You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This is an append-only chronological ledger tracking major engineering steps, research document ingestions, compiler features, and visual architecture revisions.
Action: Fully implemented four new advanced compilation gates (IF Gate, WEIGHT Node, TEMPLATE Node, VALIDATE Node) inside the Visual IDE canvas and Semantic Compiler, addressed 15 active functional bugs (BUG-1 to BUG-15) and 9 documentation specifications (DOC-1 to DOC-9).
Details:
Advanced Gates Canvas Integration: Built customized, high-fidelity visual React Custom Node components (IFNode, WEIGHTNode, TEMPLATENode, VALIDATENode) in CustomNodes.jsx and registered them in nodeTypes.
Dynamic Compiler Evaluation: Expanded the topological compiler evaluation pipeline in semanticCompiler.js to compile and dynamically evaluate conditional routing branch streams (ifGate then/else Handles), inline weight emphasis formatting (weightGate), style template formatting boilers (templateGate), and automated lexical conflict diagnosis checklists (validateGate).
App Shell State Synchronizer: Wired up sidebar Advanced Gates draggable palette buttons, reactive node properties binds (onChangeWeight, onChangeTemplate), edge-cleaning port decrement filters (BUG-12), non-hanging FileReader settle handlers (BUG-15), and synchronized compile-time validation reports in App.jsx.
Visual Port Compatibility: Upgraded isValidConnection and edge color drawing in PLGCanvas.jsx to treat the conditional then and else source handles as fully file-compatible.
Obsidian Wiki Ledger Sync: Completely refreshed the local wiki ledger (wiki/) under Dataview HSL frontmatter, documented Memory Bank and Advanced Gates layouts, and verified root raw directories.
Action: Conceived, styled, and built the complete Multi-Domain Compiler Priority Selector and Priority Manager dynamic sorting engine, implementing dynamic schema priority mapping, dynamic lexical/AI domain classification, dynamic deep visual briefs, and dynamic offline clarification questions catalog.
Details:
Dynamic Dropdown UI: Styled and rendered a premium visual <select> control (.priority-select) in the right sidebar Inspector above the Compilation Depth control, featuring dynamic color-changing visual indicator accents matching the active domain.
Auto-Detect Classifier: Implemented the dual-pathway classifier classifyDomain(text, settings) in semanticCompiler.js. It performs fast lexical keyword analysis in rule-based mode, and structured LLM API JSON classification queries in AI mode to dynamically determine the compilation profile.
Multi-Domain Priority Weights: Integrated the five core domains (image, code, debug, architecture, gui) and their 9-dimensional priority weights and keyword dictionaries from the Obsidian specification sheet.
Generic Dynamic Templates: Refactored the offline thinking paragraph generator (offlineThinkingRephrase) and the deep-thinking Visual Brief generator (offlineDeepThinkingSpec) to dynamically adapt and group prompt details under domain-specific categories.
Domain-Specific Clarification Questions: Configured the offline rule-based Ask AI Questions catalog OFFLINE_QUESTIONS to fetch targeted questions based on the active domain (such as language dependencies for code, stack traces for bugs).
Build Integrity Verified: Compiled the production bundle successfully using npm run build with zero Rollup or linting errors.
Action: Conceived and specified a multi-domain dynamic sorting architecture for the semantic compiler, introducing specialized prompt priority dimensions and offline template question banks for Programming, Debugging, System Architecture, and GUI layout tasks.
Details:
Dynamic Multi-Domain Schema: Researched and established 9-dimensional priority weights and core keyword mappings for five key prompt domains (image, code, debug, architecture, gui).
Priority Manager Specification: Designed the system architecture for a dynamic PriorityManager routing engine capable of switching priority catalogs on the fly.
Dual-Pathway Classifier: Defined a hybrid classification pipeline which performs fast offline rule-based keyword matching when executing in Normal mode, and switches to a structured LLM system classifier in Thinking or DeepThinking AI modes.
Domain-Specific Offline Question Banks: Designed 5 distinct offline question banks (10 targeted template questions each) for each compilation domain, ensuring that rule-based Ask AI Questions gates serve highly tailored queries.
Wiki & Docs Integration: Formulated wiki/priority_manager.md and updated compiler-engine.md to integrate all dynamic priorities, hybrid routing architectures, and question catalogs.
Action: Conceived and created a complete developer documentation directory containing structured markdown files covering quickstart setups, Vanilla CSS and state rules, and copy-pasteable AI prompts to automate logs, wikis, and indices.
