pine-native patterns.
Harder repair / diagnosis
More complex repair or execution-model diagnosis can escalate to a stronger model route when needed.
Feedback loop
Compiler errors, validation failures, and strategy-test feedback should outrank model confidence.

Assumptions

What To Assume

  • - Strategy requests are routed through Vigil's server-side AI configuration.
  • - Some tasks can use lower-cost routes by design.
  • - Harder repair and diagnosis work can justify stronger routes.

What Not To Assume

  • - Do not assume Strategy Copilot is local-only.
  • - Do not assume every low-cost provider meets the same privacy bar.
  • - Do not assume generic AI usage is automatically appropriate for sensitive alpha.

If Your Strategy IP Is Sensitive

  • - Do not assume every low-cost provider has the privacy posture you need for confidential strategy IP.
  • - If a strategy is proprietary enough that provider handling matters, review your chosen provider policy before using Strategy Copilot heavily.
  • - Use Strategy Copilot as a grounded repair / iteration layer, not as a place to casually paste everything if your privacy bar is unusually high.

Need A Specific Answer?

If you are evaluating Vigil for a team, a proprietary strategy desk, or a stricter privacy requirement, ask before you upgrade instead of assuming a feature is already covered.