@mindrian_os/install 1.13.0-beta.11

package/references/design/graph-visualization-standard.md

@@ -0,0 +1,178 @@
+# Graph Visualization Standard
+
+**Decision:** vis-network (vis.js) is the standard graph visualization library for ALL MindrianOS exports, snapshots, dashboards, and visual presentations.
+
+**Decided:** 2026-04-06
+**Supersedes:** Cytoscape.js for exports (Cytoscape.js retained for internal plugin dashboard only)
+
+## Why vis-network
+
+| Requirement | vis-network | Cytoscape.js |
+|-------------|-------------|--------------|
+| Physics simulation | Built-in ForceAtlas2, excellent | Available but less polished |
+| Drag & drop | Excellent out-of-box | Good |
+| Learning curve | Simple API | Steep |
+| Clustering | Built-in | Requires extensions |
+| Visual quality | Beautiful by default | Needs heavy styling |
+| Setup | `new Network(container, data, options)` | Verbose config |
+
+Cytoscape.js has superior graph algorithms (50+ built-in) but exports don't need algorithms -- they need beautiful interactive exploration.
+
+## When to Use What
+
+| Context | Library | Reason |
+|---------|---------|--------|
+| SnapshotHub exports | vis-network | Beauty + interaction for stakeholders |
+| /mos:present views | vis-network | Same as SnapshotHub |
+| /mos:dashboard live server | vis-network | Consistency with exports |
+| /mos:visualize command | vis-network | Standard graph output |
+| Plugin internal (Brain queries) | Cytoscape.js | Graph algorithms needed |
+
+## Standard Configuration
+
+```javascript
+var options = {
+  physics: {
+    enabled: true,
+    solver: 'forceAtlas2Based',
+    forceAtlas2Based: {
+      gravitationalConstant: -40,
+      centralGravity: 0.008,
+      springLength: 120,
+      springConstant: 0.04,
+      damping: 0.4
+    },
+    stabilization: { iterations: 200, fit: true }
+  },
+  interaction: {
+    hover: true,
+    tooltipDelay: 200,
+    zoomView: true,
+    dragView: true,
+    keyboard: { enabled: true }
+  },
+  nodes: {
+    font: {
+      color: '#ccc',
+      size: 11,
+      face: 'Inter, system-ui, sans-serif',
+      strokeWidth: 3,
+      strokeColor: '#0a0a0f'
+    },
+    borderWidth: 2,
+    shadow: { enabled: true, color: 'rgba(0,0,0,0.3)', size: 6 }
+  },
+  edges: {
+    smooth: { type: 'continuous', roundness: 0.3 },
+    hoverWidth: 2
+  },
+  layout: { improvedLayout: true }
+};
+```
+
+## Node Shapes
+
+| Node Type | Shape | Size | Description |
+|-----------|-------|------|-------------|
+| Section header | diamond | 20 | Room section nodes (problem-definition, market-analysis, etc.) |
+| Artifact | dot | 12 | Individual entries/documents |
+| Framework | triangle | 16 | Methodology framework nodes (for Brain views) |
+| Person | square | 14 | Team member or speaker nodes |
+
+## Edge Colors
+
+| Edge Type | Color | Opacity | Description |
+|-----------|-------|---------|-------------|
+| CONVERGES | #2ECC71 | 0.4 | Themes align across sections |
+| CONTRADICTS | #E74C3C | 0.4 | Tension detected between entries |
+| INFORMS | #3498DB | 0.4 | One entry informs another |
+| INVALIDATES | #E67E22 | 0.5 | One entry invalidates another |
+| FEEDS_INTO | #9B59B6 | 0.5 | Framework chains (Brain views) |
+
+## Section Colors
+
+| Section | Color | Hex |
+|---------|-------|-----|
+| problem-definition | Red | #E74C3C |
+| market-analysis | Yellow/Gold | #F39C12 |
+| solution-design | Blue | #3498DB |
+| business-model | Green | #2ECC71 |
+| competitive-analysis | Orange | #E67E22 |
+| team-execution | Purple | #9B59B6 |
