Knowledge
Read the two Sources ARCY drafted from your own pages, add your own, train them, see when ARCYBot follows robots.txt, and put a retired page back.
Add and train the sources the agent answers from on Knowledge. Two text Sources are there on your first read, Product Information and the Visitor FAQ, from your own pages. ARCY's Reports for your team are on Reports, and no answer quotes those.
Before you start
- A Web crawler reads a Site under a verified Domain (Domains) with no page cap, the Product URL on Product or a name under it, and once per Product the creator's work email domain. The last two stop at 50 pages each.
- A Walkthrough needs a verified Domain.
Setting it up
Add source takes five types:
- Web crawler. Give it a Starting URL; it follows your sitemap and the links under that path. Turn off Allow linking to this source in answers to keep pages unlinked. For Follow robots.txt, see ARCYBot.
- Flows. Pick Flows to suggest (Flows).
- Text. 15,000 characters at most, Markdown included. Facts the agent should know, such as prices and plans, go here: Persona holds instructions only.
- File upload. PDF, DOC, DOCX, TXT or MD, 10 MB at most. To replace one, delete and add it again.
- Walkthrough. Pick a Site and path under Starting page, then Start walkthrough. Your Product opens in a new tab and each screen becomes one document of its controls and fields. A dialog or tab is its own screen, a menu joins its screen. Never captured: what you type, table rows, select options, on or off states, screenshots. One per Site; Stop saves and the row reads Captured.
The agent buttons open the Add training sources entry on ARCY MCP. A proposed source reads Proposed by your agent; Accept adds and trains it, Reject drops it.
Press Train on the row. Runs are under Training log, pages under Documents. Opening one renders its formatting, and a text Source keeps its own line breaks. The two ARCY wrote read From your site until you edit them, and an edit changes what ARCY answers next.
A second Train reads only new pages. To reread a changed one, use Fetch this document again, or Retrain from the start, which deletes every saved page and crawls again. Leave out of future crawls drops a page for good; Stop training keeps what it saved.
Walk again replaces every screen on Stop; Leave this screen out of training survives one.
What it changes
- One crawl saves 1,000 pages; a larger site continues on the next Train.
- A page over 5 MB is skipped and named in the log, and one Web crawler reads one Site.
- Each Train drops pages robots.txt now closes, and rereads them once reopened.
- A crawled or walked page cannot be edited: leave it out, add a Text source instead.
- A text source ARCY wrote names, under its own name, the question and how many visitors asked it (Conversations). Edit or delete it like any other; ARCY does not write it again.
- A newly trained page retires an older one from a different Source that names the same thing and repeats most of the same words, so one procedure is never answered two ways. The older page reads Retired under Documents, stays there to open, and is no longer answered from. Restore puts it back. Two pages of one Source never retire each other.
- A walk holds 300 screens and stops after 30 minutes.
- Deleting a source removes its content at once.
ARCYBot
The crawler is ARCYBot/1.0. On a Domain you proved, or your work email's domain, it sets aside robots.txt rules written for every bot (User-agent: *) unless the source has Follow robots.txt on. On the Product URL every rule binds, and so does a group naming ARCYBot, Crawl-delay included. To keep it out:
User-agent: ARCYBot
Disallow: /When it goes wrong
- An address is refused, as covered by no proof or as read once already. Set it as the Product URL on Product, or verify its Domain.
- A crawl fails. Point at Failed for why (more).
- Far fewer pages than your site has. Pages are unlinked and there is no sitemap.
- The walkthrough tab has no ARCY bar. The link expired, the Publishable key is another environment's, or arcy.js is missing.
See also
Preview
Try ARCY on a page of your own site before anything is installed, and see which jobs work in it now and which are ready once your product is connected.
Conversations
Review every conversation, read every step behind an answer, find the questions ARCY could not answer, see what last night's check flagged, and read how your chat is doing.