Frequently Asked Questions

The Writers Factory is a significant software engineering project, not a simple wrapper. It consists of approximately 105,000 lines of code:

• Backend (~50k LOC): A complex Python FastAPI application with 36+ modular services. It features:
  • Multi-Agent Orchestration: Logic for routing tasks between local (Ollama) and cloud (Claude/GPT) models.
  • Graph Database Logic: Extensive use of NetworkX to model narrative state and detect inconsistencies.
  • Asynchronous Pipelines: Heavy use of asyncio for parallel model execution (Tournaments).
• Frontend (~55k LOC): A modern SvelteKit & TypeScript application using Tauri for desktop integration. It includes complex state management for real-time writing and visualization.

Opportunities for Architecture Track Improvements:

• Algorithm Optimization: Improve the graph_health_service.py to detect plot holes faster in large graphs.
• Model Routing: Enhance the model_orchestrator.py to better select models based on cost/latency trade-offs.
• Visualization: Create better D3.js/Canvas visualizations for the "Story Graph" in the frontend.
• RAG Pipelines: Optimize how the Research Graph and NotebookLM integration retrieves context for the writer.

It depends entirely on your chosen track:

• Writer Track: No technical background required. We strongly recommend completing the NotebookLM Research before Day 1 to ensure you have raw material to work with.
• Architect Track: Requires real Python/TypeScript experience (not just tutorials), comfort reading 10,000+ line codebases, and willingness to invest serious time understanding the architecture before contributing. See the Tracks page for the honest requirements.

See the Tracks Comparison page for a detailed breakdown.

Not for the Writer Track. The application handles all the technical complexity. You interact through natural language conversations with the built-in AI assistant.

The Architect Track requires coding skills, as you'll be modifying the codebase directly.

The course runs 8 sessions across 2 weeks, with 3 hours of class time per session (12:30-15:30). Total workload: 33 hours (25 contact hours + 8 hours homework). The compressed timeline is intentional—it creates urgency that pushes through perfectionism and procrastination. Ambitious writers who want to hit 40,000+ words will need to invest additional time outside the minimum requirements.

Week 1 (Jan 14-16): Foundation

• Day 1: NotebookLM setup and research gathering
• Day 2: Distillation and the 5 Research Categories
• Day 3: Story Bible generation and Voice Calibration

Weekend: Start drafting (2,000-3,000 words homework)

Week 2 (Jan 19-23): Production

• Days 4-6: High-velocity drafting with Tournament Mode
• Days 7-8: Presentations and showcase

See the full schedule for details.

Everyone produces a complete novella. The minimum requirement is 20,000 words, but ambitious writers should aim for 40,000-50,000 words - a full short novel. This is a first draft, polished enough to share but rough enough to improve later.

Architect Track participants also contribute technical improvements (bug fixes, prompt optimizations, documentation) via Pull Requests.

The course is graded Pass/Fail. You must complete 4 out of 6 criteria to pass:

1. Environment Setup (Days 1-3): NotebookLM with 5+ sources, Writers Factory installed, Story Bible generated
2. Story Specification (Days 3-4): Submitted story spec (logline, character, beat sheet) and completed Voice Calibration
3. Drafting Progress (Days 5-6): Generated at least 10,000 words of narrative (or identified technical improvement area for Architects)
4. Revision & Analysis (Day 6): Used at least 1 diagnostic tool and produced polished excerpt (2-3 pages)
5. Final Deliverable (Days 7-8): Submitted complete manuscript (20,000+ words) or technical proposal (3+ pages with working code)
6. Presentation (Days 7-8): Delivered 10-15 minute presentation and responded to Q&A

See the full rubric for details.

Writers Factory supports 15+ AI models across multiple regions:

• US: OpenAI (GPT-4o), Anthropic (Claude), Google (Gemini), XAI (Grok)
• Europe: Mistral AI
• China/Asia: DeepSeek, Alibaba (Qwen), Moonshot (Kimi), Zhipu, Tencent
• Russia: Yandex AI (YandexGPT 4)
• Local: Ollama (Llama 3, Mistral, Gemma - runs on your machine, no API costs)

Core functionality works with local models only. The course includes "tournaments" where you compare how different models handle the same creative task.

See the Models Overview for full details.

Bring Your Own Key is required for first-tier US models (OpenAI, Anthropic, XAI, Google). All other models—including Chinese, Russian, and European providers—are included during the MVP trial period. A shared Yandex key is provided for Russian-language writing.

Core functionality works entirely with Ollama (local, free). Cloud APIs unlock advanced features and tournament comparisons.

See the Models Overview for the full access policy.

The application includes a built-in chat interface designed to help you through creative blocks. It will challenge weak ideas, suggest alternatives, and maintain context across your entire project. If you're truly stuck, the instructor is available during class hours or via Telegram.

Yes, and you should! Come with a concept you're excited about. The methodology works best when you have genuine passion for your story.

In practice, the system builds your novel from research you provide via NotebookLM. The more personal material you feed into your research notebooks (your notes, inspirations, character sketches), the more the generated story will reflect your unique vision. The course will guide you through this process.

It's yours. You retain full rights to everything you create. The application stores everything locally on your machine. Nothing is sent to external servers without your explicit action.

The Writers Factory codebase is closed source and proprietary. Architect Track participants receive access to the private GitHub repository for the duration of the course.

To contribute, you must sign a contributor agreement (provided upon access). Contributions are licensed under the project terms, and contributors may receive public acknowledgment for significant improvements.

See the GitHub Access page for how to request repository access.