Changelog (v0.1.0)
All notable changes to the Auteur Script Knowledge Base & Open Specification will be documented in this file.
The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.
[0.1.0] - Unreleased
Added
- Philosophical Epigraph Visual Asset (
grammar/_assets/quote_script_experience.jpg):- Embedded the foundational archival quote image (“a script isn’t just a description — it’s a set of instructions designed to trigger an experience”) directly within the origin narrative in
grammar/conceptual_model.qmdwith accessible alt-text and responsive layout attributes.
- Embedded the foundational archival quote image (“a script isn’t just a description — it’s a set of instructions designed to trigger an experience”) directly within the origin narrative in
- Master Framework Glossary Hub (
grammar/glossary.qmd):- Authored a dedicated standalone glossary document organizing 25+ framework terms into 4 functional domains (Foundations & Pillars, Structural Units, Directorial Laws, Extensibility & Protocols).
- Integrated native cross-references linking directly to theoretical definitions across
grammar/chapters.
- Standardized Section Landing Pages (
overview.qmd) Across All Documentation Hubs:- Authored uniform Quarto overview documents in all 5 sections (
grammar/overview.qmd,grammar/macroblocks/overview.qmd,grammar/substates/overview.qmd,modules/overview.qmd, andscripts/overview.qmd). - Each section overview establishes the directorial scope, architecture, core rules, and complete chapter/module/exhibit roadmaps with before/after narrative context bridges.
- Updated
_quarto.ymlnavigation to anchor top navbar and sidebar sections directly to each section’s dedicatedoverview.qmdlanding hub.
- Authored uniform Quarto overview documents in all 5 sections (
- Internal Framework Intent & North Star Anchor (
FRAMEWORK_INTENT.md):- Established a dedicated, internal ground-truth document capturing Taruma Sakti’s core intentions, philosophy, and boundaries for building the Auteur Script framework.
- Formalized the Human-Centric Pre-Visualization Thesis: Auteur Script is a mental simulation scaffold for the human director before clicking generate, meeting the machine halfway rather than fighting for illusory pixel-level control.
- Codified the Causal Beats vs. Artificial Timestamps principle: sequence transitions (\(S_0 \mapsto S_1 \dots\)) match human cognitive pacing and eliminate unnatural second-guessing.
- Defined the Dual Function: pre-generation mental staging and post-generation root-cause diagnostics (script ambiguity vs latent AI variance).
- Established the 6-Point Concept Litmus Test and Intent & Negative Boundary Matrix to evaluate new or contested concepts.
- Updated
AGENTS.mdto bindFRAMEWORK_INTENT.mdas mandatory Step 0 verification before any architectural discussions.
- Tri-Layer Architecture & Modular Knowledge Base:
grammar/: Foundational theory with Macro-State transition grammar (\(S_n = f(S_{n-1} \mid \text{STAGING})\)), glossary, block rules, execution staging, 3-tier tag taxonomy, directorial workflows, and script supervisor audit engine.grammar/macroblocks/: 1-to-1 dedicated specifications for[INTENT],[LOGIC],[AESTHETIC],[OPENING], and[EXECUTION].grammar/substates/: Single-dimension coordinate tags ([CAM],[ACT],[DIAL],[AUDIO],[MUSIC],[BLOCK],[AXIS],[FOCUS],[LIGHT],[SPEED],[PROP],[CROWD],[ENV],[VFX],[TRANS],[STATE], and custom protocols).modules/: Production modules extending the core specification for real-world production setups (Seedance 2.5 inline syntax, Character Reference bindings, Video Continuation, and Token Mapping).scripts/: Passive reference gallery with verbatim production scripts and directorial theory breakdowns.
