llm-wiki-choose

LLM-Wiki Choose

Goal

Interview the user enough to recommend a practical path: adopt a ready-made solution, assemble a local-first workflow, roll out a team process, or build a custom LLM-Wiki.

When to use

• The user asks which LLM-Wiki approach fits: a ready-made tool, a local-first git workflow, a team rollout, a retrieval/GraphRAG system, or a custom build.
• The user explicitly asks whether OpenWiki, Obsidian, qmd, or a GraphRAG system suits their case.
• The user is comparing infrastructure options before committing to an architecture or before starting llm-wiki-paf-adoption.
• The user is still skeptical whether an LLM-Wiki is worth building at all.

Inputs

• User type (individual, team, company, product builder) and corpus type/scale.
• Privacy tier (public, internal, sensitive, regulated) and update/freshness needs.
• Preferred coding agent and existing tooling (Obsidian, GitHub, qmd, vector DB, GraphRAG, OpenWiki, custom scripts).
• Write tolerance (read-only advisory, PR-based writes, direct writes) and infrastructure/maintenance budget.

Procedure

1. Gather requirements

Ask only the questions that materially change the recommendation. Cover:

• user type: individual, team, company, product builder;
• corpus: code repo, research papers, docs folder, Obsidian vault, chat exports, web clips, PDFs, media transcripts;
• scale: rough number of sources/pages;
• update frequency and freshness needs;
• privacy: public, internal, sensitive, regulated;
• preferred agent: Claude Code, Codex, Cursor, OpenCode, GitHub Copilot, other;
• local-first and git requirements;
• write tolerance: read-only advisory, PR-based writes, direct writes;
• existing tools: Obsidian, GitHub, qmd, vector DB, GraphRAG, OpenWiki, custom scripts;
• budget for infrastructure and maintenance.

If the user already provided these facts, do not ask again.

2. Classify the case

Use this routing:

Case	Default recommendation
Small personal vault, local-first	Markdown + git + agent skills + rg; add qmd later.
Code repository docs	OpenWiki-style repo wiki or custom docs-as-code workflow.
Existing Obsidian vault	Refactor in place with protected human sections and lint gates.
Team/company knowledge	PR-based agent writes, CODEOWNERS, permissions, lint reports.
Large retrieval-heavy corpus	Hybrid search or graph/RAG system, with wiki as compiled surface if human review matters.
Product/plugin idea	Custom architecture with provenance, staged review and safe writes from day one.