Overview
RonanCodes/llm-wiki is a minimal boilerplate combining Claude Code + Obsidian — the recommended starting point for users who “can edit a text file.” It is a Tier 4 starter template in the comprehensive report.
Key Facts
- Form: minimal boilerplate (Claude Code + Obsidian)
- Audience: beginners who can edit a text file, then add MCP when index.md breaks
- Role: lowest-friction on-ramp to the pattern
- Stack: Claude Code + Obsidian + markdown
Features
Core Architecture
- Minimal boilerplate — just Claude Code + Obsidian + markdown, nothing else
- No dependencies — zero npm packages, zero databases, zero APIs
- File-based — plain markdown files
Ingest & Processing
- Agent-driven ingest — Claude Code processes sources into wiki pages
- Manual drop — user adds sources to raw/ folder
Quality & Governance
- Obsidian graph view — visual navigation of wikilinks
- Manual lint — user or agent checks page health
Integration
- Claude Code native — designed for Claude Code skill system
- Obsidian compatible — wiki works as Obsidian vault from day one
Data & Storage
- Markdown files — standard wiki output
- Git-backed — version history via git
Ralph Loop Connection
RonanCodes/llm-wiki documentation explicitly states that the ralph-loop pattern was used to build the LLM Wiki implementation itself:
The Ralph Loop is used in two ways:
- Build methodology — LLM Wiki itself was built using Ralph loops (
/ralphskill,prd.json,progress.txt)- Wiki operations — ingest/lint could run as a Ralph-style loop: process sources one at a time, each in a fresh context, progress tracked in
log.md
This means the RonanCodes implementation already embodies the exact pattern this vault documents: a three-layer architecture (raw sources, wiki pages, schema) with a cyclic ingest-query-lint workflow. The build methodology used Ralph loops, and the wiki operations could continue running as Ralph-style loops. This vault has a ralph-converter skill installed but has never used or documented that connection.
Implications
The RonanCodes documentation provides the only explicit bridge found so far between Ralph loops and LLM Wiki operations. It validates this vault’s own agent-loop pattern and suggests the ingest-query-lint cycle could itself be run as a Ralph-style loop with fresh context per source.
Open Questions
- Could this vault’s ingest pipeline benefit from running as a Ralph loop (one source per iteration, lint-verify, log-track)?
- Should the orphaned
ralph-converterskill be adopted and integrated, given RonanCodes’ documented precedent?
Relationships
- llm-wiki — (type: belongs-to) starter template
- claude-code — (type: operates-with) the agent it wires up
- obsidian — (type: operates-with) the viewer it pairs with
- llm-wiki-mcp — (type: extends) the graduate step when the wiki grows
- ralph-loop — (type: used) built the implementation itself via Ralph loops, per its own docs
Implications
RonanCodes is essentially what this vault already is (Claude Code + Obsidian + markdown), packaged as a template. Useful as a reference for the “minimal” end of the spectrum.
External Links
- Repository URL unavailable in the current freshness check; the implementation summary is preserved in
raw/articles/llm-wiki-okf-comprehensive-report-2026.md. - https://ronancodes.github.io/llm-wiki/docs/research/ralph-loop
Recent Updates (2026-09-05)
As of 2026-09-05, the repository reports 0 stars and 0 forks; no tagged release; last push 2026-04-18; 0 open issues. Recent commits: feat: add video-copy.md and video-script.json (2026-04-18), refactor: auto-discover projects via register.tsx (2026-04-18), feat: KineticPitch polish (hype cycle terms, (2026-04-15). Note: the repository was renamed from RonanCodes/llm-wiki to RonanCodes/remotion-studio-llm-wiki.
Sources
^[raw/articles/llm-wiki-okf-comprehensive-report-2026.md]