Existence_and_Lifecycle

Existence and Lifecycle

Orrin's runtime is finite on purpose: it has a persistent lifetime budget, phases of horizon-awareness, a terminal cycle, sealed run evidence, and a recoverable reset. This page maps that machinery.

The runtime-lifetime clock

brain/cognition/runtime_lifetime.py is the single source of truth for the lifetime budget:

• Rolled once at first run — a random span inside a band of roughly 365–730 days (bounded by ORRIN_LIFESPAN_MIN_DAYS / ORRIN_LIFESPAN_MAX_DAYS), persisted in data/lifespan.json, and counted in wall-clock days across restarts.
• The runtime's internal estimate of remaining lifetime is deliberately approximate: a small noise offset biases it, so the figure it acts on is not the true one.
• Horizon-awareness grows through four phases — early → middle → late → terminal — which progressively colour long-term prioritization.
• When the deadline arrives, the runtime runs its final cycle and the loop exits; the desktop app shows an end-of-life screen.

Do not confuse this with the supervisor's LifespanByCycles (supervisor/lifespan.py) — that is a per-process uptime cutoff that resets on every restart.

Restoration

Idle cycles are not dead time: the idle/consolidation phase runs memory consolidation, replay, and closed-time accounting at a low-power cadence (see The Cognitive Loop).

Life capsules (the run record)

brain/evidence/life_capsule.py seals one run into a single self-describing .orrinlife.zip (under exports/life_capsules/): raw streams preserved, plus cleaned tables, a queryable SQLite DB, computed metrics, a claims ledger, and a token-budgeted LLM bundle. Hand someone the file and they can understand the run without running Orrin.

The organizing rule is raw → cleaned → derived → interpreted, each layer downstream-only; the builder is a pure function of the raw layer, so anyone can deterministically rebuild the derived artifacts and check the claims. life_capsule_ingest.py and life_capsule_metrics.py handle ingestion and metrics.

Reset (a new existence)

reset_orrin.py starts from fresh state — it is not routine maintenance, because state is self-bounding (capped logs, windowed history, decaying memory; see docs/CONFIGURATION.md).

python reset_orrin.py                # snapshot + reset (recoverable)
python reset_orrin.py --dry-run      # show what would change
python reset_orrin.py --hard         # also clear bandit / decision learning
python reset_orrin.py --no-snapshot  # skip the snapshot (irrecoverable)

A reset rolls a new lifetime budget; identity, memory, and learning start over.

Code pointers

• brain/cognition/runtime_lifetime.py — the clock
• brain/evidence/life_capsule.py — the capsule builder
• reset_orrin.py — reset with snapshot
• frontend/src/pages/Life.tsx — the Life room in the UI