+| legal-ip | Gold | #D4AC0D |
+| financial-model | Teal | #1ABC9C |
+| personas | Deep Purple | #8E44AD |
+| opportunity-bank | Forest Green | #27AE60 |
+| meetings | Gray | #7F8C8D |
+
+## CDN
+
+```html
+<script src="https://unpkg.com/vis-network@9.1.9/standalone/umd/vis-network.min.js"></script>
+```
+
+## Required UI Elements
+
+Every graph view MUST include:
+
+1. **Sidebar** with section filter toggles and edge type toggles
+2. **Detail panel** that appears on node click (shows connections)
+3. **Controls bar** at bottom: Fit, Zoom +/-, Physics toggle, Stabilize
+4. **Back to Hub** link in top bar
+5. **Mindrian branding** with Mondrian color bar
+6. **Legend** explaining edge types and node shapes
+7. **Stats** showing node/edge counts
+
+## Dark Theme (Standard)
+
+```css
+body { background: #0a0a0f; }
+sidebar { background: #0d0d14; border-right: 1px solid #1a1a2a; }
+top-bar { background: #0d0d12; }
+node-font-color: #ccc
+node-font-stroke: #0a0a0f (3px outline for readability)
+detail-panel: #12121a with backdrop-filter: blur(8px)
+```
+
+## Integration Points
+
+- `scripts/generate-hub.cjs` -- embeds vis-network graph in SnapshotHub
+- `scripts/generate-snapshot.cjs` -- constellation view uses vis-network
+- `scripts/generate-presentation.cjs` -- graph view uses vis-network
+- `lib/quickview/hub-server.cjs` -- live dashboard graph uses vis-network
+- `dashboard/index.html` -- localhost dashboard uses vis-network
+
+---
+
+## Pitch Deck Standard (v1.9.3)
+
+The SnapshotHub pitch deck uses the `/slides` skill structure:
+- 10-slide YC Seed Deck format
+- Chart.js for all data visualizations (minimum 3 charts)
+- 16:9 aspect ratio with keyboard navigation (arrows + click)
+- Progress bar at top
+- De Stijl dark palette (#0a0a0f background, Mondrian accents)
+- Bebas Neue headings, Inter body
+- MINDRIAN logo + Back to Hub link
+- Mobile responsive (breakpoints at 900px, 600px)
+- Mondrian color bar footer
+
+### Slide Structure
+1. Title/Hook (AIDA formula)
+2. Problem (PAS formula)
+3. Convergence/IKA (Chart.js horizontal bar)
+4. Solution (TRL progress bars)
+5. Market & Timing (Chart.js S-curve lines)
+6. Competitive Landscape (grid with check/miss)
+7. Business Model (Chart.js stacked bar)
+8. Team (credential grid)
+9. Traction & Timeline (milestone cards)
+10. The Ask/CTA (AIDA urgency)
+
+### Generator Integration
+`generate-hub.cjs` should produce this deck as `views/deck.html` by:
+1. Reading room sections for content
+2. Extracting metrics from STATE.md
+3. Building Chart.js datasets from market-analysis/ and financial-model/
+4. Generating persona card from team/ data
+5. Using the slide template from references/templates/deck-template.html

package/references/document-generation.md

@@ -0,0 +1,179 @@
+# Document Generation Reference
+
+*Professional document structuring for MindrianOS — De Stijl formatted exports from the Data Room.*
+
+---
+
+## Architecture
+
+```
+Room Markdown Artifacts (YAML frontmatter)
+    ↓
+markdown2 (parse to HTML + extract metadata)
+    ↓
+Jinja2 (inject into document-type template)
+    ↓
+De Stijl CSS stylesheet (shared across all templates)
+    ↓
+WeasyPrint (render to PDF)
+    ↓
+PyMuPDF (optional: TOC bookmarks, merge sections, watermark)
+    ↓
+room/exports/{document-type}.pdf
+```
+
+**Zero additional dependencies.** All tools already installed:
+- WeasyPrint 67.0 (HTML+CSS → PDF, CSS3 paged media)
+- markdown2 2.5.4 (Markdown → HTML with frontmatter)
+- Jinja2 3.1.6 (template engine)
+- PyMuPDF 1.26.6 (PDF manipulation, TOC, merge)
