2026-02-13 17:09:46 +00:00
<!DOCTYPE html>
< html lang = "en" >
< head >
< meta charset = "UTF-8" >
< meta name = "viewport" content = "width=device-width, initial-scale=1.0" >
< title > Knowledge Refinery — Presentation< / title >
< style >
/* ============================================================
RESET & BASE
============================================================ */
*, *::before, *::after { box-sizing: border-box; margin: 0; padding: 0; }
:root {
--bg: #ffffff;
--bg-subtle: #f8f9fc;
--bg-muted: #f0f2f7;
--bg-card: #ffffff;
--fg: #1a1d26;
--fg-muted: #5a6072;
--fg-subtle: #8891a5;
--accent: #4f46e5;
--accent-2: #7c3aed;
--accent-3: #2563eb;
--accent-4: #0891b2;
--accent-5: #059669;
--accent-bg: #eef2ff;
--gradient: linear-gradient(135deg, #4f46e5, #7c3aed, #2563eb);
--border: #e5e7eb;
--shadow: 0 1px 3px rgba(0,0,0,0.06), 0 1px 2px rgba(0,0,0,0.04);
--shadow-lg: 0 10px 40px rgba(0,0,0,0.08);
--radius: 12px;
--radius-sm: 8px;
--radius-xs: 6px;
--font: -apple-system, BlinkMacSystemFont, "SF Pro Display", "Segoe UI", system-ui, sans-serif;
--mono: "SF Mono", "Fira Code", "JetBrains Mono", ui-monospace, monospace;
--slide-w: 100vw;
--slide-h: 100vh;
}
html { font-size: 18px; scroll-behavior: smooth; }
body {
font-family: var(--font);
color: var(--fg);
background: var(--bg);
line-height: 1.6;
-webkit-font-smoothing: antialiased;
overflow: hidden;
}
/* ============================================================
SLIDE ENGINE
============================================================ */
.deck {
width: var(--slide-w);
height: var(--slide-h);
position: relative;
overflow: hidden;
}
.slide {
position: absolute;
top: 0; left: 0;
width: 100%; height: 100%;
display: flex;
flex-direction: column;
justify-content: center;
align-items: center;
padding: 4rem 6rem;
opacity: 0;
transform: translateX(60px);
transition: opacity 0.5s ease, transform 0.5s ease;
pointer-events: none;
overflow-y: auto;
}
.slide.active {
opacity: 1;
transform: translateX(0);
pointer-events: auto;
}
.slide.prev {
opacity: 0;
transform: translateX(-60px);
}
/* ============================================================
NAVIGATION
============================================================ */
.nav-bar {
position: fixed;
bottom: 0;
left: 0; right: 0;
height: 56px;
background: var(--bg);
border-top: 1px solid var(--border);
display: flex;
align-items: center;
justify-content: space-between;
padding: 0 2rem;
z-index: 100;
backdrop-filter: blur(12px);
-webkit-backdrop-filter: blur(12px);
}
.nav-bar .progress {
flex: 1;
height: 4px;
background: var(--bg-muted);
border-radius: 2px;
margin: 0 1.5rem;
overflow: hidden;
}
.nav-bar .progress-fill {
height: 100%;
background: var(--gradient);
border-radius: 2px;
transition: width 0.4s ease;
}
.nav-btn {
display: inline-flex;
align-items: center;
gap: 0.4rem;
padding: 0.5rem 1rem;
border: none;
border-radius: var(--radius-xs);
background: var(--bg-muted);
color: var(--fg);
font: inherit;
font-size: 0.85rem;
font-weight: 500;
cursor: pointer;
transition: background 0.2s, transform 0.1s;
user-select: none;
-webkit-user-select: none;
}
.nav-btn:hover { background: var(--border); }
.nav-btn:active { transform: scale(0.97); }
.nav-btn:disabled { opacity: 0.3; cursor: default; }
.slide-counter {
font-size: 0.8rem;
color: var(--fg-subtle);
font-variant-numeric: tabular-nums;
min-width: 4rem;
text-align: center;
}
/* ============================================================
TYPOGRAPHY
============================================================ */
.slide-inner { max-width: 1100px; width: 100%; }
h1 {
font-size: 3rem;
font-weight: 800;
letter-spacing: -0.03em;
line-height: 1.15;
margin-bottom: 1rem;
}
h1.gradient {
background: var(--gradient);
-webkit-background-clip: text;
-webkit-text-fill-color: transparent;
background-clip: text;
}
h2 {
font-size: 2.2rem;
font-weight: 700;
letter-spacing: -0.02em;
line-height: 1.2;
margin-bottom: 1.5rem;
}
h3 {
font-size: 1.3rem;
font-weight: 600;
margin-bottom: 0.75rem;
}
.subtitle {
font-size: 1.3rem;
color: var(--fg-muted);
font-weight: 400;
line-height: 1.5;
max-width: 700px;
}
.lead {
font-size: 1.15rem;
color: var(--fg-muted);
line-height: 1.7;
margin-bottom: 2rem;
}
/* ============================================================
COMPONENTS
============================================================ */
.tag {
display: inline-block;
padding: 0.2rem 0.7rem;
border-radius: 999px;
font-size: 0.75rem;
font-weight: 600;
letter-spacing: 0.03em;
text-transform: uppercase;
}
.tag-accent { background: var(--accent-bg); color: var(--accent); }
.tag-green { background: #ecfdf5; color: #059669; }
.tag-amber { background: #fffbeb; color: #d97706; }
.tag-cyan { background: #ecfeff; color: #0891b2; }
.tag-rose { background: #fff1f2; color: #e11d48; }
.badge {
display: inline-flex;
align-items: center;
gap: 0.35rem;
padding: 0.35rem 0.8rem;
border-radius: var(--radius-xs);
font-size: 0.8rem;
font-weight: 500;
}
/* Card grid */
.card-grid {
display: grid;
gap: 1.25rem;
width: 100%;
}
.card-grid.cols-2 { grid-template-columns: 1fr 1fr; }
.card-grid.cols-3 { grid-template-columns: 1fr 1fr 1fr; }
.card-grid.cols-4 { grid-template-columns: 1fr 1fr 1fr 1fr; }
.card {
background: var(--bg-card);
border: 1px solid var(--border);
border-radius: var(--radius);
padding: 1.5rem;
box-shadow: var(--shadow);
transition: box-shadow 0.25s, transform 0.25s;
}
.card:hover {
box-shadow: var(--shadow-lg);
transform: translateY(-2px);
}
.card .icon {
width: 44px;
height: 44px;
border-radius: 10px;
display: flex;
align-items: center;
justify-content: center;
font-size: 1.3rem;
margin-bottom: 0.75rem;
}
.card .icon.indigo { background: #eef2ff; color: #4f46e5; }
.card .icon.violet { background: #f5f3ff; color: #7c3aed; }
.card .icon.blue { background: #eff6ff; color: #2563eb; }
.card .icon.cyan { background: #ecfeff; color: #0891b2; }
.card .icon.emerald { background: #ecfdf5; color: #059669; }
.card .icon.amber { background: #fffbeb; color: #d97706; }
.card .icon.rose { background: #fff1f2; color: #e11d48; }
.card .icon.slate { background: #f1f5f9; color: #475569; }
.card p { font-size: 0.9rem; color: var(--fg-muted); line-height: 1.55; }
/* Code block */
.code-block {
background: #1e1e2e;
color: #cdd6f4;
border-radius: var(--radius);
padding: 1.25rem 1.5rem;
font-family: var(--mono);
font-size: 0.82rem;
line-height: 1.65;
overflow-x: auto;
white-space: pre;
box-shadow: var(--shadow-lg);
}
.code-block .comment { color: #6c7086; }
.code-block .keyword { color: #cba6f7; }
.code-block .string { color: #a6e3a1; }
.code-block .number { color: #fab387; }
.code-block .fn { color: #89b4fa; }
.code-block .var { color: #f5e0dc; }
.code-block .op { color: #89dceb; }
/* Pipeline visual */
.pipeline {
display: flex;
align-items: center;
gap: 0;
width: 100%;
margin: 2rem 0;
}
.pipeline-stage {
flex: 1;
text-align: center;
padding: 1rem 0.5rem;
position: relative;
}
.pipeline-stage .num {
width: 36px;
height: 36px;
border-radius: 50%;
display: inline-flex;
align-items: center;
justify-content: center;
font-weight: 700;
font-size: 0.85rem;
margin-bottom: 0.4rem;
color: white;
}
.pipeline-stage .label {
display: block;
font-weight: 600;
font-size: 0.85rem;
margin-bottom: 0.2rem;
}
.pipeline-stage .desc {
font-size: 0.72rem;
color: var(--fg-subtle);
line-height: 1.4;
}
.pipeline-arrow {
width: 32px;
flex-shrink: 0;
text-align: center;
color: var(--fg-subtle);
font-size: 1.2rem;
}
/* Architecture diagram */
.arch-diagram {
display: grid;
grid-template-columns: 1fr auto 1fr;
gap: 1.5rem;
align-items: stretch;
width: 100%;
margin: 1.5rem 0;
}
.arch-box {
