INTERACTION FREQUENCY
PhysK Suki
VC-flagged
DISCORD DM EXPLORER
Observe the conversation
in context.
Durable evidence, inspectable language signals, and a clear separation between text activity and likely voice-chat periods.
READING NOTE
Message counts are not reciprocity. Use the VC lens and evidence inspector before drawing conclusions.
PERIOD COMPARISON
Select two periods above to compare observable activity without turning it into a relationship score.
MESSAGES—within current filter
SESSIONS—gap-segmented conversations
CLASSIFIED—cached structured results
VC WINDOWS—probabilistic context flags
SESSION INITIATION
Who starts sessions?
SEMANTIC EVENT STORE
Observed labels
RESPONSE LATENCY
Time to next message
VC CONFIDENCE LENS
Evidence bands
CONTACT ERAS
Active, quiet, reconnection
WORKBENCH STATUS
The dashboard is querying SQLite. Semantic classification is never started by opening this page.
EVIDENCE INSPECTOR
Messages in their surroundings
Every row is a lead into nearby context, labels, relation links, and VC confidence.
CLASSIFICATION REVIEW
Spot-check the semantic layer
Random samples make model behavior inspectable. Preview rows are clearly marked until Luna is configured.
DIRECT RELATIONSHIP / CHARACTER STATEMENTS
What We Say About Each Other
Explicit descriptions of character, intentions, trust, appreciation, frustration, closeness, and the relationship itself—preserving mixed evaluations as mixed.
These are semantic candidates, not diagnoses. A later
characterization is only surfaced as a possible change when
persisted statements genuinely differ in stance; mixed or
context-dependent views remain visible as such.
POSSIBLE STANCE CHANGES
Review, don’t overread
QUOTE-WORTHY SEMANTIC EVENTS
Notable statements
Explicit statements that may carry stronger evidence than inferred behavioral metrics. Every item expands into surrounding context.