Skip to content

Latest commit

 

History

History
81 lines (59 loc) · 4.4 KB

File metadata and controls

81 lines (59 loc) · 4.4 KB

Memex — Documentation

This folder documents the plugin itself.


Start here

Document What it covers
concepts.md Terminology, architecture, prior art, the three-layer model (raw / wiki / schema)
hook-catalog.md Every hook Memex ships, when it fires, what it enforces, how it fails
profile-authoring.md How to write a new profile (scaffolded template tree + memex.config.json)
cookbook.md Recipes: adding a custom folder, disabling a hook, extending the schema, building a custom profile
docsite.md The optional self-hosted browsable wiki — phases 1–7, configuration, auth modes, write features, per-page opt-outs
docker.md Self-hosting the docsite via Docker / Docker Compose — including Tailscale and reverse-proxy patterns

See also: ../CONTRIBUTING.md for development setup and coding conventions; ../SECURITY.md for the security policy; ../CHANGELOG.md for release notes.


Command reference

Slash commands shipped by the plugin. All are namespaced under /memex:.

Command Signature What it does
/memex:init /memex:init [profile] Scaffolds .memex/, memex.config.json, and appends to CLAUDE.md. Refuses to overwrite. Default profile: generic.
/memex:ingest /memex:ingest <path> Reads a raw source, summarises to the wiki, updates entity pages, bumps index.md, appends log.md.
/memex:query /memex:query <question> Searches wiki by index → reads matching pages → synthesises a cited answer. Offers to file the answer back.
/memex:lint /memex:lint [scope] Health-check: orphans, contradictions, stale claims, missing cross-refs, TODOs inline.
/memex:log /memex:log [--tail N] [--edit] Prints recent log; --edit opens for manual entry.
/memex:open-q /memex:open-q <title> Files a new open question under .memex/.open-questions/ with the profile's template.
/memex:promote /memex:promote <doc-path> <type> Promotes a .research/ or raw/ snippet to a permanent entity / concept / feature page.
/memex:graph /memex:graph [--format mermaid|dot] Emits a link graph. Feeds the "is this orphaned?" view.
/memex:init-profile /memex:init-profile [base-profile] Interactive: surveys the project, interviews the user, and generates a custom profile tailored to the actual folder shape. Use when none of the built-in profiles fit cleanly.
/memex:docsite /memex:docsite [serve|build|check] Wraps the memex-docsite CLI so the docsite is reachable from inside Claude Code. Requires pip install -e ".[docsite]".

Skills

Skills are auto-invoked by Claude when the conversation context matches their trigger phrases. You don't call them directly; the slash command above is usually a thin wrapper around the corresponding skill.

Skill Triggered by
ingest-source "ingest this", "process this article", paths into .memex/raw/
doc-query "what does the wiki say about", "based on my docs"
wiki-lint "lint the wiki", "audit my docs", "health-check the wiki"
open-questions-triage "triage open questions", "what's blocking me"
doc-refactor "split this page", "merge these", "rename this slug"
profile-builder "/memex:init-profile", "help me build a profile", "none of the profiles fit"

Subagents

Agent Used for
memex-ingestor Large or multi-part source ingests. Runs in worktree isolation so long ingests don't bloat the main session context.
memex-linter Wiki-lint passes on large wikis. Returns summary; original context stays clean.
memex-planner Pre-task planning: reads relevant wiki pages + open questions, produces a plan.

Configuration

Every Memex-enabled project has a memex.config.json at its root (or inside .memex/ — both paths are checked). This file drives every hook. See ../schemas/memex.config.schema.json for the full schema.

Minimal config:

{
  "$schema": "https://raw.githubusercontent.com/anthril/claude-memex/main/schemas/memex.config.schema.json",
  "version": "1",
  "profile": "generic",
  "root": ".memex"
}

See profile-authoring.md for how to extend it.