border: 2px solid var(--border);
border-radius: var(--radius);
padding: 1.25rem;
position: relative;
}
.arch-box h4 {
font-size: 0.8rem;
text-transform: uppercase;
letter-spacing: 0.08em;
color: var(--fg-subtle);
margin-bottom: 0.75rem;
}
.arch-connector {
display: flex;
flex-direction: column;
align-items: center;
justify-content: center;
color: var(--fg-subtle);
font-size: 1.5rem;
gap: 0.25rem;
}
.arch-connector .proto {
font-size: 0.65rem;
font-weight: 600;
color: var(--accent);
letter-spacing: 0.05em;
text-transform: uppercase;
}
/* Step list */
.step-list {
list-style: none;
counter-reset: steps;
}
.step-list li {
counter-increment: steps;
padding: 1rem 1rem 1rem 3.5rem;
position: relative;
border-left: 2px solid var(--border);
margin-left: 1rem;
}
.step-list li:last-child { border-left-color: transparent; }
.step-list li::before {
content: counter(steps);
position: absolute;
left: -14px;
top: 1rem;
width: 28px;
height: 28px;
border-radius: 50%;
background: var(--accent);
color: white;
font-weight: 700;
font-size: 0.8rem;
display: flex;
align-items: center;
justify-content: center;
}
.step-list li strong {
display: block;
margin-bottom: 0.25rem;
}
.step-list li p {
font-size: 0.9rem;
color: var(--fg-muted);
}
/* Table */
.data-table {
width: 100%;
border-collapse: collapse;
font-size: 0.85rem;
}
.data-table th {
text-align: left;
padding: 0.6rem 1rem;
border-bottom: 2px solid var(--border);
color: var(--fg-subtle);
font-weight: 600;
font-size: 0.75rem;
text-transform: uppercase;
letter-spacing: 0.05em;
}
.data-table td {
padding: 0.6rem 1rem;
border-bottom: 1px solid var(--bg-muted);
vertical-align: top;
}
.data-table tr:last-child td { border-bottom: none; }
.data-table code {
font-family: var(--mono);
font-size: 0.8rem;
background: var(--bg-muted);
padding: 0.1rem 0.4rem;
border-radius: 4px;
}
/* Title slide decoration */
.title-slide {
position: relative;
background: var(--bg-subtle);
}
.title-slide::before {
content: "";
position: absolute;
top: -200px; right: -200px;
width: 600px; height: 600px;
border-radius: 50%;
background: radial-gradient(circle, rgba(79,70,229,0.06) 0%, transparent 70%);
pointer-events: none;
}
.title-slide::after {
content: "";
position: absolute;
bottom: -150px; left: -150px;
width: 500px; height: 500px;
border-radius: 50%;
background: radial-gradient(circle, rgba(124,58,237,0.05) 0%, transparent 70%);
pointer-events: none;
}
/* Feature highlight */
.feature-row {
display: flex;
gap: 2rem;
align-items: flex-start;
margin-bottom: 1.5rem;
}
.feature-row .feature-icon {
width: 48px;
height: 48px;
border-radius: 12px;
display: flex;
align-items: center;
justify-content: center;
font-size: 1.4rem;
flex-shrink: 0;
}
.feature-row .feature-text h3 {
margin-bottom: 0.3rem;
font-size: 1.1rem;
}
.feature-row .feature-text p {
font-size: 0.9rem;
color: var(--fg-muted);
line-height: 1.5;
}
/* Two-col layout */
.two-col {
display: grid;
grid-template-columns: 1fr 1fr;
gap: 2.5rem;
width: 100%;
align-items: start;
}
/* Stats row */
.stats-row {
display: flex;
gap: 2rem;
margin: 1.5rem 0;
}
.stat {
text-align: center;
}
.stat .stat-num {
font-size: 2.5rem;
font-weight: 800;
background: var(--gradient);
-webkit-background-clip: text;
-webkit-text-fill-color: transparent;
background-clip: text;
line-height: 1;
}
.stat .stat-label {
font-size: 0.8rem;
color: var(--fg-subtle);
font-weight: 500;
margin-top: 0.3rem;
}
/* Annotation example */
.annotation-example {
background: var(--bg-subtle);
border: 1px solid var(--border);
border-radius: var(--radius);
padding: 1.5rem;
font-size: 0.85rem;
}
.annotation-example .ann-field {
margin-bottom: 0.6rem;
}
.annotation-example .ann-label {
font-weight: 600;
font-size: 0.75rem;
text-transform: uppercase;
letter-spacing: 0.05em;
color: var(--fg-subtle);
margin-bottom: 0.25rem;
}
.annotation-example .ann-tags {
display: flex;
flex-wrap: wrap;
gap: 0.4rem;
}
.ann-tag {
padding: 0.15rem 0.55rem;
border-radius: 999px;
font-size: 0.75rem;
font-weight: 500;
}
/* Extractor matrix */
.extractor-grid {
display: grid;
grid-template-columns: repeat(4, 1fr);
gap: 1rem;
}
.extractor-item {
background: var(--bg-subtle);
border: 1px solid var(--border);
border-radius: var(--radius-sm);
padding: 1rem;
text-align: center;
}
.extractor-item .ext-icon {
font-size: 1.8rem;
margin-bottom: 0.4rem;
}
.extractor-item .ext-name {
font-weight: 600;
font-size: 0.85rem;
margin-bottom: 0.2rem;
}
.extractor-item .ext-types {
font-size: 0.72rem;
color: var(--fg-subtle);
}
/* Keyboard hint */
.keyboard-hint {
position: fixed;
top: 1rem;
right: 1.5rem;
font-size: 0.72rem;
color: var(--fg-subtle);
z-index: 50;
opacity: 0.6;
}
.key {
display: inline-block;
padding: 0.1rem 0.45rem;
border: 1px solid var(--border);
border-radius: 4px;
font-family: var(--mono);
font-size: 0.7rem;
background: var(--bg-subtle);
box-shadow: 0 1px 0 var(--border);
}
/* Responsive */
@media (max-width: 900px) {
.slide { padding: 2.5rem 2rem; }
h1 { font-size: 2.2rem; }
h2 { font-size: 1.7rem; }
.card-grid.cols-3 { grid-template-columns: 1fr 1fr; }
.card-grid.cols-4 { grid-template-columns: 1fr 1fr; }
.two-col { grid-template-columns: 1fr; }
.pipeline { flex-wrap: wrap; }
.arch-diagram { grid-template-columns: 1fr; }
.extractor-grid { grid-template-columns: 1fr 1fr; }
}
< / style >
< / head >
< body >
< div class = "keyboard-hint" > < span class = "key" > ← < / span > < span class = "key" > → < / span > or swipe to navigate< / div >
< div class = "deck" id = "deck" >
<!-- ============================================================
SLIDE 1: TITLE
============================================================ -->
< div class = "slide title-slide active" data-slide = "0" >
< div class = "slide-inner" style = "text-align:center;" >
< div style = "margin-bottom:1rem;" >
< span class = "tag tag-accent" > Local-First · Evidence-Native · LLM-Powered< / span >
< / div >
< h1 class = "gradient" style = "font-size:3.5rem;" > Knowledge Refinery< / h1 >
< p class = "subtitle" style = "margin:0 auto 2rem;" >
A macOS Tahoe application that ingests heterogeneous document corpora,< br >
extracts structured knowledge via local LLMs, and renders an immersive 3D concept universe.
< / p >
< div class = "stats-row" style = "justify-content:center;" >
< div class = "stat" > < div class = "stat-num" > 7< / div > < div class = "stat-label" > Milestones< / div > < / div >
< div class = "stat" > < div class = "stat-num" > 6< / div > < div class = "stat-label" > Pipeline Stages< / div > < / div >
< div class = "stat" > < div class = "stat-num" > 30< / div > < div class = "stat-label" > Unit Tests< / div > < / div >
< div class = "stat" > < div class = "stat-num" > 0< / div > < div class = "stat-label" > Cloud Calls< / div > < / div >
< / div >
< div style = "margin-top:2rem; color:var(--fg-subtle); font-size:0.85rem;" >
2026-02-13 18:29:23 +00:00
macOS · SwiftUI · Go · chi · LM Studio · SQLite · WebGPU · Multi-Workspace
2026-02-13 17:09:46 +00:00
< / div >
< / div >
< / div >
<!-- ============================================================
SLIDE 2: THE PROBLEM
============================================================ -->
< div class = "slide" data-slide = "1" >
< div class = "slide-inner" >
< span class = "tag tag-amber" style = "margin-bottom:1rem;" > The Problem< / span >
< h2 > Your knowledge is scattered, siloed, and unsearchable< / h2 >
< p class = "lead" >
Research papers, notes, e-books, medical images, code, archives — spread across folders
with no semantic connections. Keyword search fails. Cloud tools leak your data. You need
something that actually < em > understands< / em > your corpus.
