Module temporal 

Temporal Constraint Agent — Agentic Controls for Temporal Intelligence

A constraint system that develops intelligence over time by:

1. Reading dodecet constraint states from sensors
2. Maintaining a temporal model (deadband funnel state)
3. Predicting future constraint states
4. Adjusting funnel shape based on prediction error
5. Detecting anomalies when predictions fail
6. Learning optimal control parameters

The Key Insight

The deadband funnel IS the agent’s temporal model. It encodes:

• Past: integral of precision energy (how much work done)
• Present: current dodecet state (where we are now)
• Future: predicted convergence time (when we’ll be done)

The finesse is the set of control parameters that tune this temporal model. The agentic controls are the API for higher-level agents.

Temporal Intelligence Stack

Layer 4: PLANNING          — Predict future constraint states, plan paths
Layer 3: LEARNING          — Adjust funnel shape from history
Layer 2: PREDICTION        — Kalman-like filter on dodecet stream
Layer 1: CONTROL           — PID on constraint error (P=error, I=energy, D=rate)
Layer 0: PERCEPTION        — Snap to lattice, encode dodecet

Structs

AgentSummary

Summary of agent state for fleet reporting.

TemporalAgent

A temporal constraint agent that reads dodecets and develops temporal intelligence.

TemporalUpdate

Output of a temporal update

Enums

AgentAction

Actions the agent can recommend

ChiralityState

Chirality state — has the agent locked into a Weyl chamber?

FunnelPhase

Which phase of the deadband funnel are we in?