Automation
Scheduled maintenance that keeps each vault tidy, current and duplicate-free — without you remembering to do anything.
Why automation matters
A knowledge base decays quietly. Files land in the vault and never get processed. The index falls behind the files it describes. Re-crawling a site you follow leaves three copies of the same page. None of it breaks loudly — you simply start getting slightly worse answers, and you never find out why.
Wikori's Automation page fixes that by scheduling the chores. Each vault has its own independent set of routines and its own schedule, so a fast-moving client vault can rebuild daily while an archive vault does nothing for weeks.
The four routines
| Routine | What it does | Good default |
|---|---|---|
| Process Pending Files | Scans the vault for files that are not yet in the index and moves them into INGEST/ so the AI pipeline picks them up. Catches anything you copied in by hand or that failed earlier. | Every 1 day |
| Rebuild Index | Regenerates the vault's index.md and SCHEMA.md from the current files, so search, the Knowledge Map and your AI agents see current reality. | Every 1 day |
| Duplicate Detection | Groups captured pages by source_url plus a content fingerprint, keeps the newest copy of each group and deletes the rest. Files without a source_url are never touched. The index is rebuilt afterwards. | Every 7 days |
| Re-run Web Crawler | Re-executes a saved crawler profile against its seed URL and submits newly discovered pages to the URL pipeline. Use it for sources that keep publishing. | Every 7 days |
Pair Re-run Web Crawler with Duplicate Detection. The crawler will happily re-capture pages it has seen before; deduplication is what stops your vault from growing three copies of the same article.
Setting up a schedule
Click Automation in the left sidebar. Routines are configured per vault, so make sure the vault you want is the active one.
Choose the routine type and the interval in days (minimum 1). For Re-run Web Crawler you also pick which saved crawler profile to use.
Toggle the routine on. It runs the next time its scheduled moment arrives while Wikori is running — keep the app in the system tray and it happens on its own.
Click Run Now to execute immediately and see the result. Note that running manually resets the routine's schedule from that moment.
Within one vault, routines run one at a time so they never collide. Across vaults they run in parallel — a long rebuild in one vault never blocks maintenance in another.
The maintenance log
Every vault keeps a short log of its last runs — which routine, when, whether it succeeded, and a one-line summary of what it did (files moved, duplicates removed, URLs submitted). The log holds the five most recent entries so it never grows without bound.
While a routine is executing, a running banner names it. Disabling a routine removes it from the queue immediately; deleting one stops it for good.
Watching health from Home
The Home page is where you check that everything is actually working. It shows a health card for each background service — File Watcher, Email Pipeline, MCP Server and Scheduler — each reading running, stopped or errored, with a link straight to the page that owns it.
Below that, Needs attention collects the things that are actually wrong right now: files that failed processing, a queue that has stopped moving, email pipeline errors, or an MCP server that is configured but not running. Each item links to the fix.
Home also carries quick actions — Process Queue, Rebuild Index and Reindex Vectors — for when you want to do by hand what the scheduler would do tonight.