SecondBrain + 2ndbrain Wiki Architecture

How the Obsidian vault, structured wiki, and PKA operational vault work together · MC-611 · 2026-04-15

SecondBrain

~/workspace/cowork/SecondBrain/

Personal Obsidian vault. Browsable on desktop + mobile. Human-optimized.

Folders

Meetings/ 34+ dated notes
people/ 43+ contact notes
projects/ 10+ active projects
tasks/ action items + due dates
daily/ YYYY-MM-DD daily notes
ideas/ captured thoughts
inbox/ raw capture, awaiting sort
docs/ reference material

Note Format

--- type: task|person|meeting status: active|done|pending tags: [topic, area] due: YYYY-MM-DD --- Body with [[wiki-links]]

Obsidian Features

Graph View (visual connections)
Backlinks (auto-generated)
Daily Notes
Mobile Sync ($5/month)
Obsidian CLI access

2ndbrain Wiki

~/.claude/vault/wiki/

Structured synthesis layer. AI-optimized indexing and cross-referencing.

sources/

Source summaries from ingested documents (reports, articles, transcripts)

entities/

Named entities (people, orgs, products) cross-referenced across sources

concepts/

Frameworks, mental models, principles linked to source evidence

synthesis/

Combined insights across multiple sources

How It's Built

/wiki-ingest processes documents
Extracts entities + concepts
Cross-links all pages
Updates indexes automatically
Shared via conrelma/claude repo

PKA Vault (Operational)

~/workspace/PKA/Vault/

Operational knowledge graph. Shared between Lucienne and Luci via git sync.

Contents

vault.db SQLite + FTS5 index
memory/ durable reference knowledge
tasks/ YAML frontmatter tasks
projects/ project tracking
notes/ general notes
schema.sql DB definition

vault.db Tables

files — indexed markdown files
edges — knowledge graph (197 edges)
search_fts — full-text search (FTS5)
activity_log — append-only audit
sessions_index — transcript search

Key Tools

index.py — scans markdown, builds DB
vault_mcp.py — 8 MCP query tools
session_indexer.py — transcript index

Skills (API Layer)

Never edit files directly — skills are the interface between human and vaults.

/brain Capture & auto-classify input into SecondBrain

/search-brain Full-text search across SecondBrain vault

/wiki-query Cross-wiki search + synthesized answers

/wiki-ingest Document → source + entity + concept pages

/wiki-lint Health checks: broken links, orphans, stale pages

Actors & Transport

Who reads, who writes, and how data moves between machines.

Elmar (Human)

Browses SecondBrain via Obsidian desktop + mobile app (Obsidian Sync). Sends tasks to Luci via Telegram.

Lucienne (Local Mac)

Chief of Staff. Owns SecondBrain + vault.db. Runs /brain, /wiki-ingest, /search-brain. Assigns work to Luci via PKA repo.

Luci (Cloud Server)

Reads PKA vault via git pull every 15 min (read-only). Queries vault.db through MCP server. Owns mc.db for operations.

Git Repos (Transport)

conrelma/cowork → SecondBrain vault
conrelma/PKA → vault.db, memory/
conrelma/claude → portable vault, skills

Data Flows

How information moves through the system.

Capture

User input → /brain skill → classifies as task/person/idea/meeting → writes markdown to SecondBrain/ with [[wiki-links]] → Obsidian detects change and updates graph.

SecondBrain/{type}/ → .md with YAML frontmatter

/search-brain searches SecondBrain only. /wiki-query searches both wikis (SecondBrain + 2ndbrain) and synthesizes an answer across all matches.

Grep + file reading → matched excerpts + source paths

Ingest

Document or URL → /wiki-ingest → creates source summary in sources/, extracts entities/ and concepts/ pages → cross-links all pages → updates indexes.

wiki/{sources,entities,concepts}/ → .md with wikilinks

Index & Sync

index.py scans all markdown → populates vault.db (FTS5 full-text search). Git syncs every 15 min. Luci pulls vault.db read-only. /wiki-lint checks health of both wikis.

index.py --rebuild → vault.db (files, edges, search_fts)