<2-3 sentences summarising what the project addresses and solves, the target audience, and the benefit.> This AI novel writing system addresses the coherence and memory limitations of standard models for long-form content. It enables consistent writing over thousands of chapters by managing story state explicitly, benefiting authors seeking deep, continuous narrative generation.

How It Works

Tianming uses a "closed-loop writing process" where each chapter is validated through a "generation gate" with six checks before its state updates a "15-dimensional fact snapshot." This state-based approach, not context windows, ensures continuity by operating on explicit, verifiable data. Long-distance recall retrieves relevant historical "slices" and "milestones" based on current chapter needs, ensuring plot threads remain accessible.

Quick Start & Requirements

• Primary install/run command: Clone the repository and run dotnet run --project Core/App/天命.csproj.
• Non-default prerequisites: Windows 10 (19041) or later, .NET 8.0 SDK. Users must provide AI API Keys for supported services (e.g., OpenAI, Anthropic).
• Built-in Models: Includes local ONNX models for semantic search (bge-small-zh-v1.5) and text refinement (chinese-roberta-tiny), which auto-load and release from memory.
• Links: QQ Group: 414086347.

Highlighted Details

• State-Based Consistency: Achieves narrative coherence over thousands of chapters via a "15-dimensional fact snapshot" and "12 types of change statements," bypassing model context limits.
• Six-Gate Generation Validation: Employs rigorous checks (protocol, citation, consistency, unknown entity, description, blueprint) to ensure AI output adheres to established rules and facts before commitment.
• Long-Distance Recall: Systematically retrieves historical context and plot points (milestones, slices) for chapter generation, overcoming model memory limitations.
• Unified Validation: Performs system-wide checks on world-building, plot, and character rules, precisely identifying affected chapters for revisions.
• Multi-Model & Local AI: Supports multiple AI providers via Microsoft Semantic Kernel and includes local ONNX models for semantic search and text refinement.

Maintenance & Community

The project is actively maintained, with a QQ group (414086347) available for bug collection and community discussion. The author, "子夜," can be contacted via QQ (229164036).

Licensing & Compatibility

• License Type: MIT License.
• Commercial Use: Commercial applications, including resale, SaaS deployment, or integration into paid products, require explicit authorization from the original author.

Limitations & Caveats

The system is exclusively designed for Windows operating systems (Windows 10 19041+). Users are responsible for providing their own API keys for external AI models, as none are included. Commercial use, despite the MIT license, is restricted and necessitates separate author permission.