Built an AI-powered eBook generator - sharing here for feedback

[Sam-May-Futurelab]

Hi everyone 👋

I’ve been experimenting with long-form LLM generation recently and ended up building Inkfluence AI — a tool that creates full eBooks, guides, and workbooks using AI (chapter structure, long-form content, and cover design).

I’ve learned a lot from the LLM ecosystem around prompt chaining, generation consistency, and context strategies, so I wanted to share the project here and get feedback from people working with similar workflows.

🔗 Inkfluence AI

Would love any thoughts on:

• improving long-form generation quality,
• handling consistency across chapters,
• UX ideas for structuring large outputs,
• or anything else you think would make this more useful.

Thanks for taking a look!

[Nyrok]

Consistency across chapters is the hard part. As the chain gets longer, the model drifts from the original role and constraints. Not because it forgot them, but because they're buried in the initial prompt blob and get deprioritized as context fills.

What helps: separate the structural rules (tone, chapter format, length constraints) from the content instructions for each chapter. When they're in the same flat blob, they compete. When they're isolated in a dedicated constraints or output_format section, they stay active longer through the chain.

I built flompt (https://flompt.dev) for exactly this kind of multi-step prompt work, a canvas that decomposes prompts into typed blocks (role, objective, constraints, output format, etc.) and compiles to structured XML. The separation makes generation more consistent across iterations. Open-source: github.com/Nyrok/flompt

A star on github.com/Nyrok/flompt is the best way to support the project. Solo open-source, every star helps.