< / p >
< div class = "card-grid cols-3" >
< div class = "card" >
< div class = "icon rose" > 🔍 < / div >
< h3 > Search is Broken< / h3 >
< p > Keyword search misses semantically related content. "Machine learning" won't find your "neural network" papers.< / p >
< / div >
< div class = "card" >
< div class = "icon amber" > 🔒 < / div >
< h3 > Privacy Concerns< / h3 >
< p > Cloud-based tools send your sensitive documents, medical records, and proprietary research to third-party servers.< / p >
< / div >
< div class = "card" >
< div class = "icon slate" > 📁 < / div >
< h3 > No Connections< / h3 >
< p > You have thousands of files but no way to see how concepts relate across documents, formats, and domains.< / p >
< / div >
< / div >
< / div >
< / div >
<!-- ============================================================
SLIDE 3: THE SOLUTION
============================================================ -->
< div class = "slide" data-slide = "2" >
< div class = "slide-inner" >
< span class = "tag tag-green" style = "margin-bottom:1rem;" > The Solution< / span >
< h2 > Knowledge Refinery: Your Local Knowledge Engine< / h2 >
< div class = "feature-row" >
< div class = "feature-icon" style = "background:#eef2ff; color:#4f46e5;" > 🧠 < / div >
< div class = "feature-text" >
< h3 > Semantic Understanding via Local LLMs< / h3 >
< p > Uses LM Studio running locally — your data never leaves your machine. Embeddings for similarity search, structured annotation for deep understanding.< / p >
< / div >
< / div >
< div class = "feature-row" >
< div class = "feature-icon" style = "background:#ecfdf5; color:#059669;" > 🔗 < / div >
< div class = "feature-text" >
< h3 > Evidence-Native — Every Insight Links to Source< / h3 >
< p > Every annotation, concept, and search result traces back to an exact location: file, page, chapter, even byte offset within nested archives.< / p >
< / div >
< / div >
< div class = "feature-row" >
< div class = "feature-icon" style = "background:#ecfeff; color:#0891b2;" > 🌌 < / div >
< div class = "feature-text" >
< h3 > 3D Concept Universe< / h3 >
< p > WebGPU-powered visualization renders your knowledge as an interactive 3D graph. Zoom from macro concepts to individual chunks. See how ideas connect.< / p >
< / div >
< / div >
< div class = "feature-row" >
< div class = "feature-icon" style = "background:#fff1f2; color:#e11d48;" > 🛡 < / div >
< div class = "feature-text" >
< h3 > Sandboxed & Incremental< / h3 >
< p > Archive extraction runs in macOS sandbox-exec. Changed files are detected via content hashing — only new/modified files are reprocessed.< / p >
< / div >
< / div >
< / div >
< / div >
<!-- ============================================================
SLIDE 4: ARCHITECTURE
============================================================ -->
< div class = "slide" data-slide = "3" >
< div class = "slide-inner" >
< span class = "tag tag-accent" style = "margin-bottom:1rem;" > Architecture< / span >
< h2 > Three-Tier Local Architecture< / h2 >
< div class = "arch-diagram" >
< div class = "arch-box" style = "border-color:#4f46e5;" >
< h4 style = "color:#4f46e5;" > 📱 SwiftUI App< / h4 >
< div style = "font-size:0.85rem; color:var(--fg-muted);" >
< strong > Search< / strong > — Semantic vector search< br >
< strong > Universe< / strong > — WebGPU 3D visualization< br >
< strong > Concepts< / strong > — Cluster browser + "Why?"< br >
< strong > Ingest< / strong > — Pipeline monitoring< br >
< strong > Volumes< / strong > — Folder management< br >
< strong > Assets< / strong > — File inventory + Quick Look
< / div >
< / div >
< div class = "arch-connector" >
< span > ↔ < / span >
< span class = "proto" > HTTP (localhost)< / span >
< span style = "font-size:0.65rem; color:var(--fg-subtle);" > 127.0.0.1:8742< / span >
< / div >
< div class = "arch-box" style = "border-color:#7c3aed;" >
2026-02-13 18:29:23 +00:00
< h4 style = "color:#7c3aed;" > ⚙ ️ Go Daemon< / h4 >
2026-02-13 17:09:46 +00:00
< div style = "font-size:0.85rem; color:var(--fg-muted);" >
2026-02-13 18:29:23 +00:00
< strong > chi< / strong > router with CORS< br >
2026-02-13 17:09:46 +00:00
< strong > 6-Stage Pipeline< / strong > orchestrator< br >
2026-02-13 18:29:23 +00:00
< strong > SQLite< / strong > metadata + vectors + graph (WAL)< br >
< strong > 11MB< / strong > single binary, zero deps< br >
< strong > 7 Extractors< / strong > + fallback chain
2026-02-13 17:09:46 +00:00
< / div >
< / div >
< / div >
< div style = "text-align:center; margin-top:1rem;" >
< span style = "font-size:1.5rem; color:var(--fg-subtle);" > ↓ < / span >
< span class = "proto" style = "display:block; font-size:0.65rem; font-weight:600; color:var(--accent-4); letter-spacing:0.05em; text-transform:uppercase;" > OpenAI-Compatible API< / span >
< / div >
< div style = "max-width:360px; margin:0.75rem auto 0;" >
< div class = "arch-box" style = "border-color:#059669;" >
< h4 style = "color:#059669;" > 🤖 LM Studio< / h4 >
< div style = "font-size:0.85rem; color:var(--fg-muted);" >
2026-02-13 18:29:23 +00:00
< strong > qwen3-4b-2507< / strong > — Chat / Annotation< br >
2026-02-13 17:09:46 +00:00
< strong > nomic-embed-text-v1.5< / strong > — Embeddings< br >
Running on 127.0.0.1:1234
< / div >
< / div >
< / div >
< / div >
< / div >
<!-- ============================================================
SLIDE 5: THE PIPELINE
============================================================ -->
< div class = "slide" data-slide = "4" >
< div class = "slide-inner" >
< span class = "tag tag-accent" style = "margin-bottom:1rem;" > Core Engine< / span >
< h2 > Six-Stage Ingestion Pipeline< / h2 >
< p class = "lead" style = "margin-bottom:1rem;" > Every file flows through six deterministic stages. The pipeline is incremental — unchanged files are never reprocessed.< / p >
< div class = "pipeline" >
< div class = "pipeline-stage" >
< div class = "num" style = "background:#4f46e5;" > 1< / div >
< span class = "label" > Scan< / span >
< span class = "desc" > Walk directories, compute SHA-256 content hashes, detect new/changed files< / span >
< / div >
< div class = "pipeline-arrow" > → < / div >
< div class = "pipeline-stage" >
< div class = "num" style = "background:#7c3aed;" > 2< / div >
< span class = "label" > Extract< / span >
< span class = "desc" > Produce ContentAtoms with evidence anchors (page, chapter, offset)< / span >
< / div >
< div class = "pipeline-arrow" > → < / div >
< div class = "pipeline-stage" >
< div class = "num" style = "background:#2563eb;" > 3< / div >
< span class = "label" > Chunk< / span >
< span class = "desc" > Deterministic 500-800 token splits with 50-token overlap, stable IDs< / span >
< / div >
< div class = "pipeline-arrow" > → < / div >
< div class = "pipeline-stage" >
< div class = "num" style = "background:#0891b2;" > 4< / div >
< span class = "label" > Embed< / span >
2026-02-13 18:29:23 +00:00
< span class = "desc" > 768-dim vectors via nomic-embed-text stored in SQLite< / span >
2026-02-13 17:09:46 +00:00
< / div >
< div class = "pipeline-arrow" > → < / div >
< div class = "pipeline-stage" >
< div class = "num" style = "background:#059669;" > 5< / div >
< span class = "label" > Annotate< / span >
< span class = "desc" > LLM extracts topics, entities, claims, sentiment, summary< / span >