+- PyYAML 6.0.3 (frontmatter parsing)
+
+---
+
+## Document Types
+
+| Document | Use Case | Template Style |
+|----------|----------|---------------|
+| **Investment Thesis** | Structured narrative for investors — problem, market, solution, financials | Multi-page portrait, section breaks, De Stijl accent bars |
+| **Executive Summary** | 1-2 page dense overview of entire Data Room | Two-column layout, tight margins, summary boxes |
+| **Due Diligence Report** | Comprehensive multi-section with TOC | Running headers, page numbers, section numbering |
+| **Research Paper** | Academic-style with citations and methodology | Two-column body, footnotes, bibliography |
+| **PWS Profile** | Professional profile/resume for cohort members | Single-page grid layout, accent bars, skill visualization |
+| **Pitch Deck** | Landscape slides for presentation | Landscape @page, large headings, one-concept-per-page |
+| **Venture Brief** | Quick 1-pager for stakeholder review | Constrained single page, key highlights only |
+
+---
+
+## De Stijl CSS for PDF
+
+```css
+@page {
+  size: A4;
+  margin: 24mm 20mm;
+  background: #0D0D0D;
+
+  @top-center {
+    content: string(doctitle);
+    font-family: 'Bebas Neue', sans-serif;
+    color: #A09A90;
+    font-size: 10px;
+  }
+
+  @bottom-right {
+    content: counter(page) ' / ' counter(pages);
+    font-family: 'Inter', sans-serif;
+    color: #A09A90;
+    font-size: 10px;
+  }
+}
+
+:root {
+  --ds-bg: #0D0D0D;
+  --ds-surface: #1A1A1A;
+  --ds-elevated: #2A2A2A;
+  --ds-cream: #F5F0E8;
+  --ds-muted: #A09A90;
+  --ds-red: #A63D2F;
+  --ds-blue: #1E3A6E;
+  --ds-yellow: #C8A43C;
+  --ds-green: #2D6B4A;
+  --ds-sienna: #B5602A;
+  --ds-gray: #5C5A56;
+  --ds-amethyst: #6B4E8B;
+  --ds-teal: #2A6B5E;
+}
+
+body {
+  font-family: 'Inter', sans-serif;
+  color: var(--ds-cream);
+  background: var(--ds-bg);
+  line-height: 1.6;
+}
+
+h1, h2, h3 {
+  font-family: 'Bebas Neue', sans-serif;
+  letter-spacing: 0.05em;
+}
+
+* { border-radius: 0; }
+```
+
+---
+
+## Room Section → Document Section Mapping
+
+| Room Section | Document Section | Accent Color |
+|-------------|-----------------|-------------|
+| problem-definition | Problem Statement / Opportunity | Red #A63D2F |
+| market-analysis | Market Analysis / TAM-SAM-SOM | Yellow #C8A43C |
+| solution-design | Solution Architecture / Technical Approach | Gray #5C5A56 |
+| business-model | Business Model / Revenue Strategy | Green #2D6B4A |
+| competitive-analysis | Competitive Landscape / Moat | Sienna #B5602A |
+| team-execution | Team / Execution Plan | Blue #1E3A6E |
+| legal-ip | Legal / IP Strategy | Amethyst #6B4E8B |
+| financial-model | Financial Projections / Investment Ask | Teal #2A6B5E |
+
+---
+
+## Implementation Plan
+
+### Plugin Files Needed
+
+```
+templates/
+  destijl-base.css          # Shared De Stijl stylesheet for PDF
+  investment-thesis.html    # Jinja2 template
+  executive-summary.html    # Jinja2 template
+  due-diligence.html        # Jinja2 template
+  research-paper.html       # Jinja2 template
+  profile.html              # Jinja2 template (PWS profiles)
+  pitch-deck.html           # Jinja2 template (landscape)
+  venture-brief.html        # Jinja2 template (1-page)
+assets/
+  fonts/BebasNeue-Regular.ttf  # Bundled font (~50KB)
+scripts/
+  render-pdf                # Orchestrator: reads room, selects template, renders PDF
+commands/
+  export.md                 # /mos:export thesis|summary|report|profile|deck|brief
+```
+
+### Command UX
+
+```
