Commands

Quick Start

# Step 1: Configure your LLM provider
wakawiki --init

# Step 2: Generate documentation (interactive)
wakawiki

# Or: One-shot non-interactive
wakawiki -p "Summarize the architecture of this project"

# Or: Update existing documentation
wakawiki --update

CLI Reference

wakawiki [OPTIONS] [PROMPT]

Option	Description
--init	Interactive setup: choose provider, set API key, pick model
-p, --print	Non-interactive one-shot mode (CI-friendly)
--scan	Heuristic scan (no LLM) — fast, deterministic docs
--index	Build structured JSON index of all symbols and files
--embed	Generate embeddings for semantic search
-q, --query	Query the index for symbols or files matching a pattern
-s, --semantic	Semantic search query (requires embeddings)
--serve	Start MCP server (JSON-RPC 2.0 over stdio)
--update	Refresh existing wakawiki/ docs with incremental diff
-h, --help	Show help
--version	Print version and exit

Interactive Mode

Running wakawiki without -p starts an interactive chat session with the LLM agent. You can guide the documentation process by asking questions or giving instructions.

wakawiki
# > Please focus on the API layer
# > Add more detail about error handling
# > Generate mermaid diagrams for the architecture

One-Shot Mode

Use -p / --print for non-interactive, single-prompt runs. This is ideal for scripts and CI pipelines.

wakawiki -p "Document the public API surface"

Incremental Updates

Once you have an existing wakawiki/ directory, run --update to refresh only the files that changed since the last run. This avoids regenerating everything from scratch.

wakawiki --update

Heuristic Scan Mode

Run --scan for instant, deterministic documentation without an LLM. Ideal for CI or when you just need a structural overview.

wakawiki --scan

The scan mode:

• Parses Rust source files for pub items and doc comments
• Reads project metadata from Cargo.toml
• Generates index.md (dependencies, directory tree) and architecture.md (API reference)
• Respects your .gitignore for file filtering
• Runs in milliseconds

Code Indexing

Build a structured JSON index of all symbols and files in your project.

wakawiki --index

Output: wakawiki/index.json containing:

• Project metadata (name, version, description)
• All files with paths, sizes, hashes, and language
• All public symbols (fn, struct, enum, trait, etc.) with line numbers and docs

Query Index

Search the index by symbol name, kind, or file path.

wakawiki -q "Config"        # search by name
wakawiki -q "fn"            # search by kind
wakawiki -q "provider"      # search by path

Semantic Search

Generate embeddings and perform semantic search.

wakawiki --embed             # generate embeddings first
wakawiki -s "configuration"  # semantic search

Uses TF-IDF vectorization — no external API required.

MCP Server

Start an MCP (Model Context Protocol) server that exposes your codebase to AI agents.

wakawiki --index             # build index first
wakawiki --serve             # start MCP server on stdio

Available MCP tools:

• query_symbols — search by pattern
• get_symbol — lookup by name
• list_files — list all indexed files
• get_file_info — file details + symbols
• get_project_info — project metadata
• semantic_search — vector-based search (requires --embed)