Caveman Writer Studio v28.0 — Naming canon: what you dictate is canon

v28.0 — naming canon: what you dictate is canon

Naming — the names and places you dictate are canon; the genre becomes the default:

• Register detection — from your set protagonist / antagonist the app infers the naming register (greek / western / invented). Set «Leopold» and the western register follows; set «Δημήτρης» and everyday Greek; set «Μοραΐνα» and invented — the cast stays consistent
• The local roll (🎲 offline) follows your book — people come from your dictated hero's register (no longer the genre's pool), and places are reused from what you already dictated
• Your dictated places are canon — place names found in your glossary / world / bible / premise are extracted and reused (labelled “canon — from your plan”) instead of a fixed imaginary pool
• CANON block — the names & places and character prompts carry your fixed names/places with an explicit “never mix registers” rule; the genre rule is demoted to the default (it applies only where you haven't dictated)
• The main brief carries a Naming (canon) line under the protagonist; the mythos prompt carries your dictated places; character suggestion chips switch register with your hero (greek / western / invented sets)

Stronger commission:

• Commission strength meter — a craft rubric, not a field count; the top gaps appear as clickable diagnostics that focus the missing field
• Dramatic question (the engine) — the one line the whole book answers
• Craft locks — must include / must never do / comps → a CONSTRAINTS section in every prompt (the trap named, the ban named; comps calibrate craft, never content)
• Character sheet: the need (what the book will force them to become) + the lie they believe
• Stage pack (📦) — five self-contained stage prompts (commission → world → souls → voice → chapter 1); each stage embeds the canon facts — paste one at a time, in order

Repairs — byte-level fixes for Greek typos (double tonics, tonic placement, a broken verb, mixed-script remnants) plus subtle detector bugs (U+0390/U+03CA normalization, placeholder double-replace, proper-phrase trimming).

Tests: bulletproof 98 + v28smoke 64 + naming 30 — all green, stable across repeated runs; CI, GitHub Pages and release assets are automated.

Caveman Writer Studio v27.0 — Spark v3: segment sparks for every segment

v27.0 — the spark technique for every segment, genuine + morphology-matched

The spark technique (v25) now applies to all suggestion segments:

• ✨ chips next to Title / Story / Character suggestions — each builds a crafted prompt for YOUR AI, with full context (genre, book language, premise/world consistency, the selected masters' rhythm — King, Tolkien, Herbert, Lovecraft… — or the blog's DNA when no master is selected)
• GENUINE block in every prompt — never reuses existing suggestions (they are listed) or famous works; every suggestion must be publishable
• Fixed-label outputs — TITLE 1–6 / STORY 1–3 (multi-line) / CHAR 1–3 (with the genre naming discipline); ⤵ Load detects the answer type on its own and fills the matching segment — pools persist, chips re-render, idea flow unchanged
• Craft baked into every segment prompt: precise opening, people-first (a flaw that is a door), the mundane as threshold, the bodily detail, dated documents, the one-line cover-up, the image-ending

Also fixed a latent test flake (usage counter substring) — 95/95 checks, stable across runs.

Craft-loop note: the park demo's Act 1 (graded 9.5/10 line-by-line) validated the sealed 5-act prompt — the blog's DNA is now in every spark.

Caveman Writer Studio v26.0 — Short story + EPUB/PDF export

v26.0

Fix in this release: restored a missing </style> closing tag that had rendered the app dead on GitHub Pages (the page loaded but no JavaScript ran). The app file in this release is the fixed build; 92/92 checks including new <style>/<script> balance guards.

📓 Short story format — the size you actually publish

• 4th format: 3–7 acts × ~1,200 words (3,000–12,000 total) — the standalone territory (Μύηση, Ο Δυρρών, Μαρεστάν…)
• Its brief: one beat per act (set-up, turn, aftertaste), ends on a hook; one world, one cast, one thread; image-ending (the last sentence reframes the whole story)
• Word slider now starts at 3,000; hint shows acts for this format

📚 Export — hand a book to an editor

• 📖 EPUB — dependency-free EPUB3 (store-only ZIP writer + CRC32 + metadata + nav + title page + one XHTML chapter per chapter, XML-escaped Greek). Opens in any e-book reader. The single-file/offline contract holds — zero libraries.
• 🖨️ PDF — print-ready A4 manuscript layout (title page, page-break per chapter, justified serif) → browser print dialog → Save as PDF

Also in this release (v23–v25, previously unreleased)

• v25: the spark technique for Names & places and the Character sheet (crafted AI prompts + ⤵ Load paste-back, 🎲 local offline roll kept as fallback)
• v24 + UI: signature techniques (blog DNA) in every brief, two demos, the app file renamed to caveman-writer-studio.html, and the stone & fire UI (carved stone-tablet cards, ember buttons, tally-stick savings bar, guided first view)

92/92 checks · single file · offline · EN/ΕΛ

Caveman Writer Studio v24.0 — the app learns its own DNA

v24.0 — signature techniques + author-craft spark + two demos + named app file

Built from a full read of the blog (all 7 series, all standalones):

