What a forest is

An organization has lands. A land is a server that runs containers and services. Each land grows a forest: an embedded NATS server with JetStream. All forests in the organization connect through NATS clustering. Together they form one system.

The forest is not a central server. It is the living system that appears when lands connect.

The parts you will meet

Leaf. An immutable event record. Every Leaf has a subject, JSON data, a source, and a timestamp. Leaves are the only way events move.

Wind. The pub/sub layer over NATS Core. Leaves travel on the Wind. A component drops a Leaf to publish it and catches a subject pattern to receive.

River. A durable stream for raw inbound data. External systems push into the River. Trees parse River data into typed Leaves.

Soil. The shared key-value store that holds current state. One bucket, one truth, shared across all lands. Any component can read Soil. Only the Decomposer and Fossil write it.

Humus. The durable stream of state changes. A component that wants to change state emits compost into Humus. The Decomposer applies the compost to Soil with optimistic locking. This single writer prevents concurrent overwrites.

Taproot. The durable outbound dispatcher. Writes that must reach an external system leave the forest as tap Leaves. Rootlets deliver them with retries and record the outcome.

Bedrock. Persistent file storage: git repositories, filesystems, Google Drive. Soil holds metadata; Bedrock holds the files. Never store secrets in Bedrock.

Why this matters to you

Every role in the forest works through these parts. You do not call a service directly. You drop a Leaf, read Soil, or emit compost. The data flow rules protect the whole organization: when you follow them, your work is observable, replayable, and safe.