Forge Pool

Appearance

Adapters ​

Adapters are the boundary layer of Forge.

They connect external systems to the Forge core — without redefining how computation works.

Core Principle ​

Forge is strict at the core and open at the edge.

  • The core defines execution truth (primitives, profiles, distributed compute)
  • The adapter layer defines how that truth is accessed and used

Adapters exist at that edge.

What Adapters Do ​

Adapters translate intent into execution.

They:

  • accept input from users or systems
  • validate and normalize data
  • map intent to primitives and profiles
  • trigger execution through the core
  • shape results for external use

What Adapters Do Not Do ​

Adapters do not define compute.

They do not:

  • implement primitive logic
  • redefine execution semantics
  • alter meaning of results
  • act as execution engines

All computation lives inside Forge core.

System Position ​

text
External Systems / Users
        ↓
      Adapter
        ↓
Forge Core (Hub → Agents → Primitives)
        ↓
Distributed Execution
        ↓
      Adapter
        ↓
External Systems / Users

Adapters are entry and exit points — not the system itself.

Adapter vs Core ​

LayerResponsibility
AdapterInput, translation, orchestration, output
CoreExecution, computation, primitives

Why Adapters Exist ​

Forge is built around canonical primitives.

Adapters exist to:

  • make the system usable
  • support domain-specific workflows
  • enable integration without modifying the core
  • allow the system to expand without breaking consistency

Flexibility Model ​

Adapters are intentionally flexible.

They can:

  • be simple or complex
  • call zero or multiple primitives
  • operate as single-step or multi-stage workflows

There is no required structure.

But there is one rule:

Adapters must never redefine compute truth.

Types of Adapters ​

Adapters are categorized by responsibility:

  • Ingest — normalize external input
  • Execution — trigger and coordinate primitives
  • Aggregation — combine results
  • Output — shape results for delivery
  • Bridge — connect Forge to external systems

These are patterns, not strict classes.

Invariants ​

All adapters must respect system invariants:

  • compute truth is external
  • execution boundaries are explicit
  • traceability is preserved
  • results are not distorted
  • no hidden state affects outcomes

See Invariants for full rules.

Mental Model ​

Adapters are:

  • translators between systems
  • coordinators of execution
  • boundaries of responsibility

They are not the engine.

Navigating This Section ​

Final Note ​

Adapters make Forge usable.

The core makes Forge correct.