Installation & setup

Get wiki-builder installed and running in a few minutes.

Requirements

Python 3.9 or higher
An API key for Anthropic or OpenAI
pdftotext (optional) — required only if you want to ingest PDFs directly

Install

pip install wiki-builder

This installs the wiki command into your environment.

To install from source:

git clone https://github.com/Turi-Labs/wiki-builder
cd wiki-builder
pip install -e .

Set an API key

wiki-builder needs an API key to call the LLM. You can provide it in three ways:

Option 1 — environment variable (recommended)

# Anthropic
export ANTHROPIC_API_KEY=sk-ant-...

# OpenAI
export OPENAI_API_KEY=sk-...

Add the export to your ~/.zshrc or ~/.bashrc to persist it across sessions.

Option 2 — CLI config

wiki config --provider anthropic --api-key sk-ant-...
# or
wiki config --provider openai --api-key sk-...

This saves the key to ~/.wiki-builder/config.json.

Option 3 — per-command flag

wiki ingest raw/paper.md --provider openai

The tool resolves provider and key in this order: CLI flags → environment variables → saved config → defaults.

Quickstart

Initialize a project, add a document, ingest it, and query the result.

# 1. Create a new wiki
wiki init my-research/ --name "AI Research"

# 2. Copy in a source document
cp ~/papers/paper.md my-research/raw/

# 3. Ingest the document
cd my-research
wiki ingest raw/paper.md

# 4. Query your knowledge base
wiki query "what are the main contributions of this paper?"

# 5. Check wiki health
wiki status
wiki lint

What wiki init creates

my-research/
├── schema.md          # LLM behavioral spec (you can edit this)
├── .gitignore
├── raw/               # drop source documents here
│   └── assets/
└── wiki/
    ├── index.md       # master catalog
    ├── log.md         # operation log
    ├── concepts/
    ├── entities/
    └── sources/

PDF support

To ingest PDFs, install pdftotext from the Poppler utilities package:

# macOS
brew install poppler

# Ubuntu / Debian
sudo apt install poppler-utils

Then pass a PDF directly to wiki ingest:

wiki ingest raw/paper.pdf

# Delete the PDF after converting to text
wiki ingest raw/paper.pdf --delete-pdf

Tips

Ingest one source at a time. The LLM produces better synthesis when it focuses on a single document.
Edit schema.md as you go. Changing naming conventions or adding page types is just a text edit.
Keep raw/ immutable. It's your source of truth. The LLM only writes inside wiki/.
Use git. Every ingest and query-save is a natural commit point. You get full history and easy rollback.
Open the wiki in Obsidian. The graph view and Dataview plugin work great with this layout.

Next: Command reference