Glossary

Canonical definitions for the nouns and concepts used across the CLI, documentation, and agent instructions.

Encyclopedia structure

Page

A wiki article in the main namespace. The fundamental unit of the encyclopedia. Has a title, content written in wikitext, a revision history, and belongs to one or more categories.

Page type

A grouping of pages that share a common structure and editorial approach. The current page types are Person and Episode, but new types may emerge as different kinds of personal data are added to the encyclopedia. See Page Types for details on each.

Section

A subdivision within a page, delimited by == Heading == syntax. Sections can be listed, read, and updated independently via wai section.

Lead paragraph

The opening prose before the first section heading. Summarizes the page topic and front-loads the most important information. Every page should have one.

Infobox

A structured sidebar template at the top of a page containing key-value facts. Format: {{Infobox Type | field=value}}. Each page type has its own infobox template.

Category

A tag grouping pages by type or theme. Syntax: [[Category:Name]]. Pages can belong to multiple categories. Use wai category to browse.

Wikilink

An internal link between pages. Syntax: [[Page Name]]. Wikilinks are how the encyclopedia cross-references itself.

Red link

A wikilink to a page that doesn't exist yet. Indicates a gap in the encyclopedia — a person, place, or event that's been mentioned but not yet documented.

Discussion

Talk page

A discussion page paired with a main page, in the Talk: namespace (e.g., Talk:Coorg Trip (2012)). Used for editorial notes, open questions, and coordination between agents and humans.

Thread

A topic within a talk page. Each thread has a subject line and content, and can be marked {{Open}} or {{Closed}}.

Data and storage

Vault

The content-addressed store on disk where source files are kept. Contains two subdirectories:

  • objects/ — individual files stored by their SHA-256 hash
  • snapshots/ — manifest files describing captured datasets

Files in the vault are immutable and deduplicated. The same file imported from multiple sources is stored only once.

Object

A single file stored in the vault, identified by its SHA-256 hash. Stored at objects/{prefix}/{hash} where prefix is the first two characters of the hash. Immutable once written.

Snapshot

A captured copy of a directory at a point in time. Created by wai snapshot <dir>. Produces:

  1. Objects — individual files hashed and stored in the vault
  2. Manifest — a JSON file mapping file paths to their hashes
  3. Source page — a wiki page in the Source: namespace documenting the dataset

Identified by a deterministic snapshot ID.

Snapshot ID

A 16-character hex string derived from the SHA-256 hash of a snapshot's manifest. Uniquely identifies a snapshot and its contents.

Manifest

A JSON file listing all files in a snapshot with their relative paths and SHA-256 hashes. Stored in vault/snapshots/{snapshotId}.json. Deterministic — the same set of files always produces the same manifest.

Source page

A wiki page in the Source: namespace (e.g., Source:WhatsApp Alice) documenting an ingested dataset. Created automatically by wai snapshot. Contains metadata (snapshot ID, file count, total size, file type breakdown) and querying instructions for programmatic access to the data in the vault.

Data source

A type of raw personal data that can be snapshotted and used to write pages. Supported types: photos, videos, messages and chats, voice notes, location data, financial data, and social media archives. See Data Sources for details on each.

Citations

Citation

A reference linking a factual claim in a page to a specific piece of evidence in the vault. Uses <ref>{{cite type | params}}</ref> syntax with type-specific templates: cite message, cite voice, cite photo, cite video. Each citation includes a hash field pointing to an immutable object.

Reference

A collected citation displayed in the == References == section at the bottom of a page via {{reflist}}. References are numbered footnotes rendered by MediaWiki.

Hash

The SHA-256 content-addressed identifier of a file in the vault. Used in citations to link claims to specific, immutable source files. Even if the original file is moved or renamed, the hash still resolves.

Tasks

Task

A work item tracking editorial work, stored as a page in the Task: namespace (e.g., Task:0001). Contains an {{Infobox Task}} with metadata and a description of what needs to be done.

Task queue

The system of Task pages and category-based status tracking used to coordinate work between agents and humans. Enables parallel work, retry on failure, and human oversight.

Task lifecycle

The state machine a task moves through:

  • pending — waiting to be picked up
  • in-progress — claimed by an agent or human, work underway
  • done — completed successfully
  • failed — could not be completed (can be requeued back to pending)

System components

Wiki

The local MediaWiki instance that stores the encyclopedia. Runs at http://localhost:8080. Stores pages, revisions, images, and categories in a SQLite database.

CLI (wai)

The command-line tool for managing the wiki, snapshotting data, and operating the task queue. The primary interface for both humans and agent harnesses.

Agent harness

An MCP (Model Context Protocol) server that gives AI coding tools read/write access to the wiki via wai commands. The bridge between an AI agent and the encyclopedia.

Plugin

The installable package that provides the agent harness for a specific coding tool. Installed via wai plugin install <tool> (e.g., claude-code, codex, opencode).

Namespace

A MediaWiki organizational prefix for pages. The project uses four namespaces:

  • Main (no prefix) — regular encyclopedia pages
  • Talk (Talk:) — editorial discussion pages
  • Source (Source:) — documentation of ingested datasets
  • Task (Task:) — editorial work items