Details:
Decoupled Folder Structure: Created a dedicated documentation/development/ directory containing three rich markdown files:
setup.md: Complete setup guide mapping prerequisites (Node.js v18+), package installs (npm install), Vite dev server setups (npm run dev), and compiled production bundles (npm run build).
standards.md: Outlines pure Vanilla CSS styling variables (no Tailwind), React Flow parent-state callbacks, topological DFS sorting, and static conflict override matrices.
llm-automation.md: Explains ingest, refine, lint, and design protocols from AGENTS.md, and provides a copy-pasteable AI system prompt template to allow LLMs to auto-compile logs and index links.
Catalogs Updated: Integrated references to the new markdown suite in wiki/index.md and the root repository README.md to assist new contributors and LLM agents.
Action: Conceived and integrated an editable Workspace Name control directly in the IDE Topbar, creating an automated two-way synchronization loop that governs all downloaded .json graph configurations and exported .txt prompt briefs, with automated space-to-underscore sanitization.
Details:
Topbar Workspace Control: Designed and styled a premium, glassmorphic .workspace-name-container component in the App.jsx Topbar. Equipped it with hover animations, reactive focus-glowing cyan borders, and integrated file extension indicators (.json / .txt).
Two-Way Binding Engine: Implemented an automated synchronization system in App.jsx. Changing the Workspace input automatically propagates the name down to rename any active fileNode filenames on the canvas; conversely, editing the filename on the canvas dynamically updates the Workspace input in the Topbar.
Automated Space-to-Underscore Substitution: Configured real-time input sanitizers across all entry streams—including Topbar workspace input, File Node filename change handlers, and nodes sync hooks—to automatically convert space () characters into underscores (_) to maintain strict operating system and filesystem filename safety.
Persistent LocalStorage Sync: Updated the mount-restore and auto-save useEffect hooks in App.jsx to preserve workspaceName under plg_last_project, ensuring it restores seamlessly on page reloads.
Automated Filename Saves: Refactored handleSaveGraph and handleExportTxt in App.jsx to compile filenames dynamically as ${workspaceName}.json and ${workspaceName}.txt respectively, addressing the user's need for consistent, workspace-aligned file naming.
Vite & CSS Compilation Verified: Rectified a minor unbalanced CSS bracket in index.css and compiled the production build using Vite to guarantee zero warning alerts.
[2026-05-30] refine | Implement Prompt Box Rephrase Undo (Ctrl+Z) History System
Action: Conceived and built a custom keystroke-level undo and checkpointing history mechanism for Prompt Box AI rephrasing, enabling users to press Ctrl+Z to revert programmatically rephrased prompts back to their raw typed baseline.
Details:
Undo State Capture: Configured the onRephrasePrompt callback in App.jsx to capture and store the current, human-written text in the node's state under data.userText before launching the AI optimizer.
Keystroke Listener Interceptor: Attached a handleKeyDown event listener to the <textarea> component inside the PromptBoxNode custom component in CustomNodes.jsx. It catches Ctrl+Z (or Cmd+Z), checks if there is a recorded userText distinct from the current rephrased text, and dynamically restores the user's pre-rephrased baseline.
Cascading Rephrase Memory: Enabled continuous multi-rephrasing history. If the user manually modifies an AI-rephrased text and triggers rephrasing again, their manual edits automatically overwrite userText as the new history checkpoint.
Tooltips Updated: Added tooltip titles explicitly calling out the Ctrl+Z undo functionality on Prompt Box rephrase controls.
[2026-05-30] refine | Add AI Rephrase Button to Prompt Box Custom Node
Action: Conceived and integrated an AI Rephrase Fragment button directly into the PromptBoxNode custom component to automatically optimize raw user inputs, correct typos, grammar issues, and enforce a commanding visual tone.
Details:
Rephrase Handler Bonded: Created an asynchronous callback onRephrasePrompt in App.jsx bound inside the bindNodeCallbacks wrapper.
AI Rephrasing Prompt Formulated: Formulated a targeted system prompt inside onRephrasePrompt directing the LLM to scan prompt fragments, fix structural typos or spelling errors, and output a concise, commanding rephrased fragment without meta-text.
Offline Rephraser Fallback: Implemented an offline fallback that cleans double spaces, handles first-letter capitalization, and outputs the sanitized baseline text if AI calls fail.
