Battlecard — nashsu/llm_wiki (15.7k★)

Battlecard — nashsu/llm_wiki

Entities Compared

• nashsu-llm-wiki — most popular LLM Wiki repo (~15.7k★, LGPL v3, Tauri v2 desktop)
• open-knowledge-format — the standard nashsu does not use

One-Line Positioning

nashsu is the most feature-rich desktop LLM Wiki on GitHub, but it locks your brain into a proprietary format behind a GUI. This vault is the open, deterministic, non-technical-owner alternative: OKF-compliant, lint-gated, auditably captured, and yours in plain markdown.

Where nashsu wins

1. Product surface — polished Tauri v2 desktop app (Windows/Mac/Linux), while we ship a terminal + dashboard script.
2. Automatic clustering — Louvain community detection + 4-signal relevance model out of the box; our graph analysis requires running graph-export.
3. Browser capture loop — their user adds a page, wiki absorbs it; our external_links.py capture is a manual CLI step.
4. Rich features — Mermaid rendering, Deep Research, Rust chat agent, Agent Skills integration, MCP server.
5. Network effect / discoverability — 15.7k★ and growing means tutorials, issues, and mindshare we cannot match.

Where we win

1. Openness & portability — OKF v0.2 frontmatter + plain markdown; nashsu uses a proprietary format (open-knowledge-format contradiction). Your knowledge survives any tool.
2. Governance & audit — wikilint gated ingest, provenance on every source (sources: + content-hash + status), log.md action trail. nashsu has no such trust layer.
3. External-link evidence — URLs become immutable, hash-tracked snapshots (scripts/external_links.py); link rot is detected, not ignored.
4. Deterministic routing — main/prompt wiki classification before every query and ingest; a decision engine, not just a chat.
5. Ownership model — LGPL v3 vs our MIT-ish freedom; and we are built for a non-technical owner who does not want to maintain a Rust/Tauri codebase.
6. Multi-wiki discipline — no competitor routes between two governed wikis.

Objection Handling

Objection	Response
”nashsu is more popular, so it’s better”	Popularity ≠ longevity. nashsu’s proprietary format is a migration trap; our plain-markdown OKF pages survive any app.
”I need a desktop GUI”	Valid gap — roadmap is dashboard.ps1 → guided no-terminal flow. Not there yet; be honest.
”I want one-click browser capture”	external_links.py capture <url> is the primitive; wiring it into the dashboard is a planned step.
”nashsu clusters my notes automatically”	We can reproduce the insight (Louvain + bridge detection) on top of graph-export, without the lock-in.
”Why should I trust your wiki tool?”	Because every page cites its source, every change is logged, and nothing is auto-created by regex — an LLM decides, you approve.

Proof Points (for the vault owner)

• Vault runs on plain markdown + OKF frontmatter — open any file in any editor.
• scripts/route-classifier.ps1 decides which of two wikis a query/ingest belongs to before any pipeline runs.
• Every external URL cited in a page is either captured to raw/external/ or flagged as a reference, never silently presented as evidence (Guardrail #16).
• wikilint.ps1 gates commits with scoped, deterministic checks.

Implications

The gap to close is product surface, not architecture: dashboard UX, browser capture, and one-click onboarding turn the discipline we already have into something a non-technical owner can live in. See competitive-brief-llm-wiki-ecosystem for the ranked roadmap.

Sources

^[wiki/entities/nashsu-llm-wiki.md]