AI-Powered Editorial Review Workflow: Local LLM Automates Manuscript Quality Check

Project Core Information This is an open-source project developed by Zoidberg2021 (GitHub link: https://github.com/Zoidberg2021/AI-Powered-Editorial-Review-Workflow, released on June 12, 2026), aiming to automate manuscript review using the n8n workflow orchestration tool and Ollama local large language model (running Qwen 14B). Key advantages include:

• Significantly reduced review time: per chapter from 45-60 minutes to 2-5 minutes
• Privacy protection: all content processed locally, no risk of third-party leakage

Pain Points of Manual Review Developmental editing is time-consuming and error-prone for novel authors and editors:

• Each chapter requires 45-60 minutes of manual review, with repeated switching between manuscripts, character setting sheets, plot outlines, and style guides
• Easy to overlook details, making consistency hard to guarantee

System Workflow and Architecture The complete process has four steps:

1. Document Acquisition and Integration: Retrieve character bibles, story outlines, style guides, and manuscripts to be reviewed from Google Docs, integrating them into an analysis package
2. Intelligent Chapter Segmentation: Automatically split manuscripts into independent chapters for precise problem localization
3. Multi-dimensional AI Analysis: Local LLM performs character continuity checks, outline compliance reviews, style guide adherence assessments, narrative structure analysis, and AI-generated text detection
4. Structured Feedback Generation: Output an integrated report containing continuity issues, editorial notes, AI text markers, missing outline nodes, and a summary of strengths

Tech Stack Selection

• n8n: Visual workflow orchestration with strong integration capabilities and easy maintenance
• Ollama + Qwen14B: Local deployment; 14B parameters can run on consumer-grade hardware; excellent performance in both Chinese and English; privacy guaranteed
• Google Docs Integration: Connects to commonly used writing platforms without changing existing habits
• JavaScript/Node.js: Natively supported by n8n, facilitating custom logic

Effectiveness Evaluation Data

Metric	Manual Review	Automated Review	Improvement Rate
Per-chapter review time	45-60 mins	2-5 mins	90-95%
Continuity error omission rate	High	Significantly reduced	-
Review consistency	Dependent on editor's state	Stable output	-
Example: Review time for a 20-chapter novel reduced from 15-20 hours to 1-2 hours, freeing up editors' creative energy

Practical Insights and Future Directions

• Key Insights: Context quality determines output effectiveness; structured prompts improve result consistency; workflow design is more important than model selection; document segmentation enhances quality and robustness
• Future Plans: Features like automatic severity scoring, character relationship tracking, multi-model comparison review, PDF export generation, etc., evolving into a complete collaboration platform

Applicable Scenarios and Value

• Applicable Scenarios: Series novel creation (vast worldviews/characters), collaborative writing (style unification), editing services (efficiency improvement), writing workshops (teaching tools)
• Core Value: Provides privacy-first practical AI auxiliary tools for independent authors and small publishing teams, liberating mechanical work and focusing on creative judgment