@substrate-ai/factory

Software factory capabilities for the Substrate platform — the graph execution engine, scenario-based validation, convergence loops, digital twin orchestration, LLM client with middleware, and pipeline template library.

Depends on @substrate-ai/core for transport-agnostic infrastructure.

Install

npm install @substrate-ai/factory

Node 22 or later is required.

Modules

| Namespace | Representative exports | |---|---| | graph | parseGraph, createValidator, createGraphExecutor, edge selector and stylesheet helpers | | handlers | HandlerRegistry, createDefaultRegistry, manager-loop handler | | scenarios | ScenarioStore, ScenarioRunner, computeSatisfactionScore, registerScenariosCommand | | convergence | Goal gates, four-level retry chain, budget controls, plateau detection, remediation context injection | | backend | Direct API backend — unified LLM client, codergen loop, provider-aligned tool sets | | llm | LLMClient with middleware chain (logging, retry, cost-tracking) and provider adapters (Anthropic, OpenAI, Gemini) | | twins | Digital twin registry, Docker Compose orchestration, health monitoring, persistence, pre-built templates | | config | FactoryConfigSchema, loadFactoryConfig, resolveConfigPath |

Usage

import { parseGraph, createGraphExecutor } from '@substrate-ai/factory'

const graph = parseGraph(dotSource)
const executor = createGraphExecutor({ /* deps */ })

Versioning

Releases in lockstep with substrate-ai. See the main repository for full documentation.