<2-3 sentences summarising what the project addresses and solves, the target audience, and the benefit.> InkOS is a multi-agent system designed for autonomous novel production, addressing common AI writing limitations like memory loss and plot incoherence. It enables AI agents to write, audit, and revise novels, incorporating human review gates, targeting engineers and power users seeking robust AI-driven narrative generation.

How It Works

The system employs a five-agent pipeline (Radar, Architect, Writer, Auditor, Reviser) that iteratively generates and refines chapters. Core to its approach are seven "truth files" (e.g., current_state.md, particle_ledger.md) that serve as a persistent, verifiable source of truth for world state, character knowledge, and plot progression, mitigating LLM hallucinations and ensuring long-term narrative consistency.

Quick Start & Requirements

• Installation: npm i -g @actalk/inkos
• Prerequisites: Node.js, npm, and access to an LLM provider (OpenAI, Anthropic, or compatible) with API key and base URL configured via global or project .env files.
• Key Commands: inkos book create, inkos write next, inkos agent, inkos config set-global.

Highlighted Details

• Persistent State Management: Utilizes seven "truth files" to meticulously track world state, resources, plot hooks, and character relationships, ensuring continuity across extensive narratives.
• Comprehensive Auditing & AI Mitigation: Features a 26-dimension audit system, including rule-based AI artifact detection and anti-detection rewriting pipelines, alongside style fingerprinting for de-AIing text.
• Deep Customization: Allows genre-specific creation rules (themes, pacing, taboos) and per-book overrides for protagonist traits, numerical systems, and narrative prohibitions.
• State Snapshots & Rewrites: Generates state snapshots for each chapter, enabling rollbacks and complete rewrites of any chapter with full state restoration.
• External Agent Integration: Atomic CLI commands with --json output facilitate integration with external AI agents (e.g., OpenClaw) for complex workflow orchestration.

Maintenance & Community

The project exhibits active development with a detailed roadmap including planned features like a Web UI (studio/) and custom agent plugins. Contributions are welcomed via issues and PRs, with development instructions provided. No specific community links (Discord, Slack) are listed.

Licensing & Compatibility

• License: MIT License.
• Compatibility: The MIT license is permissive, generally allowing for commercial use and integration into closed-source projects.

Limitations & Caveats

The planned Web UI (studio/) and advanced features like multi-model routing for agents and a custom agent plugin system are still in development, indicating current limitations in user interface and agent extensibility.