Command cheatsheet

Everything you need on one page. Slash commands work inside Claude Code / Codex CLI; CLI commands run at your terminal.

30-second setup

From Claude Code, run these slash commands in order:

/wiki-init — scaffold raw/ wiki/ site/
/wiki-sync — ingest sessions from every auto-detected agent
/wiki-graph — build the AI knowledge graph
/wiki-build — compile the static site
/wiki-serve — browse at http://127.0.0.1:8765/

Daily flow

What you want	Slash command	CLI equivalent
Convert new session transcripts	`/wiki-sync`	`llmwiki sync`
Ingest a source file into the wiki	`/wiki-ingest <path>`	--
Ask the wiki a question	`/wiki-query <question>`	--
Edit one page surgically	`/wiki-update <page>`	--
Find orphans + broken links	`/wiki-lint`	`llmwiki lint`
Triage candidate pages	--	`llmwiki candidates list`
Build / rebuild the site	`/wiki-build`	`llmwiki build`
Serve locally	`/wiki-serve`	`llmwiki serve --open`
Interactive graph	`/wiki-graph`	`llmwiki graph`
AI knowledge graph	`/wiki-graph`	`llmwiki graph --engine graphify`
Self-reflection on wiki gaps	`/wiki-reflect`	--

12 CLI commands

Command	Purpose
`init`	Scaffold `raw/` `wiki/` `site/` + seed 9 nav files
`sync`	Convert `.jsonl` sessions -> markdown -> wiki -> site
`build`	Compile `wiki/` markdown into `site/` HTML
`serve`	Start local HTTP server (default `:8765`)
`adapters`	List every adapter + its status
`graph`	Build the knowledge graph (Graphify default, builtin fallback)
`query`	Search the knowledge graph with a question
`export`	AI-consumable exports (llms-txt, jsonld, sitemap, rss, marp, ...)
`lint`	Run 14 wiki-quality rules
`candidates`	Approval workflow (list / promote / merge / discard)
`synthesize`	LLM-backed source-page synthesis with auto-tagging
`version`	Print version

Knowledge graph

llmwiki graph                          # builtin wikilink graph (stdlib, zero deps)
llmwiki graph --engine graphify        # AI-powered: Leiden communities, confidence edges, god nodes
llmwiki graph --format json            # json only
llmwiki graph --format html            # interactive HTML only

Install Graphify: pip install llm-notebook[graph]

Graphify outputs to graphify-out/: graph.json, graph.html, GRAPH_REPORT.md. Features: tree-sitter AST extraction, semantic analysis, community detection, confidence-scored edges.

Exports

llmwiki export all                     # all 7 formats at once
llmwiki export llms-txt                # llmstxt.org spec
llmwiki export llms-full-txt           # full plain-text corpus (<= 5 MB)
llmwiki export jsonld                  # schema.org entity graph
llmwiki export sitemap                 # sitemap.xml
llmwiki export rss                     # rss.xml
llmwiki export robots                  # robots.txt
llmwiki export ai-readme               # ai-readme.md

Quality

llmwiki lint                           # 14 rules (8 structural + 3 LLM + 2 v1.1)
llmwiki lint --json --fail-on-errors   # CI-friendly
llmwiki lint --rules link_integrity,orphan_detection
llmwiki lint --include-llm             # + contradiction/claim/summary checks

Candidate workflow

llmwiki candidates list                # show all candidates
llmwiki candidates list --stale        # only stale (>30 days)
llmwiki candidates promote --slug Foo  # promote to wiki
llmwiki candidates merge --slug A --into B
llmwiki candidates discard --slug X --reason "hallucinated"

LLM synthesis

llmwiki synthesize                     # synthesize source pages
llmwiki synthesize --check             # probe backend (exit 0 if ok)
llmwiki synthesize --estimate          # cost preview, no API calls
llmwiki synthesize --dry-run           # list what would be processed
llmwiki synthesize --force             # re-synth everything

Auto-tags pages (up to 5 AI tags per page, near-dup rejection, stop-word filter).