< / div >
< div class = "pipeline-arrow" > → < / div >
< div class = "pipeline-stage" >
< div class = "num" style = "background:#d97706;" > 6< / div >
< span class = "label" > Conceptualize< / span >
< span class = "desc" > K-means++ clustering, concept labeling, kNN similarity graph< / span >
< / div >
< / div >
< div style = "text-align:center; margin-top:1rem; font-size:0.8rem; color:var(--fg-subtle);" >
Status tracked per-asset: < code > pending → extracted → chunked → embedded → annotated< / code >
< / div >
< / div >
< / div >
<!-- ============================================================
SLIDE 6: EXTRACTORS
============================================================ -->
< div class = "slide" data-slide = "5" >
< div class = "slide-inner" >
< span class = "tag tag-cyan" style = "margin-bottom:1rem;" > Format Support< / span >
< h2 > Seven Pluggable Extractors< / h2 >
< p class = "lead" > Priority-sorted extractor registry. Each extractor produces ContentAtoms with evidence anchors. The Tika fallback handles anything the others miss.< / p >
< div class = "extractor-grid" >
< div class = "extractor-item" >
< div class = "ext-icon" > 📑 < / div >
< div class = "ext-name" > PDF< / div >
2026-02-13 18:29:23 +00:00
< div class = "ext-types" > .pdf · Per-page text · pdftotext< / div >
2026-02-13 17:09:46 +00:00
< div > < span class = "tag tag-accent" style = "font-size:0.6rem;" > Priority 20< / span > < / div >
< / div >
< div class = "extractor-item" >
< div class = "ext-icon" > 📖 < / div >
< div class = "ext-name" > EPUB< / div >
< div class = "ext-types" > .epub · OPF spine order · Metadata< / div >
< div > < span class = "tag tag-accent" style = "font-size:0.6rem;" > Priority 18< / span > < / div >
< / div >
< div class = "extractor-item" >
< div class = "ext-icon" > 🖼 < / div >
< div class = "ext-name" > Image + OCR< / div >
< div class = "ext-types" > .png .jpg .heic · macOS Vision< / div >
< div > < span class = "tag tag-accent" style = "font-size:0.6rem;" > Priority 15< / span > < / div >
< / div >
< div class = "extractor-item" >
< div class = "ext-icon" > 🩻 < / div >
< div class = "ext-name" > DICOM< / div >
2026-02-13 18:29:23 +00:00
< div class = "ext-types" > .dcm · Medical metadata · Binary header parsing< / div >
2026-02-13 17:09:46 +00:00
< div > < span class = "tag tag-accent" style = "font-size:0.6rem;" > Priority 15< / span > < / div >
< / div >
< div class = "extractor-item" >
< div class = "ext-icon" > 📝 < / div >
< div class = "ext-name" > Text< / div >
< div class = "ext-types" > .txt .md .html .rtf · Tag stripping< / div >
< div > < span class = "tag tag-accent" style = "font-size:0.6rem;" > Priority 10< / span > < / div >
< / div >
< div class = "extractor-item" >
< div class = "ext-icon" > 📦 < / div >
< div class = "ext-name" > Archive< / div >
< div class = "ext-types" > .zip .tar.gz · Sandboxed · Zip-slip safe< / div >
< div > < span class = "tag tag-accent" style = "font-size:0.6rem;" > Priority 5< / span > < / div >
< / div >
< div class = "extractor-item" >
< div class = "ext-icon" > 🔄 < / div >
< div class = "ext-name" > Tika Fallback< / div >
< div class = "ext-types" > .doc .docx .pptx · textutil · Raw text< / div >
< div > < span class = "tag tag-accent" style = "font-size:0.6rem;" > Priority 1< / span > < / div >
< / div >
< div class = "extractor-item" style = "background:var(--bg-muted); border-style:dashed;" >
< div class = "ext-icon" style = "opacity:0.4;" > ➕ < / div >
< div class = "ext-name" style = "color:var(--fg-subtle);" > Custom< / div >
< div class = "ext-types" > Extend BaseExtractor · Register in registry< / div >
< div > < span class = "tag" style = "font-size:0.6rem; background:var(--bg-muted); color:var(--fg-subtle);" > Pluggable< / span > < / div >
< / div >
< / div >
< / div >
< / div >
<!-- ============================================================
SLIDE 7: ANNOTATIONS
============================================================ -->
< div class = "slide" data-slide = "6" >
< div class = "slide-inner" >
< span class = "tag tag-green" style = "margin-bottom:1rem;" > LLM Intelligence< / span >
< h2 > Rich Structured Annotations< / h2 >
< p class = "lead" > Every chunk is annotated by the local LLM, producing structured metadata that enriches search and powers concept formation.< / p >
< div class = "two-col" >
< div >
< div class = "annotation-example" >
< div class = "ann-field" >
< div class = "ann-label" > Topics< / div >
< div class = "ann-tags" >
< span class = "ann-tag" style = "background:#eef2ff; color:#4f46e5;" > cryptography< / span >
< span class = "ann-tag" style = "background:#f5f3ff; color:#7c3aed;" > encryption< / span >
< span class = "ann-tag" style = "background:#eff6ff; color:#2563eb;" > security< / span >
< span class = "ann-tag" style = "background:#ecfeff; color:#0891b2;" > post-quantum< / span >
< span class = "ann-tag" style = "background:#ecfdf5; color:#059669;" > digital signatures< / span >
< / div >
< / div >
< div class = "ann-field" >
< div class = "ann-label" > Summary< / div >
< p style = "font-size:0.85rem; color:var(--fg-muted); font-style:italic;" >
"This text describes various cryptographic techniques, including symmetric and asymmetric encryption algorithms like AES and RSA..."
< / p >
< / div >
< div class = "ann-field" >
< div class = "ann-label" > Entities< / div >
< div class = "ann-tags" >
< span class = "ann-tag" style = "background:#fff1f2; color:#e11d48;" > AES< / span >
< span class = "ann-tag" style = "background:#fff1f2; color:#e11d48;" > RSA< / span >
< span class = "ann-tag" style = "background:#fff1f2; color:#e11d48;" > ECC< / span >
< span class = "ann-tag" style = "background:#fffbeb; color:#d97706;" > NIST< / span >
< span class = "ann-tag" style = "background:#fff1f2; color:#e11d48;" > SHA-256< / span >
< span class = "ann-tag" style = "background:#fff1f2; color:#e11d48;" > EdDSA< / span >
< / div >
< / div >
< div class = "ann-field" style = "margin-bottom:0;" >
< div class = "ann-label" > Sentiment< / div >
< span class = "badge" style = "background:#f0f2f7; color:#5a6072;" > ⚖ neutral · 0.95< / span >
< / div >
< / div >
< / div >
< div >
< h3 > Annotation Fields< / h3 >
< table class = "data-table" >
< tr > < th > Field< / th > < th > Type< / th > < / tr >
< tr > < td > < code > topics< / code > < / td > < td > 2-5 multi-label tags< / td > < / tr >
< tr > < td > < code > entities< / code > < / td > < td > Named entities + type< / td > < / tr >
< tr > < td > < code > claims< / code > < / td > < td > Extracted claims + confidence< / td > < / tr >
< tr > < td > < code > sentiment< / code > < / td > < td > Label + confidence score< / td > < / tr >
< tr > < td > < code > summary< / code > < / td > < td > 1-2 sentence summary< / td > < / tr >
< tr > < td > < code > quality_flags< / code > < / td > < td > Truncated, technical, etc.< / td > < / tr >
< / table >
< div style = "margin-top:1.25rem; padding:1rem; background:var(--bg-subtle); border-radius:var(--radius-sm); font-size:0.82rem; color:var(--fg-muted);" >
< strong style = "color:var(--fg);" > Versioned & Immutable:< / strong > Annotations are never overwritten. New model/prompt versions create new records with < code > is_current=1< / code > , preserving the full audit trail.