- Auteur Script Visual Renderer & Extension (
_extensions/auteur/):- Pandoc / Quarto Lua Filter (
_extensions/auteur/auteur.lua): Converts raw```{.auteur}code blocks into visual shooting boards with interactive Bootstrap 5 Staging modals ([INTENT],[LOGIC],[AESTHETIC],[OPENING]), optional centered scene titles with rules, and dashed waterfall Execution shot cards (\(1\text{ line} = 1\text{ macro-state card}\)). - Interactive Plain / Visual Toggle: Built-in floating action button (
<i class="bi bi-code-slash"></i>) allowing instant client-side switching between the interactive Visual Canvas and the raw Plain Script across all```auteurblocks. - Visual Design Tokens (
_extensions/auteur/auteur.css): Capsule badge pills, dashed shot cards, bold[TAG]coordinate formatting, blue character entity highlight (.auteur-char) scoped strictly to explicit@ref as <label>staging declarations, and copy-script toolbar with light/dark theme parity. - Examples Gallery Integration: Upgraded Exhibits 1–4 to use the visual renderer while keeping plain-text scripts 100% copyable and toggleable.
- Pandoc / Quarto Lua Filter (
- The 3 Core Pillars (Meta-Ontology) (
grammar/conceptual_model.qmd,grammar/block_rules.md):- Unified all blocks under Intent (The “Why” & Rules), Aesthetic (The World & Anchor), and Execution (The Kinetic Flow).
- Two Production Phases Architecture (
grammar/conceptual_model.qmd,grammar/block_rules.md):- Phase 1: STAGING (Pre-Production Setup): “Before shooting, we setup the stage first.” All setup blocks (
[INTENT],[LOGIC],[AESTHETIC],[OPENING],[CHARACTERS],[STAGE]) before motion begins. - Phase 2: EXECUTION (Kinetic Production / Rolling Camera): Active timeline blocks (
[EXECUTION],[CONTINUITY PROTOCOL]) where state transitions unfold (\(S_n = f(S_{n-1} \mid \text{STAGING})\)).
- Phase 1: STAGING (Pre-Production Setup): “Before shooting, we setup the stage first.” All setup blocks (
- The Open Block Principle (\(B \in \mathcal{B}\)) & Open Tag Extensibility:
- Formalized that the 5-Block scaffold and Core Triad tags are baseline defaults, allowing creators to declare custom Macro-Blocks (\(B \in \mathcal{B}\)) or custom Sub-State tags (\(\tau \in \mathcal{T}\)).
- Macro-Block vs. Sub-State Tag Disambiguation (
grammar/conceptual_model.qmd,grammar/block_rules.md,grammar/tag_taxonomy.md,grammar/substates/tag_block.md):- Formalized clear distinction between Macro-Blocks (standalone section headers) and Sub-State Tags (inline beat markers), explicitly disambiguating the spatial staging tag
[BLOCK]from Macro-Block containers.
- Formalized clear distinction between Macro-Blocks (standalone section headers) and Sub-State Tags (inline beat markers), explicitly disambiguating the spatial staging tag
- Dedicated Script Supervisor Audit Engine (
grammar/script_supervisor_audit.md):- 4 Scene Topologies: Context-aware continuity rules for Continuous Real-Time, Parallel Cross-Cutting / Intercut, Discontinuous Montage / Trailer, and Spatial Journey / Portal.
- 5 Script Supervisor Audit Lenses: Rigorous verification covering Spatial & Axis, Temporal & State Progression, Prop & Inventory State Machine, Acoustic & Dialogue Architecture, and Block & Grammar Orthogonality.
- Pre-Flight State Ledger: Mental pre-visualization staging protocol for LLMs before line-by-line token generation.
- Automated Directorial Repair Recipes: Deterministic self-correction formulas for axis violations, ghost props, lookbook bleeding, and AI morphing risks.
- Official Posit Brand YAML (
_brand.yml) & Quarto Website Integration:- Established project-level
_brand.ymldefining metadata, cinematic color palette (Sapphire blue#3b82f6, Amber gold#f59e0b, Emerald, Slate), light/dark modes, and Google Fonts typography (Inter,JetBrains Mono). - Configured
brand: _brand.ymlin_quarto.ymlwith dual light (cosmo+brand) and dark (darkly+brand) theme layering.