UI/UX Node Component Upgrades: Modified PromptBoxNode in CustomNodes.jsx to display a glassmorphic button equipped with a Brain icon and responsive spinner animation. Adjusted target handle vertically to top: "50%" to center dynamically.
Documentation & Wiki Updated: Updated the nodes manual (nodes-guide.md) and the wiki guide (nodes.md) to reflect Prompt Box rephrasing controls.
[2026-05-30] refine | Redesign Compiled Prompts for Strict Commanding and Directive Output
Action: Completely overhauled the semantic prompt compilers (both AI and offline modes) to ensure all generated outputs are strictly directive visual briefs or commanding sentences, eliminating passive phrasings, parenthesized technical metadata, and document references.
Details:
AI Prompt Architectures Redesigned: Refactored system prompts for AND (aiGateAnd), OR (aiGateOr), NOT (aiGateNot), Q&A Synthesis (aiRefineAnswers), and Memory Alignment (aiAlignPromptWithMemory) to speak in a visual director's command voice, strictly banning phrases like "according to memory", "based on fragment A", "merged prompt", or meta-labels.
Deep & Normal Thinking Prompts Refined: Modified thinking and deep-thinking mode compiler prompts to strictly command the downstream generator model via active, imperative verbs and structure Premium Markdown briefs without meta-commentary or document references.
Offline Memory Appends Redesigned: Dismantled the parenthesized technical metadata formula (Memory Constraints: [[Term] rule: [Description]]) in both primary and error fallback compilation pathways. Replaced it with a direct, active sentence-injection format: Enforce [Term]: [Description]. which blends seamlessly into the final prompt.
Documentation & Wiki Manuals Synced: Updated the compiler engine spec (documentation/compiler-engine.md), workflows handbook (documentation/workflows-and-tutorials.md), and memory wiki (wiki/memory.md) to reflect the new commanding prompt syntax and expected outputs.
[2026-05-30] lint | Reconcile 10-page user manual suite, align wiki hubs, and synchronize with Git
Action: Updated the entire documentation/ user manual suite and aligned relative link hubs to reconcile all developer documentation with fuchsia Context Memory nodes and singleCompiledPrompt explicit inline negation architecture.
Details:
10-Page Manual Suite Synced: Overwrote README, Getting Started, How It Works, Nodes Guide, Compiler Engine, UI Reference, Workflows, Project Structure, and Design System.
Context Memory API Documented: Detailed handles (.memory), drag-and-drop file uploaders, lexical casing swappers, zero-shot AI catalogers, and the parenthesized rule constraint injection formula.
Unified Compiled Prompt Specs: Completely excised obsolete references to a separate positive/negative dual-pipeline; updated all tutorials and UI references to feature a single Compiled Prompt showing inline negation constraints.
Star-Graph Links Integrated: Added relative connection grids at the end of wiki/overview.md to form a perfect Obsidian star-graph representation.
[2026-05-30] refine | Upgrade Visual Context Memory system to MemPalace-style ledger & complete prompt expansion
Action: Upgraded the Visual Context Memory system to model a high-fidelity indexing catalog (styled after MemPalace) and enabled complete prompt design/expansion based on matching memory specifications.
Details:
Memory Indexer Upgraded: Upgraded onExtractMemory in App.jsx to index files into a rigorous HSL-friendly hierarchy containing Wings, Rooms, and Halls.
Offline Lexical Crawler: Implemented an advanced line-by-line regex crawler that parses JSON keys, multi-casing variables (camelCase, PascalCase, snake_case, CONSTANT_CASE), function signatures, and semantic guidelines (using must, should, always, never, avoid regex matching) to construct detailed visual entity hints and constraints.
AI Prompt Indexer Redesigned: Redesigned the AI system extractor instructions to output Obsidian-friendly workspace hierarchy logs, variable definition tables, and semantic visual constraints.
Complete Prompt Design & Expansion: Upgraded the Context Memory compiler gate (aiAlignPromptWithMemory and offline fallbacks in semanticCompiler.js) to intercept high-level prompt keywords or references, automatically expanding them into detailed specifications utilizing exact case-sensitive nomenclature.
Offline Semantic Rules Injection: Implemented automatic extraction of memory rule statements which are appended directly as formatted parenthesized constraints on rule-based compiler cycles when terms are matched.
Wiki Verification Guide: Overwrote Context Memory Reference Specifications to document the Wings/Rooms/Halls workspace structure, expansion engines, and validation card.
