strategy-scaling-architecture

This skill provides a structured framework for evolving the spectre-build CLI tool into an industrial-grade product. It focuses on breaking the 'three walls' that hinder scaling: a monolithic god-module architecture, print-coupled output, and synchronous execution limitations. Targeted at developers and architects working on the spectre-labs ecosystem, this skill guides the transition from script-based execution to robust service-oriented architecture.

• Orchestrates the transition from CLI-bound execution to a modular PipelineOrchestrator, enabling support for web interfaces, API integration, and headless operations.

• Implements an asynchronous event bus system to replace stdout printing with structured event streams, facilitating real-time telemetry, live steering, and GUI integration.

• Provides strategies for multi-model abstraction, enabling seamless switching between LLM backends and custom completion strategies like FunctionCallCompletion or ExitCodeCompletion.

• Supports the development of advanced features including adversarial code review, automated validation cycles, and complex pipeline chaining.

• Facilitates industrial-grade scalability through multi-tenant support, server-layer implementations (REST/WebSocket), and persistent state stores (SQLite/Postgres).

• Enables node-based pipeline editing and complex scheduling via crons, webhooks, and asynchronous subprocess handling.

• Use this skill when planning product-wide architectural shifts, adding new model adapters, or optimizing the build loop for complex, long-running tasks.

• Inputs typically include scope documents, architecture diagrams, and system requirements, while outputs involve refined pipeline configuration (YAML), new service modules, and structural refactoring plans.

• Requires familiarity with Pydantic for data validation, asynchronous Python programming (asyncio), and existing spectre-build pipeline logic (executor.py, stage.py).

• Adhere to the established phased development path: Foundation Extraction, Model Abstraction, Server Layer development, Live Steering, Adversarial Reviews, and Telemetry integration.