Overview

This page describes the current operating shape of vipin wiki.

Layers

• Public wiki: wiki/
• Private local-only wiki: maintained in a separate local-only layer
• Raw sources: raw/
• Reader layer: reader-context.md
• Operating layer: AGENTS.md, .wiki-schema.md, WORKFLOWS.md, CONTRIBUTIONS.md, .codex/skills/vipin-wiki/, and scripts/
• Public website layer: site/ publishes the public wiki/ layer through Quartz and GitHub Pages.

Active Public Domains

• llm-wiki
• llm-based-recommendation
• personal-knowledge-systems
• 2026-04-21-llm-rec-research-map

Section Entrypoints

• home
• index
• queries-home
• synthesis-home
• timelines-home
• topics-home
• comparisons-home

Automation Modules (2026-05-24)

Five new operational scripts added to scripts/:

• wiki-scheduler.py — automated ingest scheduler (10 jobs, cron-style, daemon mode, Windows Task Scheduler install)
• wiki-graph.py — knowledge graph query CLI (search, neighbors, path, communities, god nodes, bridges, export)
• wiki-dashboard.py — agent performance dashboard (terminal + HTML, token summary inline)
• wiki-tokens.py — token usage tracker (per-agent cost, 7-day rolling, HTML section)
• ingest_github.py + ingest_karpathy.py — Python ingest base replacing PowerShell scripts

Notes

• Search and context assembly now use a shared parser/index core so catalog, search, context, status, and lint share one interpretation of pages.
• The default query loop is fast answer first, durable ingest second.
• Private holdings stay outside public indexes and public logs.
• Collection-style ingest is now a first-class workflow alongside one-by-one ingest.
• Generated graph artifacts remain optional because Quartz provides website search, backlinks, and graph navigation.
• A project-local Codex skill now routes future agents through the wiki’s query, ingest, crystallization, maintenance, and public/private safety workflows.
• Knowledge graph (graphify) is updated incrementally via post-commit hook; wiki-graph.py provides query interface.