Knowledge Map — Trenton Richards

A live, self-updating visualization of my knowledge state across every topic, skill, and concept I've worked through with Claude. Hosted at knowledge-map.trentbrichards.com.

---

What This Is

Every Claude Code session ends with an update to this system. Claude parses what was covered, what was new vs. reinforced, and writes to two files:

• data/knowledge-map-public.json — sanitized skill/tool/concept graph (public, powers the website)
• ~/Documents/my-vault/Notes/knowledge-map-private.md — full context including philosophies, gaps, session trails (private, never leaves the vault)

The website renders the public graph as an interactive D3.js force-directed visualization with a timeline scrubber — drag it and watch the knowledge graph grow from day one.

---

Architecture

Session End (every Claude session)
  │
  ├─→ knowledge-map-private.md (vault — full brain, private)
  │     - philosophies, goals, mental models
  │     - knowledge gaps
  │     - session summaries with full context
  │     - company + career targeting logic
  │
  └─→ data/knowledge-map-public.json (this repo — sanitized, public)
        - skills, tools, concepts, projects
        - proficiency scores (1-10 internal)
        - first_seen + history for timeline
        - connections between nodes
              ↓
        Next.js app at knowledge-map.trentbrichards.com
        D3.js force-directed graph
        ├── node size + glow = proficiency score (1-10)
        ├── node label = Novice / Working / Fluent
        ├── color = domain (AI/ML, Trading, Web, etc.)
        ├── hover = proficiency label + session count
        ├── click = related nodes + connected projects
        └── timeline scrubber = drag to see graph at any past date

---

Data Schema

Node types

Type	Examples
skill	Python, TypeScript, D3.js, SQL
tool	Ollama, Alpaca, Vercel, Claude Code
concept	RAG, multi-agent, bracket orders, OAuth
domain	AI/ML Engineering, Trading, Web Dev
project	AutoTrader, Clipper, Portfolio, Knowledge Map

Node schema

{
  "id": "python",
  "label": "Python",
  "type": "skill",
  "domain": "engineering",
  "score": 8,
  "proficiency": "fluent",
  "first_seen": "2026-03-28",
  "last_updated": "2026-04-02",
  "session_count": 12,
  "history": [
    { "date": "2026-03-28", "score": 6 },
    { "date": "2026-04-02", "score": 8 }
  ],
  "connections": ["autotrader", "clipper", "job-search-pipeline", "ollama"]
}

Proficiency scale

1-3   Novice   — aware of it, used it once or twice
4-6   Working  — can use it independently, needs docs occasionally
7-10  Fluent   — default choice, deep understanding, can debug

The score drives visual weight (node size, glow, edge thickness). The label shown publicly is always Novice/Working/Fluent. Score stays internal.

Domain color map

Domain	Color
AI/ML Engineering	#d3968c (salmon — matches portfolio accent)
Trading & Finance	#839958 (olive)
Web Development	#105666 (teal)
Video & Automation	#7a6e9c (purple)
Strategy & Business	#c4a35a (gold)
Systems & DevOps	#6e8c7a (muted green)

---

Visualization Features

Force-directed graph

• Nodes repel each other naturally, edges pull connected nodes together
• Skills cluster around the domains + projects that use them
• Highly connected nodes (Python, Claude Code) gravitate toward center
• Orphan nodes (new skills with no connections yet) float at edges

Timeline scrubber

• Horizontal slider at bottom: left = first session, right = today
• Drag left → nodes learned after that date fade out and disappear
• Drag right → nodes animate in as they were first introduced
• Shows the actual growth trajectory, not just current state

Interaction

• Hover: show label, proficiency (Novice/Working/Fluent), session count
• Click: highlight connected nodes, dim everything else, show project links
• Double-click a project node: see all skills it required
• Filter buttons by domain (show only AI/ML, only Web Dev, etc.)

---

Tech Stack

Layer	Choice	Why
Visualization	D3.js v7	Force simulation, timeline scrubber, full control
Framework	Next.js (App Router)	Consistent with portfolio, easy Vercel deploy
Styling	Tailwind + portfolio design tokens	Matches trentbrichards.com aesthetic
Data	Static JSON (this repo)	No backend needed — Claude writes it at session end
Hosting	Vercel	Same account as portfolio
Domain	Cloudflare subdomain (knowledge-map.trentbrichards.com)	CNAME → Vercel

---

Session-End Protocol

At the end of every Claude Code session, Claude:

1. Reviews what was covered in the session
2. Identifies: new nodes to add, existing nodes to score up, new connections to draw
3. Writes to data/knowledge-map-public.json:
   • New nodes with first_seen = today, initial score
   • Updated score, last_updated, session_count, history entry for existing nodes
   • New connections between nodes that were used together
4. Writes a summary to ~/Documents/my-vault/Notes/knowledge-map-private.md:
   • Full session context, concepts introduced, mental models built
   • Knowledge gaps surfaced (what wasn't known, what was explained)
   • Private notes (not in public JSON)

This means the public graph updates every session automatically, with no manual intervention.

---

Deployment Plan

1. ✅ Seed initial knowledge-map-public.json with current knowledge state
2. Build Next.js app with D3.js graph (web/)
3. Deploy to Vercel as separate project
4. Add CNAME record in Cloudflare: knowledge-map → Vercel deployment URL
5. Add custom domain in Vercel project settings
6. Link from portfolio Projects section

---

File Structure

~/projects/knowledge-map/
├── README.md                          ← this file
├── data/
│   └── knowledge-map-public.json      ← Claude updates this every session
├── web/                               ← Next.js app (to be built)
│   ├── src/
│   │   ├── app/
│   │   │   └── page.tsx
│   │   └── components/
│   │       ├── KnowledgeGraph.tsx
│   │       └── TimelineScrubber.tsx
│   └── package.json
└── scripts/
    └── seed.py                        ← one-time seed helper (optional)

---

Privacy Model

Public (website): Skills, tools, concepts, projects, proficiency levels, growth timeline. What Trenton knows. Nothing personal.

Private (vault only): Session content, philosophy, goals, mental models, knowledge gaps, company targeting strategy, personal context. How Trenton thinks. Never leaves ~/Documents/my-vault/.

---

Status

Item	Status
Project structure	✅ Created
Initial data seed	✅ Done (2026-04-02)
Next.js app	🔴 Not started
D3.js graph component	🔴 Not started
Timeline scrubber	🔴 Not started
Vercel deployment	🔴 Not started
Cloudflare subdomain	🔴 Not started