A multi-agent system designed for fully automated, long-form AI novel generation. It addresses the challenge of creating complete novels from a single prompt, requiring minimal human intervention. The target audience includes engineers, researchers, and power users interested in AI-driven creative writing automation, offering a robust framework for generating extensive narratives autonomously.

How It Works

The core architecture features an LLM-driven "Coordinator" that orchestrates "Architect," "Writer," and "Editor" sub-agents in a long-loop process. This design prioritizes simplicity and stability, avoiding complex orchestration. A novel "rolling planning" strategy generates future plot arcs and chapters on demand, referencing past context and character states, crucial for maintaining coherence in novels exceeding 500 chapters. Adaptive context management and a multi-dimensional quality review system further enhance creative output.

Quick Start & Requirements

Installation is performed via Go: go install github.com/voocel/ainovel-cli/cmd/ainovel-cli@latest. Local development can be initiated with go run ./cmd/ainovel-cli. Initial setup requires configuring LLM providers (e.g., OpenRouter, OpenAI, Anthropic, Gemini) and API keys, managed through a guided process or configuration files (~/.ainovel/config.json, ./ainovel.json). The system supports various LLM providers and local models like Ollama.

Highlighted Details

• Multi-Agent Orchestration: Coordinator, Architect, Writer, and Editor agents collaborate autonomously.
• Long-Loop Creation: Single-prompt initiation with LLM-driven decision-making for the entire novel lifecycle.
• Step-Level Recovery: Robust breakpoint recovery ensures progress is saved after each tool execution.
• Rolling Planning: Dynamic generation of plot arcs and chapters, adapting to story progression.
• Context Management: Advanced strategies for handling long narratives (500+ chapters) using tiered summarization and context compression.
• Seven-Dimensional Quality Review: Editor agent rigorously assesses content across multiple criteria.
• Real-time Intervention: Users can inject feedback during writing without pausing the process.
• LLM Flexibility: Supports a wide array of LLM providers and models.

Maintenance & Community

The project actively participates in the linux.do community. No other specific community channels or contributor details are immediately apparent from the provided text.

Licensing & Compatibility

The project is released under the MIT License, which is highly permissive for commercial use and integration into closed-source projects.

Limitations & Caveats

The system's effectiveness is inherently tied to the performance and coherence of the underlying LLMs. Initial setup requires careful configuration of LLM API access. The design philosophy emphasizes simplicity and LLM-driven decision-making, which may limit highly granular, code-level control over the creative process.