Writer ghost · research preview

AI Library

Turn a source book or long text into a reusable writer ghost — not a chat knowledge base. Then generate a new manuscript on your chosen topic in the author's lexical style, review it, and optionally share it in the public library.

  • 1 free book per email
  • Additional books $1.99 · public listing $0.99
  • Output must be ≥ 50% of source word count
  • Workflow: draft → edited → published
How it works

Research preview — email only for your first free book. Security keyword optional for unlimited access.

Source book pages transform into a writer ghost made of words, then into a new manuscript — AI Library style transfer for books
Source text → writer style model → planned manuscript → your edits → library

What AI Library is — and is not

Many users confuse this with a RAG chatbot or plagiarism tool. Here is the distinction.

This is

  • Style mining from your uploaded source
  • A new book on a topic you choose
  • Draft generation you must review and edit
  • Optional public library sharing

This is not

  • A Q&A knowledge base about your PDF
  • Automatic publishing without your review
  • A substitute for rights clearance on source material
  • Guaranteed commercially publishable output

How AI Library works

Five clear steps from upload to publish. The workspace guides you through each one after you sign in.

Five steps: Upload source, Train writer style, Generate draft, Validate quality, Edit and publish

Upload source text

Paste text or upload a .txt or .pdf. We count words and enforce the 50% minimum output size.

Train the writer ghost

We learn writing patterns from your source and build a reusable style profile for new manuscripts.

Generate your book

Set title, topic, and target pages. An async pipeline plans chapters and composes a full draft manuscript.

Validate quality

Automated checks for verbatim copy, ghost-style match, and hallucination risk before you publish.

Edit & publish

Revise the markdown draft, choose public or private, then publish to your library when validation passes.

Built in-house · Insight IT Solutions

Proprietary retrieval technology

Powered by PrismResonance

AI Library runs on PrismResonance — a memory-retrieval library designed and built entirely in-house at Insight IT Solutions. No third-party RAG framework. No vendor lock-in. We own the full stack.

How it works — in plain terms

  1. Standard RAG retrieves context by finding the closest embedding vector to your query — like a nearest-neighbour search. It treats every chapter in isolation, with no memory of what was retrieved before.
  2. PrismResonance encodes each knowledge-graph community as a wavepacket with an amplitude (what the community is about) and a phase angle (what pipeline stage it belongs to). Retrieval happens via wave interference — communities resonate with the chapter topic if their waves align, and cancel out if they don't.
  3. The result: communities that get cited together in chapter 2 drift toward the same phase and naturally surface together again in chapter 7 — giving your manuscript a coherent thematic arc without any manual chapter planning.
  • Cross-chapter coherence — recurring themes self-reinforce automatically
  • Phase-tagged stages — active (NORMAL), retry (ALERT), done (ARCHIVE)
  • No torch dependency — pure ONNX runtime, runs on CPU in production
  • pgvector native — reads your existing embeddings, zero migration

Each wave is a knowledge-graph community. Constructive interference with the chapter query = high relevance. Destructive interference = filtered out by the coherence shield.

Standard RAG Cosine similarity · no memory · every chapter starts cold
vs
PrismResonance Wave interference · Hebbian phase drift · thematic arcs form naturally

Enterprise-grade composition

Beyond the free-tier writer ghost — controls for quality, genre, and voice fidelity at scale.

Style Intensity

A 0–100% slider controls how strictly the author's voice is enforced. Low intensity gives the AI creative freedom; high intensity locks to the source cadence and vocabulary.

Genre Presets

Auto-detected from your domain prompt or manually selected: Business, Memoir, Fiction, Technical, Science, Self-Help. Each preset tunes prompts, temperature, and structural expectations.

Auto Chapter Retry

Chapters that fall below the word-count target are automatically retried once with an extended token budget — no manual intervention needed.

Knowledge Graph RAG

Entity extraction → Louvain community detection → PrismResonance wave retrieval. The knowledge graph ensures every chapter is grounded in real concepts from the source, not hallucinated filler.

6-Point Validation

Automated quality checks on every manuscript: copy ratio, verbatim run detection, ghost lexicon match, cadence match, novel content ratio, and optional Gemini hallucination review.

Audit Trail

Every generation is logged with job ID, pipeline stage timings, validation scores, and entitlement events — ready for enterprise compliance review.

Sample output

Generated from The Little Prince by Antoine de Saint-Exupéry (17,977 words). The AI Library extracted the author's voice and wrote a new manuscript in the same style.

New title The Little Prince — A New Journey Source: 17,977 words  ·  Output: 6,911 words  ·  4 chapters  ·  35 pages

Chapter 1: The Weight of the World and a Child's Gaze

When I was six years old, I saw a magnificent picture in a book. It was about the primeval forest. The book was called "True Stories from Nature." It showed a boa constrictor swallowing a wild beast. The book said that boa constrictors swallow their prey whole. They do not chew it. After that, they are unable to move. They sleep for six months to digest.

I showed my masterpiece to the grown-ups. I asked if my drawing frightened them. They answered, "Why should anyone be frightened by a hat?" My drawing was not a hat. It was a boa constrictor digesting an elephant. Grown-ups never understand anything by themselves. It is tiresome for children to be always explaining things.

So I chose another profession. I learned to fly airplanes. In the course of my life I met many people who seemed clear-sighted — but they always said "That is a hat." They would say, "You should talk about bridges and golf." So I kept to myself. I talked to no one about my true interests. This was until six years ago, when my plane had a breakdown in the Sahara Desert, a thousand miles from any inhabited land…

— Excerpt from the AI-generated manuscript. The full 6,911-word manuscript is available in the public library below.

Public library archive

Books shared by the community — read without signing in. Search, sort, or export; click Read to open the full manuscript.

Loading public library…

Common questions

Quick answers so you know what to expect before entering the workspace.

What is the security keyword?

A shared workspace access keyword (provided by your administrator) plus your email. The keyword tracks entitlements — one free book per email — and unlocks payments.

Why must the new book be at least 50% of the source length?

AI Library is designed for substantial re-authoring in the source voice, not short summaries. The UI blocks targets below half the source word count.

Can I use someone else's book as source?

Only if you have the legal right. You must accept a rights attestation before generation. We log acceptance for audit purposes.

What happens after I click Generate?

A background pipeline runs (typically a few minutes). Stay on the page — progress updates automatically, then the editor opens with your draft.

Ready to try AI Library?

Enter the workspace with your security keyword and email. Your first book is free.

  • Upload PDF or pasted source text
  • Generate, validate, and edit in one place
  • Share finished work in the public library (optional)

Start now

Need help? Email info@insightits.com