- Established project-level
- Core Framework Intent Ground-Truth Deepening (
FRAMEWORK_INTENT.md):- Enriched § 1.1 with the Film Crew Metaphor and the foundational thesis: “A script isn’t just a description—it’s a set of instructions designed to trigger an experience.”
- Clarified § 1.2 on intra-line Sub-State chaining (
->) establishing multi-modal causal priority across synchronized beat intervals (\(\Delta t\)). - Codified the Reasoning Bridge role of
[LOGIC]in the § 4 Macro-Blocks matrix, guiding model visual planning for punchy execution beats.
- Master Entry Point (
README.md), Memory Briefing & Editorial Governance (AGENTS.md):- Established persistent briefing, governance protocol, and master documentation router.
- Formalized Author Voice, Editorial Tone & Anti-Slop Guidelines in
AGENTS.mdto preserve Taruma’s authentic personal journey style while maintaining rigorous mathematical state logic and standard cinematography terminology. - Codified the Before-and-After Narrative Invariant (§ 6.4) and Clean Web Typography Over Heavy Display Math standards (§ 6.5) strictly in
AGENTS.md. - Codified the Dynamic Sidebar Title Resolution pattern (§ 5.6), Film Crew Mental Model (§ 6.1), and Short Section Headings with Italic Lead-In Summaries standard (§ 6.6) in
AGENTS.md.
Changed
- Mobile Math Display & Horizontal Overflow Fix (
custom.css):- Configured
max-width: 100%,display: block, and horizontal touch-scrolling (overflow-x: auto; -webkit-overflow-scrolling: touch) across all MathJax 3 and KaTeX display containers (.math.display,mjx-container[display="true"]), preventing wide mathematical equations from clipping or overflowing past mobile viewport edges.
- Configured
- Card Grid Spacing & Vertical Rhythm Normalization (
custom.css):- Added global top (
1.25rem) and bottom (1.75rem) margins to.gridcontainers and normalized.pillar-cardchild margins to ensure comfortable breathing room between card grids and succeeding body paragraphs across all pages.
- Added global top (
- Dynamic Sidebar Title Resolution (
_quarto.yml, all documentation files):- Configured section landing pages (
overview.qmd) to explicitly label as"Overview"in the sidebar navigation while maintaining full descriptive document titles inside each file’s frontmatter. - Removed hardcoded
text:attributes from all article links in_quarto.ymlsidebar navigation so they dynamically bind to each document’s YAMLtitle:. - Added clean YAML frontmatter
title:metadata across all.mdand.qmddocumentation files. - Editing any document title updates the sidebar navigation seamlessly without requiring edits to
_quarto.yml.
- Configured section landing pages (
- Conceptual Model Theoretical Deepening & Dual-Role Visual Workflow (
grammar/conceptual_model.qmd):- Formalized the mathematical state equation \(S_n = f(S_{n-1} \mid \text{STAGING})\) to define \(S_n\) as a beat-state trajectory across duration \(\Delta t\), with \(\text{STAGING}\) operating as the global invariant conditioning baseline.
- Clarified the intra-line Sub-State chaining operator (
->) as establishing multi-modal synchronization and causal priority (framing \(\to\) action \(\to\) dialogue \(\to\) audio) for model visual planning. - Added a responsive Quarto visual component (
::::: {.grid}with.workflow-pipelinebanners) illustrating the dual-loop workflow: Pre-Visualization Scaffold (forward mental staging) and Diagnostic Post-Mortem (root-cause debugger for output divergence vs. stochastic AI variance).
- Conceptual Model Quarto Migration & Practitioner Voice (
grammar/conceptual_model.qmd):- Migrated
grammar/conceptual_model.mdtogrammar/conceptual_model.qmdwith native Quarto cross-references (@sec-staging-equation,@sec-three-pillars,@sec-structural-units,@sec-directorial-laws). - Rewrote the narrative flow in Taruma’s authentic practitioner voice (matching
index.qmd), opening with the cognitive pre-visualization challenge and embedding the interactive.workflow-pipelinebanner and 2-column feature grid (::::: {.grid}). - Extracted the 25-row glossary dictionary out of the chapter body into the dedicated
grammar/glossary.qmdhub.