+/mos:export thesis      → Investment thesis PDF from full room
+/mos:export summary     → 1-2 page executive summary
+/mos:export report      → Due diligence report with TOC
+/mos:export profile     → PWS profile for a cohort member
+/mos:export deck        → Landscape pitch deck
+/mos:export brief       → 1-page venture brief
+```
+
+### Tri-Polar Surface Behavior
+
+| Surface | Behavior |
+|---------|----------|
+| **CLI** | `render-pdf` script generates PDF, opens with default viewer |
+| **Desktop** | Larry offers "Want me to export this as a PDF?" after key milestones |
+| **Cowork** | Export to shared `00_Context/exports/` directory, visible to all team members |
+
+---
+
+## LaTeX Skill Pattern Reference
+
+The "LaTeX Handouts" Claude Code skill (mcpmarket.com) demonstrates the pattern:
+- Skill reads source content (slides/markdown)
+- Expands bullets into research-backed prose
+- Generates structured professional document
+- Outputs as standalone file
+
+MindrianOS follows the same pattern but uses HTML→PDF (WeasyPrint) instead of LaTeX→PDF,
+keeping zero dependencies while achieving the same professional output quality.
+
+---
+
+## Font Bundling Note
+
+Bebas Neue is a free Google Font (SIL Open Font License). Bundle the .ttf in the plugin's
+assets/ directory (~50KB). WeasyPrint loads it via @font-face in the CSS template.
+No system font installation required.

package/references/hsi/HSI-TOOLS-REFERENCE.md

@@ -0,0 +1,222 @@
+# HSI & Reverse Salient Discovery — Tool Reference
+
+*Computational engines for auto-relationship building in the Data Room knowledge graph.*
+
+---
+
+## What These Tools Do for MindrianOS
+
+The HSI (Hybrid Similarity Index) and Reverse Salient Discovery framework provide the **computational backbone for auto-discovering cross-relationships** in the Data Room. Instead of relying only on keyword matching or explicit cross-references, these tools find HIDDEN connections between room artifacts using dual similarity analysis.
+
+### The Core Insight
+
+When two room artifacts share **structural similarity** (same methods/terms) but LOW **semantic similarity** (different meanings/applications), OR vice versa — that GAP is where hidden innovation connections live.
+
+```
+HSI(artifact_i, artifact_j) = |BERT_similarity - LSA_similarity| × Integrative_Factor
+
+High HSI = "These two artifacts are connected in ways nobody sees"
+```
+
+### Application to Data Room
+
+| Source Tool | Data Room Application |
+|-------------|----------------------|
+| LSA (TF-IDF + SVD) | Structural similarity between room artifacts (shared terminology) |
+| BERT embeddings | Semantic similarity between room artifacts (shared meaning) |
+| HSI differential | Hidden connections: artifacts that share structure but not meaning, or meaning but not structure |
+| Reverse Salient detection | Cross-subsystem innovation opportunities — where a solution in one room section addresses a problem in another |
+| Top pairs extraction | The most surprising connections to surface to the user |
+
+### Integration with Meeting Filing
+
+After meeting segments are filed to room sections, the HSI pipeline runs:
+1. Compute LSA similarity between ALL room artifacts (including new meeting segments)
+2. Compute BERT/embedding similarity between all artifacts
+3. Find high-differential pairs (|BERT - LSA| > threshold)
+4. Surface as ENABLES/INFORMS/CONTRADICTS edges in the knowledge graph
+5. Larry says: "I found a hidden connection between your mentor's advice and your market analysis that nobody mentioned in the meeting."
