Neural API — Adaptive Multi-Layer Orchestration

What It Is

The Neural API is 🌿🖥️ biomeOS’s orchestration layer — the central nervous system that integrates the coordination triad and enables products to consume primal capabilities through graph execution.

One-line thesis: 13 primals expose 427 methods. The Neural API collapses these into ~20 atomic signals organized by tier (Tower/Node/Nest/NUCLEUS), so complex systems like rootPulse and product compositions emerge rather than being engineered as monoliths.

The Combinatorial Problem

8 springs validate across 13 primals. Each spring needs different capabilities from different primals. Without composition collapse, every spring (and every product) must independently discover, connect to, and manage 427 methods.

The Neural API solves this with three layers:

Layer 1: Primals (427 methods, JSON-RPC)
    |
Layer 2: biomeOS Neural API (atomic signals, graph execution)
    |
Layer 3: Emergent systems (rootPulse, RPGPT, helixVision, etc.)

Products compose at Layer 3. The Neural API handles Layer 2. Primals provide Layer 1. No product needs to know about all 427 methods — only the atomic signals relevant to its domain.

Atomic Tiers

Five Coordination Patterns

Graphs are defined in TOML:

[graph]
name = "rootpulse-commit"
pattern = "sequential"

[[graph.node]]
name = "health-check"
signal = "tower.health"
order = 1

[[graph.node]]
name = "dehydrate"
signal = "nest.dehydrate"
order = 2
depends_on = ["health-check"]

[[graph.node]]
name = "sign"
signal = "tower.sign"
order = 3
depends_on = ["dehydrate"]

Bidirectional Learning

The Neural API has a learning loop:

1. Forward pass — execute the capability graph, collect metrics
2. Backward pass — feed metrics to the PathwayLearner
3. Optimization — discover patterns, collapse redundant paths, improve routing

Each execution produces data that makes the next execution more efficient. The system learns which primals respond fastest, which capability paths produce the best results, and which subgraphs can be parallelized.

Composition Collapse

The philosophical commitment: when you can express an operation as a composition of existing capabilities, you do not build a new service. You compose.

None of these systems required new primals. They required new compositions of existing primals — exactly what the Neural API enables.

Implementation Status

Emergent Systems

Systems that emerge from Neural API composition rather than being built:

The Neural API is the kernel substrate — domain-agnostic graph execution that makes 427 primal methods composable into any system. Intelligence emerges from simple components and feedback loops, not from complexity. The primals are simple. The connections are rich. The results are emergent.