Action: Conceived and built the Context Memory Node which reads codebase files, wikis, html, and json, indexing variables/functions into structured Markdown to enforce exact casing alignment during prompt compilation.
Details:
Memory Node Creation: Implemented ContextMemoryNode in CustomNodes.jsx utilizing a fuchsia-purple theme (var(--memory)). Equipped node with client-side text file loaders, files lists, a text editor preview, and a target file pin + source memory pin.
Topological Sorting Updated: Upgraded topological sorter (buildExecutionOrder in semanticCompiler.js) to resolve memory dependencies upstream, ensuring file baselines are fully generated before memory indexes them.
Visual Handles & Routing: Added purple target memory handles to AND, OR, NOT, and Provide Answers nodes in CustomNodes.jsx with custom validation and purple visual connections (#c084fc) in PLGCanvas.jsx.
Compiler Alignment: Upgraded prompt generators in semanticCompiler.js to swallow memory variables/functions and feed them as strict system constraints to the AI, ensuring exact naming matches.
Offline Memory Verification: Formulated an offline regex matching validator within compileGraph that scans generated prompts for backtick-wrapped memory keywords and outputs exact validation reports in the Pipeline Debugger stages.
[2026-05-30] refine | Update NOT Gate to Explicit Negation & Remove Negative Prompt System
Action: Converted NOT gate concept suppression to append explicit negation clauses (e.g. avoid [concept]) directly to the single Compiled Prompt baseline and fully dismantled the separate Negative Prompt system.
Details:
Core Compiler Refactoring: Updated semanticCompiler.js to strip positive references to A and append avoid A in rule-based mode, and modified aiGateNot instructions to ask the LLM to write natural clauses (like "do not use", "avoid") in one prompt.
UI/UX Cleanup: Completely removed Section 2 ("Negative Prompt") and its copy button from Inspector.jsx and the PLG-IDE.html right sidebar. Renamed "Positive Prompt" to "Compiled Prompt".
Custom Nodes Refined: Updated FileViewerNode in CustomNodes.jsx to hide the Negative Prompt box and expanded the main Compiled Prompt text area for a premium visual aesthetic. Updated NOTNode description.
Export System: Synchronized App.jsx and PLG-IDE.html plain text exporters to output a clean, single prompt under a # Compiled Prompt header.
Compiling Wiki Guides: Updated all Obsidian-compatible guides including Overview, Architecture, Nodes Reference, Compiler Specifications, and Workflows to strictly detail explicit negation routing instead of negative memory.
Action: Added 3 prompt compilation depth options (Normal, Thinking, DeepThinking) to the visual IDE's Inspector panel and Semantic Compiler.
Details:
Premium UI & Segmented Control: Implemented a glassmorphic segment selector inside Inspector.jsx utilizing Lucide icons (Zap, Brain, Cpu) with responsive active HSL colors and glowing focus rings in index.css.
State & LocalStorage Sync: Synchronized compiler modes in App.jsx to persist in local storage, automatically restoring preferences on reload and providing premium toast feedback.
Thinking Mode Pipeline: Expanded the compiler in semanticCompiler.js to rephrase prompts beautifully via LLM, and implemented a high-vocabulary offline natural paragraph synthesizer that groups items by semantic category (Subject, Environment, camera framing, etc.) as a fallback.
DeepThinking Mode Pipeline: Formulated a compiler engine that outputs an exhaustive visual spec document in Markdown layout (covering Subject Focus, Spatial Context, Lighting/Composition, and Suppression Rules) via LLM, alongside a structured offline specs document fallback.
Action: Completed rule-based semantic parser and multi-provider AI api caller.
Details:
Formulated the 8 core categorization prompt dictionaries (Subject, Environment, Action, Emotion, Lighting, Style, Camera, Effects) with distinct compilation priority weights.
Structured baseline conflict matrix resolving style incompatibilities (e.g. realistic vs cartoon) via penalty deductions.
Established topological sorting sequence building functional DAG execution paths (buildExecutionOrder) to navigate file and prompt-converter boundaries.
Connected Anthropic Claude, OpenAI ChatGPT, and Google Gemini API callers using local CORS gateway proxies.
[2026-05-28] design | Core Visual Logical Circuit Design Plan
Action: Conceived visual logical operators mapping to classic boolean logic:
AND: Intersection compilation merging context details by sorting semantic importance.
OR: Context selection routing the highest semantically compatible option.
NOT: Negative suppression routing a concept to negative memories and stripping it from positive files.