Configuration

All tuning lives in config/system.yaml.

Sections

Section	Controls
pools	Target sizes (2-7), spawn cooldown, health check intervals
mesh	Gossip rate, Hebbian decay/reward rates, signal TTL
consensus	Min votes, approval threshold, trust priors (Beta distribution), decay rate
cognitive	LLM endpoint, model tiers (fast/standard/deep), token budget, concurrency limit, attention parameters, per-tier temperature/top-p
memory	Max episodes, relevance threshold
dreaming	Idle threshold, replay count, strengthening/weakening factors, prune threshold

LLM Tiers

ProbOS supports three LLM tiers for different cognitive functions:

Tier	Use Case
fast	Classification, gist extraction, quick routing decisions
standard	General-purpose decomposition, agent instructions
deep	Complex multi-step reasoning, code generation, architecture design

Switch tiers at runtime with /tier fast|standard|deep. Each tier supports independent temperature and top-p tuning.

Standing Orders

Agent behavior is governed by a 4-tier instruction hierarchy stored in config/standing_orders/:

Tier	File	Scope
Federation Constitution	federation.md	Universal, immutable
Ship Standing Orders	ship.md	Per-instance
Department Protocols	engineering.md, science.md, etc.	Per-department
Agent Standing Orders	builder.md, architect.md, etc.	Per-agent, evolvable

Instructions are composed at call time by compose_instructions() and injected into every LLM request.

Pool Sizing

Each agent pool has a target size. The adaptive scaler adjusts pool sizes based on demand — pools handling high load grow, idle pools shrink back to their minimum.

pools:
  filesystem:
    target_size: 3
    min_size: 1
    max_size: 7