+
+---
+
+## Tool 1: HSI Computational Pipeline
+
+### Architecture
+
+```
+Room artifacts (text) → TF-IDF vectorization → TruncatedSVD (80 components) → LSA similarity matrix
+Room artifacts (text) → BERT tokenization → segment embeddings → BERT similarity matrix
+|BERT - LSA| = differential matrix → HSI score × integrative factor → top pairs = hidden connections
+```
+
+### Key Parameters
+
+| Parameter | Value | Purpose |
+|-----------|-------|---------|
+| max_features | 2000 | TF-IDF vocabulary size |
+| n_components | 80 | SVD dimensions (topic space) |
+| max_df | 0.5 | Ignore terms in >50% of docs |
+| BERT model | bert-large-cased | Semantic embeddings |
+| max_tokens | 512 | BERT segment size |
+| threshold | 0.3 | Minimum differential for reverse salient |
+
+### Adaptation for Plugin
+
+**Problem:** BERT-large requires ~1.3GB memory + GPU. Plugin must work on CPU with low memory.
+
+**Solution (already proven):** Gemini-Native HSI — replace BERT with Claude/Gemini embedding API:
+- `text-embedding-004` or Claude's native embeddings
+- API-based, no local GPU needed
+- Minutes instead of hours
+- Fits within plugin constraints
+
+**Alternative:** Use Pinecone (already integrated as Brain MCP) for embedding similarity. Room artifacts get embedded at filing time. Cross-similarity is a Pinecone query.
+
+### LSA Component (lsa.py adapted)
+
+```python
+# Adapted for room artifacts instead of papers
+from sklearn.feature_extraction.text import TfidfVectorizer
+from sklearn.decomposition import TruncatedSVD
+
+def compute_lsa_similarity(artifacts: list[str]) -> np.ndarray:
+    vectorizer = TfidfVectorizer(stop_words='english', max_features=2000, max_df=0.5)
+    X = vectorizer.fit_transform(artifacts)
+    svd = TruncatedSVD(n_components=min(80, len(artifacts)-1), random_state=256)
+    Z = svd.fit_transform(X)
+    sim = cosine_similarity(Z)
+    # Normalize to [0,1]
+    sim = (sim - sim.min()) / (sim.max() - sim.min() + 1e-12)
+    return sim
+```
+
+### Semantic Component (adapted for plugin)
+
+```python
+# Option A: Pinecone embeddings (if Brain connected)
+def compute_semantic_similarity_pinecone(artifact_ids: list[str]) -> np.ndarray:
+    # Query Pinecone for existing embeddings
+    # Compute pairwise cosine similarity
+    pass
+
+# Option B: Local lightweight embeddings (Tier 0)
+def compute_semantic_similarity_local(artifacts: list[str]) -> np.ndarray:
+    from sentence_transformers import SentenceTransformer
+    model = SentenceTransformer('all-MiniLM-L6-v2')  # 80MB, CPU-friendly
+    embeddings = model.encode(artifacts)
+    sim = cosine_similarity(embeddings)
+    sim = (sim - sim.min()) / (sim.max() - sim.min() + 1e-12)
+    return sim
+```
+
+### HSI Computation
+
+```python
+def compute_hsi(artifacts: list[str]) -> dict:
+    lsa = compute_lsa_similarity(artifacts)
+    semantic = compute_semantic_similarity_local(artifacts)  # or Pinecone
+
+    # Differential = hidden connections
+    gap = np.abs(semantic - lsa)
+    gap = (gap - gap.min()) / (gap.max() - gap.min() + 1e-12)
+
+    return {"lsa": lsa, "semantic": semantic, "hsi": gap}
+
+def top_hidden_connections(hsi: np.ndarray, artifacts: list, k=10):
+    n = hsi.shape[0]
+    pairs = []
+    for i in range(n):
+        for j in range(i+1, n):
+            if hsi[i,j] > 0.3:  # threshold
+                pairs.append((i, j, float(hsi[i,j])))
+    pairs.sort(key=lambda x: x[2], reverse=True)