• Signature techniques in every brief — the blog's structural DNA, injected into horror/weird briefs: the mundane as threshold, dated found documents that prove (never explain), one-line institutional cover-up, the bodily price of knowledge, the entity that REMEMBERS (never destroyed), the sound motif that returns changed, the image-ending. Plus precise first line + the story must smell in all genres.
• House style (My profile) +5 rules from the full read.
• Spark v2 carries the author's craft — precise opening, "it must smell", the wrong detail, the remembering thing; HOOK sharpened.
• Two demos, one picker — «Το Πάρκο που Θυμάται» (the sealed 5-act retro-liminal brief) + «Η Γέφυρα των Ψιθύρων».
• App file renamed — index.html → caveman-writer-studio.html (small branded index.html entry auto-opens it; Pages site unchanged).

v23.0 (also in this release)

• Genre-aware naming — fantasy = invented names & places only, realistic Greek = everyday names & real places, across character generator, names & places, build-from-ingredients and mythos.
• Spark v2 — no more rolling among predefined ideas: a crafted spark prompt for YOUR AI + «⤵ Load» paste-back.

74/74 checks · single file · offline · EN/ΕΛ

v1.8.0 — Build from ingredients + naming + mythos

v1.8.0 — Build from ingredients + imaginative naming + mythos

The "no idea yet" wizard

• ⚡ Build from ingredients — pick what kind of story (cosmic horror · high fantasy · post-apocalyptic · environmental SF · gothic · noir · mythic fairy tale · mirror/psychological), mood, setting, elements → the app builds a full brief: logline, world (with invented places), hero + background, antagonist, themes, glossary, lore. 🎲 Surprise me for random.

From the author's blog vocabulary

• 🏷 Names & places — western/Greek/invented names + places like Μάρεςταν / Ξαναντού / Δυρρών, objects from the blog's world
• 🌍 Mythos & worldbuilding expander + 🧠 Lore field injected into every brief
• 👤 Character background builder (origin, the day that changed everything, keepsake, the line they'd never cross)

Polish

• Intent logline now uses the full genre label; whole UI re-renders on EN/ΕΛ; always starts on the main screen; refined Parchment clarity theme.

Verified: 59/59 test suite · all wizard/name/mythos checks green.

v1.6.2 — start on main screen

v1.6.2 — always start on the main screen

Every time you open the app it now resets to the main screen:

• Back to the first tab (Scene tools)
• All collapsed cards expanded · quick-start visible
• Focus & compact modes off
• (Your theme choice is kept; your book data is untouched)

No more getting "stuck" in a submenu you can't leave.

v1.6.1 — fix bilingual toggle

v1.6.1 — fix: bilingual toggle now translates the whole UI

Bug fix — clicking EN/ΕΛ previously only switched the static labels; tabs, manuscript chips, story-memory hints and co-author POVs stayed in the old language until reload. Now the toggle re-renders the entire interface instantly (via a full refresh), with a resilient fallback so it can never be blocked by a glitch.

Verified: 59/59 test suite + dedicated toggle checks (labels, tabs, chapter chips, status, co-author, memory) pass in both directions.

v1.6.0 — UI clarity pass

v1.6.0 — UI clarity pass

The default Parchment (light) theme is now tuned for maximum clarity and reachability:

• White input fields on warm paper → fields stand out
• Deeper text + clearer borders → better contrast
• Selected states are obvious: active tabs, chips, levels and author cards fill with the accent colour
• Focus rings for keyboard users; bigger touch targets on chips/checks
• Wrapping author grid — the 19 masters no longer squeeze into tiny cards
• Quick-start card highlighted with an accent border

Dark themes (Night / Forest) kept and harmonised. Docs updated.

v1.5.0 — searchable masters + English writing

v1.5.0 — searchable masters + write in English

New

• 🔍 Author search — find any of the 19 masters by name or technique; live count + ✕ Clear all. No more scrolling through an overwhelming grid.
• 🌍 Deliverable language — choose Greek or English for the book itself, independent of the UI language. The role, language guard, continuation and extender prompts all adapt (Greek guard ⇄ English guard).

Docs: README + Guide updated.

v1.4.0 — word goals, reliability, bulletproof tests

v1.4.0 — word goals, reliability hardening, bulletproof tests

New

• 🎯 Per-chapter word goal — set a target per chapter (default = format size); chart bars and stats show words / goal
• 🔍 Chapter search — filter the chapter list instantly

Reliability ("never broke")

• Resilience-hardened $: a missing element becomes a safe no-op — nothing can crash the app
• Global error handler with a friendly notice; autosave + quota fallback; guarded empty-state actions
• Bulletproof test suite — tests/bullet.js: 59 checks · 5 angles (syntax, structure, functional, robustness, regression), all passing

Docs: README + Guide updated (word goals, search, reliability, how to run tests).