- Migrated
- Layer 1 Pure Theory Alignment & The Reasoning Bridge (
grammar/):- Aligned all core theory documents in
grammar/withFRAMEWORK_INTENT.mdand verified architectural ground truth. - Formalized the Reasoning Bridge role of
[LOGIC]inconceptual_model.qmdandblock_rules.md: guides model visual planning for brief execution beats and primes upstream LLM token prediction, reducing latent variance while degrading gracefully if omitted. - Re-anchored
[OPENING]strictly under the AESTHETIC Pillar in Phase 1: STAGING as the static first-frame canvas (\(S_0\)) with zero kinetic progression. - Calibrated the Script Supervisor Audit check (
script_supervisor_audit.md) for[INTENT]to treat the 1-sentence formula as the recommended anti-bloat baseline rather than a rigid disqualifier. - Validated the 4 Narrative Scene Topologies, 4 Script Density Profiles, and the 100% strict Single-Dimension Law.
- Aligned all core theory documents in
- Website Navigation Streamlining & Direct Scene Terminology (
_quarto.yml,scripts/,README.md):- Streamlined the global Quarto website sidebar into 5 punchy sections:
Theory & Grammar,Macro-Blocks,Sub-State Tags,Modules, andReference Scripts. - Removed all numerical prefixes (
01.,02., etc.) and square brackets from sidebar items, using clean em-dash styling (CAM — Optics & Moves,INTENT — Global Objective) to eliminate unnecessary multi-line text wrapping. - Eliminated “Exhibit” terminology across all script documents, overview hubs, and routers in favor of direct, descriptive scene titles (
Baseline Cinema (Horror POV),Seedance Dialogue (Laundromat),Video Continuation (Auditorium),Token Mapping (Comedy)).
- Streamlined the global Quarto website sidebar into 5 punchy sections:
- 6-Tier Layer Architecture & Strict Downstream Isolation Governance (
AGENTS.md):- Codified the formal 6-Tier Layer Responsibility Matrix (Layer 0 to Layer 5 + Root Router) directly into
AGENTS.md. - Enforced the Single Source of Truth and Strict Downstream Isolation invariant to eliminate duplicate maintenance across files.
- Added the Change Cascade Matrix defining exactly which files must be touched—and which must remain untouched—when modifying any documentation layer.
- Codified the formal 6-Tier Layer Responsibility Matrix (Layer 0 to Layer 5 + Root Router) directly into
- Lean Specification Router & GitHub Landing Page Refactor (
README.md):- Refactored
README.mdinto a lean, pure-prose specification router and GitHub landing page, eliminating duplicate LaTeX state math, redundant invariant lists, and ASCII boxes to enforce Strict Downstream Isolation. - Established a single source of truth where pure theory lives in
grammar/and internal governance lives inFRAMEWORK_INTENT.md, preventing cross-document drift when updating core concepts.
- Refactored
- Prefix-Free Semantic File Naming & Directory Restructuring (
grammar/,modules/,scripts/):- Standardized the entire repository on 100% prefix-free semantic file naming:
grammar/(conceptual_model.md,block_rules.md,execution_director.md,tag_taxonomy.md,directorial_workflows.md,script_supervisor_audit.md).grammar/macroblocks/(intent.md,logic.md,aesthetic.md,opening.md,execution.md).grammar/substates/(substate_grammar.md,tag_*.md).modules/(character_reference.md,seedance25_syntax.md,token_mapping.md,video_continuation.md).scripts/(rules_of_thumb.md,baseline_cinema_script.md,seedance25_dialogue_script.md,video_continuation_script.md,token_mapping_script.md).
- Streamlined Quarto navbar titles to compact labels (Grammar, Blocks, Tags, Modules, Scripts).