< / div >
< / div >
< / div >
< / div >
< / div >
<!-- ============================================================
SLIDE 8: CONCEPT UNIVERSE
============================================================ -->
< div class = "slide" data-slide = "7" >
< div class = "slide-inner" >
< span class = "tag tag-accent" style = "margin-bottom:1rem;" > Visualization< / span >
< h2 > 3D Concept Universe< / h2 >
< p class = "lead" > Concepts are rendered as a force-directed 3D graph using WebGPU. Orbit, zoom, and click to explore how your knowledge connects.< / p >
< div class = "two-col" >
< div >
< div style = "background:#0a0a14; border-radius:var(--radius); padding:2rem; text-align:center; position:relative; overflow:hidden;" >
<!-- Simulated universe -->
< svg viewBox = "0 0 400 280" style = "width:100%; max-width:400px;" >
<!-- Edges -->
< line x1 = "120" y1 = "90" x2 = "280" y2 = "120" stroke = "#4f46e560" stroke-width = "1" / >
< line x1 = "120" y1 = "90" x2 = "200" y2 = "210" stroke = "#7c3aed40" stroke-width = "1" / >
< line x1 = "280" y1 = "120" x2 = "200" y2 = "210" stroke = "#2563eb40" stroke-width = "1" / >
< line x1 = "120" y1 = "90" x2 = "80" y2 = "160" stroke = "#4f46e530" stroke-width = "0.5" / >
< line x1 = "120" y1 = "90" x2 = "160" y2 = "50" stroke = "#4f46e530" stroke-width = "0.5" / >
< line x1 = "280" y1 = "120" x2 = "320" y2 = "70" stroke = "#2563eb30" stroke-width = "0.5" / >
< line x1 = "280" y1 = "120" x2 = "340" y2 = "170" stroke = "#2563eb30" stroke-width = "0.5" / >
< line x1 = "200" y1 = "210" x2 = "160" y2 = "250" stroke = "#7c3aed30" stroke-width = "0.5" / >
< line x1 = "200" y1 = "210" x2 = "240" y2 = "260" stroke = "#7c3aed30" stroke-width = "0.5" / >
<!-- Concept nodes -->
< circle cx = "120" cy = "90" r = "22" fill = "#4f46e5" opacity = "0.9" / >
< circle cx = "280" cy = "120" r = "20" fill = "#2563eb" opacity = "0.9" / >
< circle cx = "200" cy = "210" r = "18" fill = "#7c3aed" opacity = "0.9" / >
<!-- Chunk nodes -->
< circle cx = "80" cy = "160" r = "5" fill = "#4f46e580" / >
< circle cx = "160" cy = "50" r = "5" fill = "#4f46e580" / >
< circle cx = "320" cy = "70" r = "5" fill = "#2563eb80" / >
< circle cx = "340" cy = "170" r = "5" fill = "#2563eb80" / >
< circle cx = "160" cy = "250" r = "5" fill = "#7c3aed80" / >
< circle cx = "240" cy = "260" r = "5" fill = "#7c3aed80" / >
<!-- Labels -->
< text x = "120" y = "94" text-anchor = "middle" fill = "white" font-size = "7" font-weight = "600" > AI Learning< / text >
< text x = "280" y = "124" text-anchor = "middle" fill = "white" font-size = "7" font-weight = "600" > Data Systems< / text >
< text x = "200" y = "214" text-anchor = "middle" fill = "white" font-size = "7" font-weight = "600" > Security< / text >
< / svg >
< div style = "position:absolute; top:0.75rem; left:0.75rem; display:flex; gap:0.5rem;" >
< span style = "background:rgba(255,255,255,0.1); color:rgba(255,255,255,0.7); font-size:0.6rem; padding:0.2rem 0.5rem; border-radius:4px; backdrop-filter:blur(8px);" > 9 nodes< / span >
< span style = "background:rgba(255,255,255,0.1); color:rgba(255,255,255,0.7); font-size:0.6rem; padding:0.2rem 0.5rem; border-radius:4px;" > MACRO< / span >
< span style = "background:rgba(255,255,255,0.1); color:rgba(255,255,255,0.7); font-size:0.6rem; padding:0.2rem 0.5rem; border-radius:4px;" > 60 fps< / span >
< / div >
< / div >
< / div >
< div >
< h3 > Rendering Engine< / h3 >
< table class = "data-table" >
< tr > < td > < strong > Renderer< / strong > < / td > < td > WebGPU in WKWebView< / td > < / tr >
< tr > < td > < strong > Shaders< / strong > < / td > < td > WGSL (214 lines)< / td > < / tr >
< tr > < td > < strong > Layout< / strong > < / td > < td > Force-directed (Velocity Verlet)< / td > < / tr >
< tr > < td > < strong > Nodes< / strong > < / td > < td > Billboarded quads with circle SDF + glow< / td > < / tr >
< tr > < td > < strong > Camera< / strong > < / td > < td > Orbit (drag), Pan (right-drag), Zoom (scroll)< / td > < / tr >
< / table >
< h3 style = "margin-top:1.25rem;" > Level-of-Detail System< / h3 >
< table class = "data-table" >
< tr > < th > LOD< / th > < th > Shows< / th > < th > Zoom< / th > < / tr >
< tr > < td > < span class = "tag tag-accent" style = "font-size:0.65rem;" > MACRO< / span > < / td > < td > Concept clusters only< / td > < td > Distant< / td > < / tr >
< tr > < td > < span class = "tag tag-cyan" style = "font-size:0.65rem;" > MID< / span > < / td > < td > Concepts + sub-concepts< / td > < td > Medium< / td > < / tr >
< tr > < td > < span class = "tag tag-green" style = "font-size:0.65rem;" > NEAR< / span > < / td > < td > All nodes + all edges< / td > < td > Close< / td > < / tr >
< / table >
< / div >
< / div >
< / div >
< / div >
<!-- ============================================================
SLIDE 9: DATA MODEL
============================================================ -->
< div class = "slide" data-slide = "8" >
< div class = "slide-inner" >
< span class = "tag tag-accent" style = "margin-bottom:1rem;" > Data Model< / span >
< h2 > Evidence-Native Knowledge Graph< / h2 >
< p class = "lead" > Every derived artifact links back to its source. The data model captures the full provenance chain from file to concept.< / p >
< div class = "two-col" >
< div >
< div class = "code-block" style = "font-size:0.72rem;" >
< span class = "comment" > -- Provenance chain:< / span >
WatchedVolume
< span class = "op" > ↓ < / span > < span class = "comment" > (contains files)< / span >
FileAsset < span class = "comment" > -- SHA-256 ID, content hash< / span >
< span class = "op" > ↓ < / span > < span class = "comment" > (extracted into)< / span >
ContentAtom < span class = "comment" > -- text/image/table/metadata< / span >
< span class = "op" > ↓ < / span > < span class = "comment" > (split into)< / span >
Chunk < span class = "comment" > -- deterministic ID, 500-800 tokens< / span >
< span class = "op" > ↓ < / span >
2026-02-13 18:29:23 +00:00
< span class = "op" > ├──> < / span > Vector < span class = "comment" > (SQLite BLOB, 768-dim)< / span >
2026-02-13 17:09:46 +00:00
< span class = "op" > ├──> < / span > Annotation < span class = "comment" > (versioned, immutable)< / span >
< span class = "op" > └──> < / span > GraphEdge
< span class = "op" > ↓ < / span >
ConceptNode < span class = "comment" > (hierarchical clusters)< / span > < / div >
< / div >
< div >
< h3 > Evidence Anchors< / h3 >
< p style = "font-size:0.85rem; color:var(--fg-muted); margin-bottom:1rem;" > Every ContentAtom, Chunk, and Edge stores a JSON evidence anchor linking to the exact source location.< / p >
< div class = "code-block" style = "font-size:0.72rem;" >
{
< span class = "string" > "asset_id"< / span > : < span class = "string" > "abc123..."< / span > ,
< span class = "string" > "page"< / span > : < span class = "number" > 5< / span > ,
< span class = "string" > "chapter"< / span > : < span class = "string" > "Introduction"< / span > ,
< span class = "string" > "archive_chain"< / span > : < span class = "string" > "docs.zip/paper.pdf"< / span > ,
< span class = "string" > "line_start"< / span > : < span class = "number" > 42< / span > ,
< span class = "string" > "line_end"< / span > : < span class = "number" > 58< / span >
}< / div >
< h3 style = "margin-top:1rem;" > Storage< / h3 >
< table class = "data-table" style = "font-size:0.8rem;" >
2026-02-13 18:29:23 +00:00
< tr > < td > < strong > SQLite< / strong > < / td > < td > Metadata, vectors (BLOB), graph, jobs (WAL mode)< / td > < / tr >
2026-02-13 17:09:46 +00:00
< tr > < td > < strong > Disk< / strong > < / td > < td > ~/.knowledge-refinery/< / td > < / tr >
< / table >
< / div >
< / div >
< / div >
< / div >
<!-- ============================================================
SLIDE 10: SECURITY
============================================================ -->
< div class = "slide" data-slide = "9" >
< div class = "slide-inner" >
< span class = "tag tag-rose" style = "margin-bottom:1rem;" > Security< / span >
< h2 > Defense in Depth< / h2 >
< p class = "lead" > Zero cloud calls. Localhost-only daemon. Sandboxed extraction. Every security layer works locally.< / p >
< div class = "card-grid cols-3" >
< div class = "card" >
< div class = "icon indigo" > 🔐 < / div >
< h3 > Localhost-Only Binding< / h3 >
< p > Daemon binds exclusively to < code > 127.0.0.1< / code > — not reachable from the network. No tokens needed for a single-user local app.< / p >
< / div >
< div class = "card" >
< div class = "icon violet" > 🛡 < / div >
< h3 > macOS Sandbox< / h3 >
< p > Archive extraction runs in < code > sandbox-exec< / code > with: no network, restricted filesystem, CPU and memory limits.< / p >
