Pagekin Operational

Pagekin is a free, open-source AI desktop app for writing long-form novels, with support for Windows and macOS. Pagekin breaks novel writing into a workflow of setting, architecture, blueprint, drafting, review, revision, and finalization. AI handles drafting and assistance, while the author confirms every step. Chapter-level context management helps prevent memory degradation during long-form writing. Pagekin does not include a built-in model. You need to connect an API such as DeepSeek or run a local model with Ollama. All work data is stored locally.

Follow us on WeChat and reply “开源” to join the AI open-source project group
Creative workflow: Break novel writing into seven steps: setting, architecture, blueprint, drafting, review, revision, and finalization. AI drafts while the author confirms each step.
Chapter-level context: When writing each chapter, AI reads only that chapter’s blueprint, relevant characters, summaries of earlier chapters, and the full text of the previous chapter, preventing long-form memory degradation.
Review-to-revision diff confirmation: Compare revision results with the original draft paragraph by paragraph. The author decides what to keep, so nothing is overwritten with one click.
Deconstruction and imitation: Import reference novels in .txt/.md/.epub format. AI analyzes their structure, extracts their writing style, and infers chapter blueprints for imitation.
Continuation of drafts and paragraphs: Import completed chapters as finalized text and have AI continue writing from them. You can also select a paragraph in the manuscript and ask AI to continue it.
Writing Skill system: Import prompt-based SKILL.md files and bind them to stages such as planning, drafting, and review for automatic application.
Plot and foreshadowing tracking: Record the setup, development, and resolution of main plots, subplots, and foreshadowing by chapter, so long-form writing does not lose track of clues.
Work management and version history: Manage all works in one library. Compare or roll back previous drafts for every chapter.

Download and install: Download the installer for your operating system from GitHub Releases.
Configure a model: In “Settings → Models,” enter the provider URL, model name, and API Key. You can connect a local Ollama model according to the instructions.
Create a work: Return to the library and click “New Work” to start your first novel.
Fill in the settings: Specify the genre, audience, writing language, total number of chapters, and word count per chapter.
Build the architecture: Complete the story premise, character list, worldbuilding, and plot outline in four sections. Each section can be generated by AI and then confirmed by you.
Plan the blueprints: Define the protagonist’s short-term goal, conflicts and turning points, suspense hooks, and appearing characters for each chapter. AI can generate them in batches by chapter range.
Write the manuscript: Click “Write Chapter N” to select a model and provide author guidance, or use “Batch Creation” to write up to 10 chapters consecutively.
AI review: Start a review after generating a draft. The report lists revision suggestions one by one. Select the items to address or add your own comments.
Confirm revisions: After AI rewrites the text according to the suggestions, compare it with the original paragraph by paragraph and decide which version to keep.
Finalize and export: Finalize the work after confirming it is ready. You can then export it as a combined or chapter-by-chapter Markdown or TXT file, and use WebDAV for cloud backup.

Chapter-level context: When writing each chapter, only its blueprint, relevant characters, summaries of earlier chapters, and the full text of the previous chapter are injected, addressing long-form AI writing memory degradation at the engineering level.
Structured workflow: Setting → architecture → blueprint → drafting → review → revision → finalization, with every step saved for continued work.
Deconstruction and imitation: Import a reference novel to infer its structure, extract its writing style, and generate imitation constraints. Reference materials enter only the knowledge base and do not contaminate the manuscript.
Writing Skill extensions: Import custom prompt rules such as “Remove the AI Flavor” and bind them to specific creative stages. Scripts are never executed, combining safety with customization.
Local-first and privacy-focused: Works are stored in a local SQLite database. The software includes no model and does not manage usage quotas. API Keys remain on your computer, keeping privacy boundaries clear.

GitHub repository: https://github.com/EthanYoQ/AI-Novel-Writer

Daily web-fiction serialization: Batch creation can write up to 10 chapters consecutively, with pause and resume support. Failed attempts retain recovery candidates, while foreshadowing tracking helps maintain a steady daily publishing rhythm.
Completing a long-form novel: The seven-step workflow and chapter-level context management preserve the accuracy of character relationships and foreshadowing across hundreds of chapters.
Modeling a hit novel for a new project: Import a reference novel to analyze its structure and writing style and generate imitation constraints. Reference materials enter only the knowledge base and do not mix into the manuscript.
Continuing and rewriting existing drafts: Import completed chapters as finalized text for AI to continue. Select a paragraph to continue, polish, expand, or convert it into dialogue.
Practice and writing instruction for beginners: Blueprints require conflicts and hooks to be clarified in advance. Load a “Remove the AI Flavor” Skill to learn structured writing while creating finished work.