yuanbw2025/storyforge

What it solves

StoryForge is designed for long-form fiction writers who want AI assistance without losing creative control. It solves the problem of "black box" AI writing tools that generate entire stories automatically, instead providing a structured environment where the author remains the primary decision-maker. It specifically addresses the challenges of maintaining consistency in complex, long-form narratives—such as tracking world-building, character arcs, and plot holes—by providing a structured local database for story elements.

How it works

The project is a local-first, browser-based studio. It uses IndexedDB for local storage of manuscripts and settings, ensuring data privacy and autonomy. Users "bring their own AI" by connecting to various OpenAI-compatible providers, Anthropic, Gemini, or local models via Ollama/LM Studio.

The system operates on a three-layer AI process:

1. Context Assembly: Gathers relevant project data (world-building, character sheets, outlines) to feed into the prompt.
2. Prompt Rendering: Uses customizable templates with variables and conditional blocks to generate the final prompt.
3. Structured Adoption: AI outputs are presented for review and editing before the author chooses to "adopt" them into the project database or manuscript.

Who it’s for

• Novelists and long-form creators: Those writing complex series or multi-world narratives with many characters.
• Control-oriented authors: Writers who prefer to edit and transparently manage the prompt engineering behind the AI's suggestions.
• Privacy-conscious writers: Users who want their data stored locally rather than on a cloud server.

Highlights

• Comprehensive Story Database: Dedicated tools for managing world-building, character relationship networks, plot arcs, foreshadowing (planting and recovery), and item ledgers.
• Transparent Prompting: All system prompts are visible, editable, and cloneable, allowing authors to create their own style templates.
• Local-First Architecture: Data lives in the browser's IndexedDB with optional backups to GitHub Gists or local folders.
• Style Learning: Extracts author-specific writing styles from confirmed samples to improve subsequent AI generations.
• Broad AI Compatibility: Supports a wide array of international and domestic (Chinese) AI providers and local LLMs.