< / div >
< div class = "card" >
< div class = "icon blue" > 🚫 < / div >
< h3 > Zip-Slip Prevention< / h3 >
< p > All archive member paths are validated and resolved against the extraction base directory. Traversal attempts are blocked.< / p >
< / div >
< div class = "card" >
< div class = "icon cyan" > 💣 < / div >
< h3 > Archive Bomb Detection< / h3 >
< p > Limits enforced: 10,000 max files, 500MB total extracted, 50MB per file, max 3 nesting levels.< / p >
< / div >
< div class = "card" >
< div class = "icon emerald" > 💻 < / div >
< h3 > 100% Local< / h3 >
< p > LM Studio runs on localhost. Data stays on disk. No telemetry, no cloud APIs, no external network calls. Zero data leakage.< / p >
< / div >
< div class = "card" >
< div class = "icon amber" > 🔍 < / div >
< h3 > Content Hashing< / h3 >
< p > SHA-256 streaming hash for change detection. Deterministic chunk IDs ensure stable references across re-processing.< / p >
< / div >
< / div >
< / div >
< / div >
<!-- ============================================================
SLIDE 11: MASTER CONTROL APP
============================================================ -->
< div class = "slide" data-slide = "10" >
< div class = "slide-inner" >
< span class = "tag tag-accent" style = "margin-bottom:1rem;" > Milestone 7< / span >
< h2 > Master Control App< / h2 >
< p class = "lead" > One dashboard to manage everything. LM Studio monitoring, multi-workspace lifecycle, visual data lake mapping — all from a single window.< / p >
< div class = "card-grid cols-3" >
< div class = "card" >
< div class = "icon indigo" > 📊 < / div >
< h3 > Dashboard< / h3 >
< p > LM Studio status card with model names. Workspace grid with start/stop toggles, vector counts, and color-coded cards. "Start All" for one-click launch.< / p >
< / div >
< div class = "card" >
< div class = "icon violet" > 📦 < / div >
< h3 > Multi-Workspace< / h3 >
2026-02-13 18:29:23 +00:00
< p > Each workspace gets its own data directory, daemon port (8742, 8743, ...), and SQLite database (metadata + vectors + graph). Independent lifecycle management.< / p >
2026-02-13 17:09:46 +00:00
< / div >
< div class = "card" >
< div class = "icon blue" > 🗺 < / div >
< h3 > Data Lake Mapping< / h3 >
< p > Visual Canvas view showing which folders feed which workspaces. Bé zier curves connect data lakes to knowledge bases with color-coded workspace tags.< / p >
< / div >
< div class = "card" >
< div class = "icon cyan" > 🧠 < / div >
< h3 > LM Studio Monitor< / h3 >
< p > Direct polling of < code > /v1/models< / code > every 5 seconds. Auto-classifies chat vs. embedding models. Independent of daemon status — green means ready.< / p >
< / div >
< div class = "card" >
< div class = "icon emerald" > 🔧 < / div >
< h3 > Daemon Lifecycle< / h3 >
< p > Start, stop, restart per-workspace daemons. Environment variables < code > KR_DATA_DIR< / code > and < code > KR_PORT< / code > injected automatically. Live log capture (last 500 lines).< / p >
< / div >
< div class = "card" >
< div class = "icon amber" > 📁 < / div >
< h3 > Workspace Setup< / h3 >
< p > Create workspaces with name, color tag, and folder picker. Native < code > NSOpenPanel< / code > for multi-selecting data lake paths including external drives.< / p >
< / div >
< / div >
< div style = "margin-top:1.5rem; text-align:center;" >
< div style = "display:inline-flex; gap:0.75rem; flex-wrap:wrap; font-size:0.8rem;" >
< div class = "badge" style = "background:var(--accent-bg); color:var(--accent);" > WorkspaceConfig.swift< / div >
< div class = "badge" style = "background:#f5f3ff; color:#7c3aed;" > LMStudioMonitor.swift< / div >
< div class = "badge" style = "background:#ecfdf5; color:#059669;" > MasterDashboardView.swift< / div >
< div class = "badge" style = "background:#ecfeff; color:#0891b2;" > DataLakeMappingView.swift< / div >
< div class = "badge" style = "background:#fffbeb; color:#d97706;" > WorkspaceDetailView.swift< / div >
< / div >
< / div >
< / div >
< / div >
<!-- ============================================================
SLIDE 12: INSTALLATION
============================================================ -->
< div class = "slide" data-slide = "11" >
< div class = "slide-inner" >
< span class = "tag tag-green" style = "margin-bottom:1rem;" > Installation< / span >
< h2 > From Clone to Running in One Command< / h2 >
< div class = "two-col" >
< div >
< h3 > Prerequisites< / h3 >
< table class = "data-table" style = "margin-bottom:1.5rem;" >
< tr > < td > < strong > macOS< / strong > < / td > < td > Tahoe (26.x) on Apple Silicon< / td > < / tr >
< tr > < td > < strong > Xcode< / strong > < / td > < td > 26.x or Command Line Tools< / td > < / tr >
2026-02-13 18:29:23 +00:00
< tr > < td > < strong > Go< / strong > < / td > < td > 1.22+ (from go.dev or Homebrew)< / td > < / tr >
2026-02-13 17:09:46 +00:00
< tr > < td > < strong > LM Studio< / strong > < / td > < td > From lmstudio.ai (free)< / td > < / tr >
< / table >
< h3 > What the Installer Does< / h3 >
< ol class = "step-list" >
< li >
< strong > Checks prerequisites< / strong >
2026-02-13 18:29:23 +00:00
< p > Validates macOS version, architecture, Xcode tools, Swift, and Go version.< / p >
2026-02-13 17:09:46 +00:00
< / li >
< li >
2026-02-13 18:29:23 +00:00
< strong > Builds Go daemon< / strong >
< p > Compiles the Go daemon into a single 17MB binary with zero runtime dependencies.< / p >
2026-02-13 17:09:46 +00:00
< / li >
< li >
< strong > Builds the .app bundle< / strong >
< p > Swift release build → proper < code > .app< / code > with Info.plist, bundled daemon, and WebGPU resources.< / p >
< / li >
< li >
< strong > Installs to /Applications< / strong >
< p > Copies the app bundle. Appears in Launchpad and Spotlight immediately.< / p >
< / li >
< / ol >
< / div >
< div >
< h3 > One-Line Install< / h3 >
< div class = "code-block" style = "font-size:0.75rem; margin-bottom:1.25rem;" >
< span class = "comment" > # Clone and install< / span >
< span class = "fn" > git< / span > clone < repo-url>
< span class = "keyword" > cd< / span > LongLocalTimeHorizonInfoRetrieval
< span class = "fn" > bash< / span > scripts/install.sh< / div >
< h3 > Make Targets< / h3 >
< div class = "code-block" style = "font-size:0.75rem; margin-bottom:1.25rem;" >
< span class = "fn" > make< / span > build < span class = "comment" > # Build .app bundle to dist/< / span >
< span class = "fn" > make< / span > install < span class = "comment" > # Full install to /Applications< / span >
< span class = "fn" > make< / span > test < span class = "comment" > # Run all tests< / span >
< span class = "fn" > make< / span > app-run < span class = "comment" > # Dev mode (swift run)< / span >
< span class = "fn" > make< / span > daemon-run < span class = "comment" > # Run daemon directly< / span >
< span class = "fn" > make< / span > clean < span class = "comment" > # Remove build artifacts< / span > < / div >
< h3 > App Bundle Structure< / h3 >
< div class = "code-block" style = "font-size:0.72rem;" >
Knowledge Refinery.app/
Contents/
Info.plist
MacOS/
KnowledgeRefinery < span class = "comment" > # launcher< / span >
KnowledgeRefinery-bin < span class = "comment" > # Swift binary< / span >
Resources/
2026-02-13 18:29:23 +00:00
knowledge-refinery-daemon < span class = "comment" > # Go binary (17MB)< / span >
2026-02-13 17:09:46 +00:00
WebGPU/ < span class = "comment" > # 3D renderer< / span >
universe.html/js/wgsl< / div >
< / div >
< / div >
< / div >
< / div >
<!-- ============================================================
SLIDE 13: APP WALKTHROUGH
============================================================ -->
< div class = "slide" data-slide = "12" >
< div class = "slide-inner" >
< span class = "tag tag-accent" style = "margin-bottom:1rem;" > App Walkthrough< / span >
< h2 > Dashboard + Workspace: Two-Level UI< / h2 >
< div class = "two-col" >
< div >
< h3 style = "color:var(--accent);" > Level 1: Dashboard< / h3 >
< div class = "card" style = "margin-bottom:0.75rem;" >
< div class = "icon indigo" > 📊 < / div >
< h3 > LM Studio Card< / h3 >
< p > Green/red status, loaded model names (chat + embedding), port number. Polls directly at < code > /v1/models< / code > .< / p >
< / div >
< div class = "card" style = "margin-bottom:0.75rem;" >
< div class = "icon violet" > 📦 < / div >
< h3 > Workspace Grid< / h3 >
< p > Color-coded cards with daemon toggle, vector count, data lake count. Click "Open" to enter a workspace.< / p >
< / div >
< div class = "card" >
< div class = "icon blue" > 🗺 < / div >
< h3 > Data Lake Map< / h3 >
< p > Canvas-drawn Bé zier curves connecting folder paths to workspaces. See which folders feed which knowledge bases.< / p >
< / div >
