Skip to content

Commit ed6693a

Browse files
committed
feat(nightwatch): add dual-panel evaluation for drift detection
Add DualPanelResult struct and dual_panel_evaluate function to enable two independent quality assessments on agent output. Drift is detected when panel agreement falls below 0.5. Panel A: Scores based on ReasoningCertificate quality (premises, claims, edge cases, confidence) Panel B: Scores based on output structure (sections, evidence markers, conclusion markers, minimum length) Includes comprehensive unit tests covering: - Both panels agree (no drift) - Panels disagree (drift detected) - Missing certificate scenario Refs #91
1 parent ab46bbd commit ed6693a

2 files changed

Lines changed: 386 additions & 2 deletions

File tree

crates/terraphim_orchestrator/src/lib.rs

Lines changed: 3 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -55,8 +55,9 @@ pub use error::OrchestratorError;
5555
pub use handoff::{HandoffBuffer, HandoffContext, HandoffLedger};
5656
pub use mode::{IssueMode, TimeMode};
5757
pub use nightwatch::{
58-
validate_certificate, Claim, CorrectionAction, CorrectionLevel, DriftAlert, DriftMetrics,
59-
DriftScore, NightwatchMonitor, RateLimitTracker, RateLimitWindow, ReasoningCertificate,
58+
dual_panel_evaluate, validate_certificate, Claim, CorrectionAction, CorrectionLevel,
59+
DriftAlert, DriftMetrics, DriftScore, DualPanelResult, NightwatchMonitor, RateLimitTracker,
60+
RateLimitWindow, ReasoningCertificate,
6061
};
6162
pub use persona::{MetapromptRenderError, MetapromptRenderer, PersonaRegistry};
6263
pub use scheduler::{ScheduleEvent, TimeScheduler};

0 commit comments

Comments
 (0)