A self-publisher was spending days on every low-content book — researching the Amazon shelf, designing pages, laying out the interior, checking it, drawing a cover, writing the listing. We turned the whole routine into one unattended pipeline that takes a list of titles and returns finished files.
Low-content books — planners, journals, logbooks — live or die on the Amazon shelf. Every title needed the same long routine: read the competition, decide what the book must do better, lay out a hundred-plus pages, check every one, draw a selling cover, write the listing. Done by hand, that is days per book, and the quality depends on how tired the person is.
The same path for every subject — a food journal, a pilot logbook, a homeschool planner. Nothing in the pipeline is tailored to one book.
The live Amazon search page for the title, fetched the way a buyer sees it today. Every listing is parsed — page count, price, rating, reviews — and a model first throws out the books that are about something else. Medians come only from the relevant ones.
A research brief: who buys the book, what the best competitors put on their pages, what reviewers ask for and do not get. The plan takes the best of three or more books into one list, removes duplicates and weak ideas, and adds something of its own — a synthesis, never a copy.
The design calls lay out the daily page, the front and the back; a JavaScript builder renders the 6×9" PDF with KDP margins, black ink on white paper. Checkboxes and scales are drawn shapes, not typed characters; every page carries real content.
Machine QA runs while the PDF renders. Then a second model reads the whole book against the brief and grades every finding — critical, important, not important. Only real defects hold a book back; fixes touch only what a finding names, and the book is rebuilt and audited again.
Code computes the full wrap from the final page count — trim, bleed, spine, safe zones — and lays down a stencil that differs from book to book. GPT Image draws the cover over it, text included; code only keeps everything inside the margins and nothing on the spine.
An SEO title that opens with the search phrase as typed, a selling description with keyword variants and the right disclaimer, seven keywords and two full KDP categories — plus the trim and page count. Book, cover and listing land in one archive.
The run path of the workflow, node by node — what each one does and why it exists. Seven model calls, two IF gates, one loop; everything else is code.
Kicks off a run. There are no operator forms anywhere after this point — the workflow is designed to be started and left alone.
Input & configThe list of titles for this run, one per line. Every title becomes its own item and travels the whole pipeline independently.
Input & configSingle source of settings: API endpoints behind the Kie.ai proxy, which model handles each step (provider_competition, provider_brief, provider_design, provider_fix), time budget per node, output folder on disk.
Input & configNormalises the raw list — trims, deduplicates, builds a slug for the output folder — and emits one clean item per title.
Input & configFetches the live Amazon search page for the title through ScraperAPI, so the run sees the same shelf a buyer sees today, not a cached guess.
Market researchExtracts every listing from the results HTML: title, ASIN, page count, price, rating, review count. Sixteen cards in a typical run.
Market researchOpens the top listings' own pages to read what the search results don't show: the look-inside page structure, the description, the cover.
Market researchA model reads the shelf and first drops the off-topic listings — the top for a query often contains books about something else. From the relevant ones it derives medians for page count and the features buyers expect.
Market researchTurns the streamed model answer into structured JSON and degrades gracefully if the stream arrived truncated — the run continues with what it has instead of dying.
Market researchWrites the plan for this book: who buys it, what the 3–5 best competitors do, and what this one must do better — more useful pages, more of them, the fields readers ask for in reviews.
Brief & designValidates the brief against a schema: page target, required sections, required fields on the daily page. Missing pieces are filled with defaults, not silently ignored.
Brief & designDesigns the repeating worksheet — every field, its size and order. Split out into its own call so it fits the API gateway's two-minute window.
Brief & designChecks the page design: a DATE field must exist, no owner or personal-data fields may exist, columns must add up to the page width.
Brief & designFront matter: title page, copyright page (© and year, never an author), how-to-use page, any reference pages the brief demands.
Brief & designSchema check for the front matter and the copyright line; rejects placeholder text and empty pages.
Brief & designBack matter: review pages, summaries, notes — sized to reach the page target without padding the book with duplicates.
Brief & designMerges the three designs into one BookSpec: the ordered page list with repeat counts, plus the trim size, margins and paper (always white).
Brief & designThe largest node — a JavaScript PDF builder that renders the 6×9" interior with KDP margins. While rendering it runs machine QA: duplicate pages, missing date fields, forbidden owner fields, dead vertical space, column widths, table cell text, copyright presence.
Build & audit gateA second model reads the finished book against the brief and lists findings by severity. The machine QA report is passed in so the auditor adds to it instead of repeating it.
Build & audit gateCounts what blocks a release: critical and important findings plus the builder's own blockers. This number — not a feeling — decides whether the book may ship.
Build & audit gateIF node. Blocking findings and rounds remaining → go fix. Nothing blocking → move on. Out of rounds and still blocking → the book fails with the report attached.
Build & audit gateClaude rewrites the BookSpec to resolve every finding — the audit is the diagnosis, this is the treatment. Runs up to three times per book.
Build & audit gateValidates the corrected spec and sends it back to Build interior for a full rebuild, so fixes are verified on real pages, not assumed.
Build & audit gateIF node. Guarantees the final rebuild is audited before the cover is drawn — a fixed book is never trusted on its word.
Build & audit gateComputes the full-wrap cover from the final page count: trim, bleed, spine width, safe zones — exactly as KDP's cover calculator would.
CoverEstimates KDP printing cost from page count and trim, so the listing price on the back cover and in the metadata is realistic.
CoverSends the cover brief to GPT Image through Kie.ai — create task, poll until ready — and requests three variants that follow the shelf's selling formula: big title with an accent word, subtitle band, icon row, promise on the back.
CoverMeasures each variant: text inside the safe margins, nothing on the spine fold, no reserved barcode box. Saves the three covers as PNG.
CoverWrites the interior PDF and the run's metadata into the title's folder on the server's disk.
DeliveryWrites the SEO title — search phrase first, keyword subtitle after — and a selling description with keyword variants, plus the right disclaimer (medical for health topics, general otherwise).
DeliveryCollects every step's result — timings, counts, findings, files — into one text report saved next to the book.
DeliveryLoop node. Returns to the next title in the list until the run is exhausted; each title gets a fresh budget of time and fix rounds.
DeliverySummary for the whole run: which books shipped, which failed the gate and why.
DeliveryThe rules are written once, for any subject, and checked by code before and after each build — so a mistake made in one book cannot quietly appear in the next.
No owner name, contact or reader-ID fields — not on the sheets, not on a quick-start page, nowhere.
A © line with the year on exactly one page, centred at its foot. No author, publisher, ISBN or “printed in” line the pipeline cannot know.
Medical for health subjects, general for the rest — always present, always last, never dropped because another page mentions it.
Every page carries real content. No half-empty sheets, no white voids, no two headings in a row, no rating rows stacked one under another.
Checkboxes, scales and icon circles are drawn shapes, compact and centred; a cell that exists only for numbering is square.
A fix round touches only what a finding names — no new book each round, no new pages unless asked, and the engine checks the spec before every rebuild.
Research, design, layout, audit, cover and listing run unattended. Quality is enforced by a gate that counts defects, not by a checklist someone has to remember at the end of a long day.
Describe it in two sentences — we reply within a day with a pipeline sketch.
hello@wireclad.com →