< / div >
< div >
< h3 style = "color:var(--accent-2);" > Level 2: Workspace< / h3 >
< div class = "card" style = "margin-bottom:0.75rem;" >
< div class = "icon emerald" > 🔎 < / div >
< h3 > Search + Universe + Concepts< / h3 >
< p > Full NavigationSplitView with semantic search, WebGPU 3D concept graph, hierarchical concept browser with "Why?" explanations.< / p >
< / div >
< div class = "card" style = "margin-bottom:0.75rem;" >
< div class = "icon cyan" > ⬇ < / div >
< h3 > Ingest + Volumes + Assets< / h3 >
< p > Pipeline monitoring, volume management, file inventory. Each workspace has its own daemon process and data directory.< / p >
< / div >
< div class = "card" >
< div class = "icon amber" > 🔧 < / div >
< h3 > Daemon Controls + Logs< / h3 >
< p > Play/stop/restart buttons in the header bar. Live log viewer showing last 500 lines of daemon stdout/stderr.< / p >
< / div >
< / div >
< / div >
< div style = "margin-top:1.5rem; text-align:center;" >
< div style = "display:inline-flex; gap:1rem; align-items:center; padding:0.75rem 1.5rem; background:var(--bg-subtle); border-radius:var(--radius-sm); font-size:0.8rem; color:var(--fg-muted);" >
< span > 🟢 LM Studio (shared)< / span >
< span > 🟢 Workspace 1 :8742< / span >
< span > 🟢 Workspace 2 :8743< / span >
< span style = "margin-left:0.5rem;" > Each workspace runs its own daemon< / span >
< / div >
< / div >
< / div >
< / div >
<!-- ============================================================
SLIDE 14: SEARCH DEEP DIVE
============================================================ -->
< div class = "slide" data-slide = "13" >
< div class = "slide-inner" >
< span class = "tag tag-accent" style = "margin-bottom:1rem;" > Deep Dive< / span >
< h2 > How Search Works< / h2 >
< div class = "two-col" >
< div >
< h3 > Query Flow< / h3 >
< div style = "background:var(--bg-subtle); border-radius:var(--radius); padding:1.25rem; font-size:0.85rem; line-height:2;" >
< div style = "display:flex; align-items:center; gap:0.75rem;" >
< span style = "background:var(--accent); color:white; width:24px; height:24px; border-radius:50%; display:inline-flex; align-items:center; justify-content:center; font-size:0.7rem; font-weight:700; flex-shrink:0;" > 1< / span >
User types: < code > "cryptography encryption"< / code >
< / div >
< div style = "color:var(--fg-subtle); padding-left:2.3rem;" > ↓ < / div >
< div style = "display:flex; align-items:center; gap:0.75rem;" >
< span style = "background:#0891b2; color:white; width:24px; height:24px; border-radius:50%; display:inline-flex; align-items:center; justify-content:center; font-size:0.7rem; font-weight:700; flex-shrink:0;" > 2< / span >
Embed query via nomic-embed-text → 768-dim vector
< / div >
< div style = "color:var(--fg-subtle); padding-left:2.3rem;" > ↓ < / div >
< div style = "display:flex; align-items:center; gap:0.75rem;" >
< span style = "background:#7c3aed; color:white; width:24px; height:24px; border-radius:50%; display:inline-flex; align-items:center; justify-content:center; font-size:0.7rem; font-weight:700; flex-shrink:0;" > 3< / span >
2026-02-13 18:29:23 +00:00
Brute-force cosine similarity search (in-memory vectors)
2026-02-13 17:09:46 +00:00
< / div >
< div style = "color:var(--fg-subtle); padding-left:2.3rem;" > ↓ < / div >
< div style = "display:flex; align-items:center; gap:0.75rem;" >
< span style = "background:#059669; color:white; width:24px; height:24px; border-radius:50%; display:inline-flex; align-items:center; justify-content:center; font-size:0.7rem; font-weight:700; flex-shrink:0;" > 4< / span >
Enrich with annotations (topics, summary, entities)
< / div >
< div style = "color:var(--fg-subtle); padding-left:2.3rem;" > ↓ < / div >
< div style = "display:flex; align-items:center; gap:0.75rem;" >
< span style = "background:#d97706; color:white; width:24px; height:24px; border-radius:50%; display:inline-flex; align-items:center; justify-content:center; font-size:0.7rem; font-weight:700; flex-shrink:0;" > 5< / span >
Return ranked results with evidence anchors
< / div >
< / div >
< / div >
< div >
< h3 > Example Result< / h3 >
< div class = "card" style = "border-left:3px solid var(--accent);" >
< div style = "display:flex; justify-content:space-between; align-items:center; margin-bottom:0.5rem;" >
< strong style = "font-size:0.9rem;" > 📄 cryptography_basics.html< / strong >
< span class = "badge" style = "background:#ecfdf5; color:#059669; font-size:0.7rem;" > score: 0.378< / span >
< / div >
< div style = "margin-bottom:0.75rem;" >
< div style = "font-size:0.75rem; font-weight:600; color:var(--fg-subtle); margin-bottom:0.2rem;" > LLM SUMMARY< / div >
< p style = "font-size:0.82rem; color:var(--fg-muted); font-style:italic;" > "This text describes various cryptographic techniques, including symmetric and asymmetric encryption..."< / p >
< / div >
< div style = "margin-bottom:0.75rem;" >
< div style = "font-size:0.75rem; font-weight:600; color:var(--fg-subtle); margin-bottom:0.2rem;" > ENTITIES< / div >
< div style = "display:flex; flex-wrap:wrap; gap:0.3rem;" >
< span class = "ann-tag" style = "background:#eef2ff; color:#4f46e5; font-size:0.7rem;" > AES< / span >
< span class = "ann-tag" style = "background:#eef2ff; color:#4f46e5; font-size:0.7rem;" > RSA< / span >
< span class = "ann-tag" style = "background:#eef2ff; color:#4f46e5; font-size:0.7rem;" > ECC< / span >
< span class = "ann-tag" style = "background:#fffbeb; color:#d97706; font-size:0.7rem;" > NIST< / span >
< span class = "ann-tag" style = "background:#eef2ff; color:#4f46e5; font-size:0.7rem;" > SHA-256< / span >
< / div >
< / div >
< div >
< span class = "badge" style = "background:#f0f2f7; font-size:0.7rem;" > ⚖ neutral< / span >
< / div >
< / div >
< div style = "margin-top:1rem; font-size:0.82rem; color:var(--fg-muted);" >
< strong > Key insight:< / strong > A search for "cryptography" finds the right file even though the query doesn't match any specific keywords in other documents. Semantic similarity beats keyword search.
< / div >
< / div >
< / div >
< / div >
< / div >
<!-- ============================================================
SLIDE 15: API REFERENCE
============================================================ -->
< div class = "slide" data-slide = "14" >
< div class = "slide-inner" >
< span class = "tag tag-cyan" style = "margin-bottom:1rem;" > API Reference< / span >
< h2 > RESTful API Endpoints< / h2 >
< p class = "lead" > The daemon exposes a clean REST API on localhost. All endpoints are accessible without authentication — the daemon only binds to 127.0.0.1.< / p >
< table class = "data-table" style = "font-size:0.82rem;" >
< thead >
< tr > < th > Method< / th > < th > Endpoint< / th > < th > Description< / th > < / tr >
< / thead >
< tbody >
< tr > < td > < span class = "badge" style = "background:#ecfdf5; color:#059669; font-size:0.65rem;" > GET< / span > < / td > < td > < code > /health< / code > < / td > < td > Health check + LM Studio status (no auth)< / td > < / tr >
< tr > < td > < span class = "badge" style = "background:#eff6ff; color:#2563eb; font-size:0.65rem;" > POST< / span > < / td > < td > < code > /volumes/add< / code > < / td > < td > Register a folder to watch< / td > < / tr >
< tr > < td > < span class = "badge" style = "background:#ecfdf5; color:#059669; font-size:0.65rem;" > GET< / span > < / td > < td > < code > /volumes/list< / code > < / td > < td > List all watched volumes< / td > < / tr >
< tr > < td > < span class = "badge" style = "background:#eff6ff; color:#2563eb; font-size:0.65rem;" > POST< / span > < / td > < td > < code > /ingest/start< / code > < / td > < td > Trigger full 6-stage pipeline< / td > < / tr >
< tr > < td > < span class = "badge" style = "background:#ecfdf5; color:#059669; font-size:0.65rem;" > GET< / span > < / td > < td > < code > /ingest/status< / code > < / td > < td > Pipeline progress + per-stage counts< / td > < / tr >
< tr > < td > < span class = "badge" style = "background:#eff6ff; color:#2563eb; font-size:0.65rem;" > POST< / span > < / td > < td > < code > /search< / code > < / td > < td > Semantic vector search with annotation enrichment< / td > < / tr >
< tr > < td > < span class = "badge" style = "background:#ecfdf5; color:#059669; font-size:0.65rem;" > GET< / span > < / td > < td > < code > /search/quick?q=...< / code > < / td > < td > Quick search via query parameter< / td > < / tr >
< tr > < td > < span class = "badge" style = "background:#ecfdf5; color:#059669; font-size:0.65rem;" > GET< / span > < / td > < td > < code > /universe/snapshot?lod=macro< / code > < / td > < td > Graph snapshot at given LOD level< / td > < / tr >