+    return pairs[:k]
+```
+
+---
+
+## Tool 2: Reverse Salient Discovery Framework
+
+### The 4-Phase Pipeline
+
+```
+Phase 1: ACQUIRE     — gather all room artifacts as "documents"
+Phase 2: ANALYZE     — compute dual similarity (LSA + semantic)
+Phase 3: DETECT      — find high-differential pairs (reverse salients)
+Phase 4: SYNTHESIZE  — generate innovation thesis for each pair
+```
+
+### Adaptation for Data Room Auto-Relationships
+
+Instead of analyzing academic papers across domains, we analyze ROOM ARTIFACTS across sections:
+
+| Framework Concept | Data Room Equivalent |
+|-------------------|---------------------|
+| Domain A papers | Artifacts in section A (e.g., market-analysis) |
+| Domain B papers | Artifacts in section B (e.g., solution-design) |
+| Cross-domain reverse salient | Hidden connection between sections |
+| Innovation thesis | Larry's insight: "This market trend enables that technical approach" |
+
+### Edge Types Generated
+
+| HSI Pattern | Edge Type | Example |
+|-------------|-----------|---------|
+| High structural, low semantic | ENABLES | "Same method, different application — this technique in solution-design could address the gap in market-analysis" |
+| Low structural, high semantic | INFORMS | "Different words, same concept — the mentor's advice about timing maps to the S-curve analysis" |
+| High differential + contradiction | CONTRADICTS | "These two artifacts say opposite things using different language" |
+| Multiple artifacts → same concept | CONVERGES | "Three meetings, three speakers, all pointing to the same insight from different angles" |
+| Change in one → invalidates other | INVALIDATES | "This new finding makes the assumption in financial-model stale" |
+
+### Neo4j Storage (when Brain connected)
+
+```cypher
+// Store discovered relationships
+MERGE (a1:RoomArtifact {id: $artifact1_id})
+MERGE (a2:RoomArtifact {id: $artifact2_id})
+CREATE (a1)-[:HSI_CONNECTION {
+  hsi_score: $score,
+  lsa_sim: $lsa,
+  semantic_sim: $semantic,
+  connection_type: $type,
+  discovered: datetime(),
+  human_approved: null
+}]->(a2)
+```
+
+---
+
+## Integration with Phase 6 (Cross-Relationship Discovery Loop)
+
+Plan 06-03 builds the cross-relationship discovery loop. The HSI tools integrate as:
+
+```
+Artifact filed (from meeting or methodology session)
+    ↓
+Step 6: Cross-relationship scan
+    ↓
+Option A (Tier 0, no deps): Keyword-based (existing analyze-room)
+Option B (Tier 1, lightweight): LSA + MiniLM embeddings (this reference)
+Option C (Tier 2, Brain): LSA + Pinecone + Neo4j storage
+    ↓
+Hidden connections surfaced → Larry presents → User APPROVE/REJECT/DEFER
+    ↓
+Graph edges created → Dashboard updated → Knowledge compounds
+```
+
+---
+
+## Source Files
+
+Original HSI codebase (Jonathan's research):
+- `lsa.py` — LSA similarity computation (TF-IDF + TruncatedSVD)
+- `sts_bert.py` — BERT semantic similarity (segment-aware cosine)
+- `comparison.py` — Differential analysis + reverse salient extraction
+- `paper_gathering.py` — Scopus API document acquisition
+- `paper_cleaning.py` — Text preprocessing
+- `lsa_bert_similarity.py` — Combined LSA + BERT pipeline
+- `hybrid_index.py` — HSI computation with integrative scoring
+
+Enhanced framework:
+- Reverse Salient Discovery Agent (4-phase pipeline)
+- Neo4j integration schema for storing/querying discoveries
+- Dynamic threshold adjustment based on domain characteristics