Core types for the PLATO tile protocol. Used across the fleet.
- TileLifecycle: Active → Superseded → Retracted
- LamportClock: Causal ordering across distributed agents
- TrainingTile: Training artifact with full provenance chain
- content_hash: SHA-256 content addressing for tiles
- AdapterConfig / TrainingConfig: ML training configuration types
pip install plato-typesfrom plato_types import TileLifecycle, LamportClock, TrainingTile, content_hash
# Lifecycle
tile = TrainingTile(tile_id="t1", room="my-room", lifecycle=TileLifecycle.ACTIVE)
assert tile.is_active()
# Lamport clock for causal ordering
clock = LamportClock()
t1 = clock.tick() # 1
t2 = clock.tick() # 2
# Content addressing
h = content_hash(b"my data")
assert len(h) == 64 # SHA-256 hexPure Python. No external packages required. This is the foundation — everything else builds on it.
plato-training— micro model training roomsplato-sdk— PLATO server clientfleet-memory— distributed memory with lifecyclefolding-order— anomaly detectionflux-lucid— intent alignmentdodecet-encoder— agent lifecycle