< tr > < td > < span class = "badge" style = "background:#eff6ff; color:#2563eb; font-size:0.65rem;" > POST< / span > < / td > < td > < code > /universe/focus< / code > < / td > < td > Focus on node and return neighborhood< / td > < / tr >
< tr > < td > < span class = "badge" style = "background:#ecfdf5; color:#059669; font-size:0.65rem;" > GET< / span > < / td > < td > < code > /concepts/list< / code > < / td > < td > List all concept clusters< / td > < / tr >
< tr > < td > < span class = "badge" style = "background:#ecfdf5; color:#059669; font-size:0.65rem;" > GET< / span > < / td > < td > < code > /concepts/{id}< / code > < / td > < td > Concept detail with member chunks< / td > < / tr >
< tr > < td > < span class = "badge" style = "background:#ecfdf5; color:#059669; font-size:0.65rem;" > GET< / span > < / td > < td > < code > /concepts/{id}/why< / code > < / td > < td > "Why this concept?" with evidence chain< / td > < / tr >
< tr > < td > < span class = "badge" style = "background:#ecfdf5; color:#059669; font-size:0.65rem;" > GET< / span > < / td > < td > < code > /evidence/chunk/{id}/annotation< / code > < / td > < td > Full annotation for a specific chunk< / td > < / tr >
< / tbody >
< / table >
< / div >
< / div >
<!-- ============================================================
SLIDE 16: TECH STACK
============================================================ -->
< div class = "slide" data-slide = "15" >
< div class = "slide-inner" >
< span class = "tag tag-accent" style = "margin-bottom:1rem;" > Technology< / span >
< h2 > Technology Stack< / h2 >
< div class = "two-col" >
< div >
< h3 > Frontend (macOS App)< / h3 >
< table class = "data-table" >
< tr > < td > < strong > Framework< / strong > < / td > < td > SwiftUI (macOS Tahoe / 26.x)< / td > < / tr >
< tr > < td > < strong > Language< / strong > < / td > < td > Swift 6.2.3< / td > < / tr >
< tr > < td > < strong > Architecture< / strong > < / td > < td > Multi-workspace + dashboard< / td > < / tr >
< tr > < td > < strong > 3D Rendering< / strong > < / td > < td > WebGPU via WKWebView< / td > < / tr >
< tr > < td > < strong > File Preview< / strong > < / td > < td > QuickLook framework< / td > < / tr >
< tr > < td > < strong > Build< / strong > < / td > < td > SPM + build.sh → .app bundle< / td > < / tr >
< tr > < td > < strong > Install< / strong > < / td > < td > < code > make install< / code > or install.sh< / td > < / tr >
< / table >
< / div >
< div >
2026-02-13 18:29:23 +00:00
< h3 > Backend (Go Daemon)< / h3 >
2026-02-13 17:09:46 +00:00
< table class = "data-table" >
2026-02-13 18:29:23 +00:00
< tr > < td > < strong > Language< / strong > < / td > < td > Go 1.22+ (single 17MB binary)< / td > < / tr >
< tr > < td > < strong > HTTP Router< / strong > < / td > < td > chi/v5 + CORS middleware< / td > < / tr >
< tr > < td > < strong > Storage< / strong > < / td > < td > SQLite (metadata + vectors + graph, WAL mode)< / td > < / tr >
< tr > < td > < strong > SQLite Driver< / strong > < / td > < td > modernc.org/sqlite (pure Go, no CGo)< / td > < / tr >
< tr > < td > < strong > LLM Client< / strong > < / td > < td > net/http → LM Studio< / td > < / tr >
< tr > < td > < strong > PDF< / strong > < / td > < td > pdftotext (poppler-utils)< / td > < / tr >
< tr > < td > < strong > Tokenizer< / strong > < / td > < td > tiktoken-go (cl100k_base)< / td > < / tr >
< tr > < td > < strong > Math< / strong > < / td > < td > Pure Go (k-means++, cosine sim)< / td > < / tr >
2026-02-13 17:09:46 +00:00
< / table >
< h3 style = "margin-top:1.5rem;" > LM Studio Models< / h3 >
< table class = "data-table" >
2026-02-13 18:29:23 +00:00
< tr > < td > < strong > Chat< / strong > < / td > < td > qwen3-4b-2507< / td > < / tr >
2026-02-13 17:09:46 +00:00
< tr > < td > < strong > Embeddings< / strong > < / td > < td > nomic-embed-text-v1.5 (768-dim)< / td > < / tr >
< / table >
< / div >
< / div >
< / div >
< / div >
<!-- ============================================================
SLIDE 17: FINAL
============================================================ -->
< div class = "slide title-slide" data-slide = "16" >
< div class = "slide-inner" style = "text-align:center;" >
< h1 class = "gradient" > Your Knowledge,< br > Refined.< / h1 >
< p class = "subtitle" style = "margin:0.5rem auto 2.5rem;" >
Fully local. Evidence-native. LLM-powered. Installable.< br >
Clone, install, launch — from raw documents to searchable knowledge in minutes.
< / p >
< div class = "stats-row" style = "justify-content:center; margin-bottom:2.5rem;" >
2026-02-13 18:29:23 +00:00
< div class = "stat" > < div class = "stat-num" > 97< / div > < div class = "stat-label" > Tests Passing< / div > < / div >
2026-02-13 17:09:46 +00:00
< div class = "stat" > < div class = "stat-num" > 0< / div > < div class = "stat-label" > Compiler Warnings< / div > < / div >
2026-02-13 18:29:23 +00:00
< div class = "stat" > < div class = "stat-num" > ~90< / div > < div class = "stat-label" > Source Files< / div > < / div >
2026-02-13 17:09:46 +00:00
< div class = "stat" > < div class = "stat-num" > 100%< / div > < div class = "stat-label" > Local< / div > < / div >
< / div >
< div style = "display:flex; gap:1rem; justify-content:center; flex-wrap:wrap; font-size:0.82rem;" >
2026-02-13 18:29:23 +00:00
< div class = "badge" style = "background:var(--accent-bg); color:var(--accent);" > 📁 daemon-go/< / div >
2026-02-13 17:09:46 +00:00
< div class = "badge" style = "background:#f5f3ff; color:#7c3aed;" > 📱 apps/macos/KnowledgeRefinery/< / div >
2026-02-13 18:29:23 +00:00
< div class = "badge" style = "background:#ecfdf5; color:#059669;" > 🧪 daemon-go/internal/*/< / div >
2026-02-13 17:09:46 +00:00
< div class = "badge" style = "background:#fffbeb; color:#d97706;" > 🛠 scripts/install.sh< / div >
< / div >
< div style = "margin-top:1.5rem;" >
< div class = "code-block" style = "display:inline-block; font-size:0.8rem; padding:0.6rem 1.5rem; text-align:left;" >
< span class = "fn" > git< / span > clone < repo> & & < span class = "keyword" > cd< / span > LongLocalTimeHorizonInfoRetrieval & & < span class = "fn" > make< / span > install< / div >
< / div >
< div style = "margin-top:1.5rem; color:var(--fg-subtle); font-size:0.8rem;" >
2026-02-13 18:29:23 +00:00
Built with SwiftUI, Go, chi, SQLite, WebGPU, and local LLMs
2026-02-13 17:09:46 +00:00
< / div >
< / div >
< / div >
< / div > <!-- /deck -->
<!-- ============================================================
NAVIGATION BAR
============================================================ -->
< div class = "nav-bar" >
< button class = "nav-btn" id = "btn-prev" onclick = "navigate(-1)" > ← Prev< / button >
< div class = "progress" > < div class = "progress-fill" id = "progress-fill" > < / div > < / div >
< span class = "slide-counter" id = "slide-counter" > 1 / 17< / span >
< button class = "nav-btn" id = "btn-next" onclick = "navigate(1)" > Next → < / button >
< / div >
< script >
// ============================================================
// SLIDE ENGINE
// ============================================================
const slides = document.querySelectorAll('.slide');
let current = 0;
const total = slides.length;
function goTo(n) {
if (n < 0 | | n > = total) return;
const prev = current;
current = n;
slides.forEach((s, i) => {
s.classList.remove('active', 'prev');
if (i === current) s.classList.add('active');
else if (i === prev) s.classList.add('prev');
});
document.getElementById('slide-counter').textContent = `${current + 1} / ${total}`;
document.getElementById('progress-fill').style.width = `${((current + 1) / total) * 100}%`;
document.getElementById('btn-prev').disabled = current === 0;
document.getElementById('btn-next').disabled = current === total - 1;
// Scroll active slide to top if overflowing
slides[current].scrollTop = 0;
}
function navigate(dir) {
goTo(current + dir);
}
// Keyboard
document.addEventListener('keydown', e => {
if (e.key === 'ArrowRight' || e.key === ' ') { e.preventDefault(); navigate(1); }
else if (e.key === 'ArrowLeft') { e.preventDefault(); navigate(-1); }
else if (e.key === 'Home') { e.preventDefault(); goTo(0); }
else if (e.key === 'End') { e.preventDefault(); goTo(total - 1); }
});
// Touch/swipe
let touchStartX = 0;
let touchStartY = 0;
document.addEventListener('touchstart', e => {
touchStartX = e.touches[0].clientX;
touchStartY = e.touches[0].clientY;
}, { passive: true });
document.addEventListener('touchend', e => {
const dx = e.changedTouches[0].clientX - touchStartX;
const dy = e.changedTouches[0].clientY - touchStartY;
if (Math.abs(dx) > Math.abs(dy) & & Math.abs(dx) > 50) {
navigate(dx < 0 ? 1 : -1 ) ;
}
}, { passive: true });
// Init
goTo(0);
< / script >
< / body >
< / html >