- Synchronized all navigation trees in
_quarto.yml, website landing page (index.qmd), master router (README.md), governance protocols (AGENTS.md,FRAMEWORK_INTENT.md), and intra-document relative links across all files.
- Standardized the entire repository on 100% prefix-free semantic file naming:
- Homepage Specification Landing Page Refactor (
index.qmd,custom.css):- Refactored
index.qmdinto a focused, noise-free technical standard landing page and specification router for AI filmmakers, creative technologists, and pipeline developers. - Converted the multi-column card grid in the Knowledge Base Modular Explorer into a Curated Narrative Roadmap divided into two compact thematic layers (1. Foundational Theory & Directorial Grammar and 2. Production Modules & Reference Exhibits) with Core Triad focus (
[CAM],[ACT],[AUDIO]) and an active development callout.
- Refactored
- Architectural Realignment: “Production Modules” Taxonomy:
- Replaced misleading “Platform Extensions” and “Platform Adapters” terminology across navigation manifests (
_quarto.yml), home roadmap (index.qmd), master router (README.md), intent matrix (FRAMEWORK_INTENT.md), and agent governance (AGENTS.md) with Production Modules. - Formally categorized the operational layer into practical filmmaking modules (Multimodal Reference Binding
@actor, Context Token DictionariesLOC01, Multi-Clip Continuous Pipelines@video1, and Native Syntax Adapters) rather than narrow platform-specific hacks. - Replaced the redundant reading pathways tabset with a streamlined Where to Start closing section linking directly to the Conceptual Model and Reference Examples Gallery.
- Added an early experimental development disclaimer, skill-to-knowledge-base migration note, and community connection callout linking to Taruma on X (@tarumainfo) for real-time updates, questions, and insights.
- Replaced clunky LaTeX display equations with a responsive, modern Workflow Pipeline Banner component (
.workflow-pipeline). - Added seamless before and after narrative paragraphs to every visual card, illustration, and video exhibit.
- Replaced misleading “Platform Extensions” and “Platform Adapters” terminology across navigation manifests (
- Homepage Production Exhibit Showcase (
index.qmd,custom.css):- Added three live production exhibits: Prehistoric Survival Chase (Seedance 2.5), Continuous POV War Survival (Seedance 2.5), and Netherlands Vacation Montage (Seedance 2.0) generated with Dreamina.
- Integrated native Quarto
{{< video >}}playback, floating top-left model indicator badges (Seedance 2.0/Seedance 2.5), top-right Dreamina platform link overlays, clean centered scene captions, and collapsible interactive.auteurscript canvas blocks with visual/plain toggle capabilities. - Added responsive viewport rules (
@media (max-width: 576px)) incustom.cssensuring glassmorphic overlays remain compact and never collide on mobile devices.
- Human-Centric Pre-Visualization & Platform-Agnostic Refinement:
- Formalized Auteur Script primarily as a human-first writing, cognitive staging, and pre-visualization tool for the director to mentally simulate and pace the scene before clicking generate.
- Enforced strict platform and model agnosticism across
grammar/, removing proprietary model enumerations from core definitions. - Articulated Macro-State lines (\(S_n\)) as Human Pre-Visualization Units for cognitive reading and pacing.
- Reframed foundational directorial laws (“Boundaries, Not Handcuffs”, “Literal Over Literary”, “Single-Dimension Law”) as classical directorial craft principles.
- Positioned the Script Supervisor Audit Engine as the director’s mental pre-flight checklist.
- Transition to Standalone Knowledge Base & Open Specification:
- Transformed the repository from an AI agent-specific “skill” into an authoritative, platform-agnostic Modular Knowledge Base & Open Cinematic Specification (v0.1.0).
- Replaced
SKILL.mdwithREADME.mdas the master overview, architecture index, and document router. - Refactored interaction behavior into formal Directorial Workflows, Script Density Profiles (Full/Concise/Tagless/Audit), and Pipeline Integration Standards (
grammar/05_directorial_workflows.md).