RAG source policy (customer chatbot)

This page defines allowed knowledge sources for the public AI assistant on docs.kkauto.net and kkauto.net.

Allowed for indexing

• Public content only at:
  • https://docs.kkauto.net/guide/** (Vietnamese)
  • https://docs.kkauto.net/en/guide/** (English)
• Prefer: how KKAuto works (Server + Client), quick start, wtadmin setup, KKClient, troubleshooting, FAQ.
• Canonical export: rag-export/corpus.jsonl (from npm run rag:export in the kk-docs2 repo).

Not allowed for indexing

• Internal technical docs: Go host, Redis, sockets, cron, internal website APIs.
• Legacy paths: old /user/, _archive, api/, demo/ on the docs site.
• Internal kkwiki, source code, production logs, unredacted screenshots.
• Credentials, tokens, proxies, PATs, local browser API URLs/ports.

Chatbot answering rules

• Explain the Management Website (Server) and KKClient desktop app using operational guidance only.
• Do not guess internal mechanics (Redis, socket frames, PHP queues).
• For «task pending» questions, mention Tasks on the website and Connected + Start Automation on the desktop.
• Never guide users to bypass Facebook checkpoints or CAPTCHA.
• Ask users to redact passwords/tokens/proxies before sharing support material.

Operations

Dify deployment, re-index, and QA: see plans/rag-customer-dify.md in the kk-docs2 repository.