Adapters

llmwiki adapters                       # list every adapter + who fires on next sync
llmwiki adapters --wide                # untruncated descriptions

Core (auto-discovered, always loaded):

Adapter	Source
`claude_code`	`~/.claude/projects/`
`codex_cli`	`~/.codex/sessions/`

Contrib (load on-demand with --adapter <name>):

Adapter	Source
`chatgpt`	`conversations.json` export
`copilot_chat`	VS Code workspaceStorage
`copilot_cli`	GitHub Copilot CLI
`cursor`	VS Code workspaceStorage
`gemini_cli`	`~/.gemini/`
`obsidian`	Obsidian vault `.md` files
`opencode`	OpenCode / OpenClaw sessions

Obsidian integration

# Sync wiki into an Obsidian vault
llmwiki sync --vault "~/Documents/Obsidian Vault/my-wiki"

# Build site from a vault
llmwiki build --vault "~/Documents/Obsidian Vault/my-wiki"

# Use the obsidian adapter to read from a vault
llmwiki sync --adapter obsidian

Flags you'll actually use

Flag	Command	What
`--since YYYY-MM-DD`	`sync`	Only sessions after that date
`--project <slug>`	`sync`	Restrict to one project
`--force`	`sync`, `synthesize`	Ignore state file, reconvert everything
`--fail-on-errors`	`lint`	Non-zero exit on error-severity issues
`--vault <path>`	`sync`, `build`	Operate on an Obsidian / Logseq vault
`--engine graphify`	`graph`	AI-powered knowledge graph
`--host 0.0.0.0`	`serve`	Bind LAN-accessible (default: loopback-only)
`--status`	`sync`	Show last sync + per-adapter counters

Config files

File	Purpose
`sessions_config.json`	All settings (see below)
`.llmwikiignore`	Exclude patterns (git-ignore format)
`.llmwiki-state.json`	Per-source mtime cache (auto; gitignored)
`.llmwiki-quarantine.json`	Convert failures (auto; gitignored)
`.env`	`ANTHROPIC_API_KEY` for API-mode synth

All configurable settings (sessions_config.json)

Section	Key	Default	What
`vault`	`default_path`	`""` (must pass `--vault`)	Default Obsidian/Logseq vault path
`vault.layout`	`entities`	`Wiki/Entities`	Vault subfolder for entity pages
`vault.layout`	`concepts`	`Wiki/Concepts`	Vault subfolder for concept pages
`vault.layout`	`sources`	`Wiki/Sources`	Vault subfolder for source pages
`vault.layout`	`syntheses`	`Wiki/Syntheses`	Vault subfolder for synthesis pages
`vault`	`allow_overwrite`	`false`	Allow clobbering existing vault pages
`graph`	`default_engine`	`builtin`	Graph engine: `builtin` or `graphify`
`graph`	`format`	`both`	Graph output: `json`, `html`, or `both`
`serve`	`port`	`8765`	Dev server port
`serve`	`host`	`127.0.0.1`	Dev server bind address
`serve`	`open_browser`	`false`	Auto-open browser on serve
`build`	`out_dir`	`site`	Build output directory
`build`	`search_mode`	`auto`	Search index mode: `auto`, `tree`, `flat`
`build`	`synthesize`	`false`	Auto-synthesize overview on build
`schedule`	`build`	`on-sync`	When to auto-build: `on-sync`, `daily`, `manual`
`schedule`	`lint`	`manual`	When to auto-lint: `on-sync`, `daily`, `manual`
`synthesis`	`backend`	`dummy`	LLM backend: `dummy`, `ollama`
`synthesis`	`model`	`llama3.1:8b`	Model for synthesis
`synthesis`	`base_url`	`http://127.0.0.1:11434`	Ollama server URL
`synthesis`	`timeout`	`60`	Synthesis timeout (seconds)
`filters`	`live_session_minutes`	`60`	Skip sessions younger than N minutes
`truncation`	`tool_result_chars`	`500`	Max chars for tool results in output

Three-layer architecture

raw/     IMMUTABLE transcripts (source of truth, never modify)
wiki/    LLM-generated pages (you own this)
  sources/      one summary per raw source
  entities/     people, projects, tools (TitleCase.md)
  concepts/     ideas, patterns, decisions (TitleCase.md)
  syntheses/    saved query answers
  comparisons/  side-by-side diffs
  questions/    first-class open questions
site/    GENERATED static HTML (don't edit by hand)

Common recipes

# Daily: sync + serve
llmwiki sync && llmwiki serve --open

# Nightly cron (one project)
llmwiki sync --project my-project --no-auto-lint --since $(date -v-1d +%Y-%m-%d)

# AI knowledge graph
pip install llm-notebook[graph]
llmwiki graph --engine graphify

# CI quality gate
llmwiki lint --json --fail-on-errors

# Export wiki to Obsidian vault
llmwiki sync --vault "~/Documents/Obsidian Vault/my-wiki"

# Full site rebuild with AI synthesis
llmwiki build --synthesize && llmwiki export all

Exit codes

Code	Meaning
`0`	Success
`1`	Operation failed
`2`	Usage error (bad flags)