A local-first CLI that converts English-language requirements into Attractor pipelines and executes them step-by-step using tool-equipped coding agents in isolated git worktrees.
A comprehensive Python implementation claiming 100% spec coverage — features a custom recursive-descent DOT parser, 9 node types, middleware-chain LLM client, HTTP server with SSE streaming, cooperative cancellation, and a goal-gate circuit breaker for safe autonomous iteration.
A Kotlin pipeline runner that orchestrates multi-stage AI workflows using DOT-format directed graphs — integrates with Claude, GPT, and Gemini, with conditional branching, parallel execution, human-in-the-loop gates, SQLite checkpoint/resume, and a real-time SSE-powered dashboard.
A spec-first Rust library implementing the Attractor specification as layered crates — a unified multi-provider LLM client and a coding-agent loop — with deterministic conformance testing across providers.
A TypeScript pipeline orchestration engine that executes multi-stage AI workflows defined as DOT-syntax directed graphs — with checkpoints, parallel branches, human-in-the-loop gates, and an HTTP API for remote submission.
A Python pipeline execution framework that interprets DOT-syntax directed graphs with a handler-based architecture — supporting human-in-the-loop gates, parallel branches, checkpoint recovery, and an HTTP server with Server-Sent Events streaming.
A Ruby gem implementing all three Attractor spec layers — a provider-agnostic LLM client, an autonomous agent loop with dev tools, and a DOT-based pipeline orchestrator with conditional branching, parallel execution, and a five-step edge-selection algorithm.
A TypeScript monorepo structured as three workspace packages — core engine, pi.dev LLM backend, and CLI — with a backend-agnostic CodergenBackend interface that lets providers be swapped without touching the execution engine.
A Scala 3 pipeline orchestration engine built on cats-effect and fs2 — type-safe concurrency via IO monads, stream processing with automatic backpressure, sealed-trait pattern matching, and an immutable case-class architecture implementing the full Attractor spec with a provider-agnostic LLM client.
A multi-agent Software Factory that replaces pass/fail testing with probabilistic satisfaction scoring — orchestrates specialized agents (Coding, Validator, Debugger, Planner) against holdout scenario sets and Digital Twin API replicas to iterate autonomously toward a confidence threshold.
An F# pipeline execution engine targeting .NET 10 — three libraries (Attractor engine, unified multi-provider LLM client, coding agent) with CSS-like stylesheets for LLM routing, automatic retry loops with goal gates, full checkpoint/resume, and 317 tests across three test projects.
A pure C11 implementation compiling to a static library and CLI binary — includes a hand-written DOT parser, multi-provider LLM client over libcurl, and self-referential validation pipelines that use Attractor to audit its own spec compliance.
A clone of strongdm/attractor with extensive documentation and 5 new DOT pipeline blueprints — demonstrates the software factory concept with ready-to-run blueprints for login pages, REST APIs, CLI tools, landing pages, data pipelines, and static site generators.
A Ruby DOT-based workflow orchestration engine for AI-driven development — parses Graphviz DOT files into directed graphs with LLM execution, human approval gates, conditional branching, parallel fan-out/fan-in, automatic retries, checkpointing, and 13 built-in linting rules for pipeline validation.
PHP and Tcl implementations of the Attractor spec layers — including unified multi-provider LLM clients, autonomous coding-agent loops, and DOT-based pipeline runners for software-factory workflows.
A spec-driven DOT pipeline engine that coordinates AI coding agents — uses convergence loops where specifications drive verification, with fresh context windows per attempt, persistent learnings from failures, holdout test scenarios, Effect.ts error handling, and a web dashboard for monitoring.
A Kubernetes-native Dark Factory pipeline for the Bravo Zero platform — implements Judge-01 Scenario Eval with trained D3N models, plus phased Spec Engine, Attractor, Scenario Executor, and DTU Controller components per ADR-151.
A C# implementation of the Attractor spec with a DOT-based pipeline engine, an agentic coding loop, and a unified multi-provider LLM client supporting Anthropic, OpenAI, and Gemini.