Yes. Echo's multi-KB setup is fully supported. SleekAI lets you run a separate bot per KB, each with its own vector store, prompt, and display conditions. A typical setup runs a customer bot bound to the public knowledge base and a staff bot bound to the internal KB, with role-based display conditions enforcing the separation. The bots never see each other's articles.

Only if you choose to. Display conditions and post-type filters let you exclude restricted articles from indexing or expose them only on bots that are gated by user role. Internal-only articles published in a private Echo KB stay private; the bot pointed at the public KB never sees them. Treat the chat surface the same way you treat the article visibility settings on the KB itself.

Yes. Category and tag taxonomies are read alongside article content. That lets you scope a bot to one category, like a specific product line or a single topic, by adding a display condition. The bot will only retrieve articles inside that category, which is essential when one KB houses multiple products and you want the chat to behave like a per-product help desk rather than a single firehose.

Yes. With WPML, Polylang, or TranslatePress installed, the bot reads the active language and replies in it. Translated Echo articles in the multilingual setup feed the bot's context, so a visitor browsing the German version of a KB gets German answers grounded in the German articles, not in machine-translated English ones. The translation layer is owned by the multilingual plugin, not SleekAI.

On save. When an Echo article is published or updated, SleekAI sends the new content to OpenAI Files vector storage and queries it at runtime. Deletions remove the article from the index. There is no scheduled sync, no manual import, and no chance of a stale answer because someone forgot to run a job. The bot's library tracks the live KB to within seconds.

OpenAI, Anthropic, Google, or OpenRouter for inference. Vector storage runs on OpenAI Files. Bringing your own key keeps usage cost on your account and gives you direct control over rate limits, model choice, and data residency commitments. Most Echo customers pick OpenAI for retrieval and one of the larger Anthropic or OpenAI models for chat, but any combination works.

Yes. Define an escalation path in the prompt that routes to your support form, live chat, or ticket system when retrieval scores are low or the user explicitly asks. The bot's job is to answer when it can ground the reply in a real article and to hand off cleanly when it cannot. SleekAI ships with conversation logging so the support team can pick up exactly where the bot left off.

Echo's built-in rating widgets continue to work on the article pages themselves; the chatbot is a parallel surface, not a replacement. What SleekAI adds is a conversation log: every question that hits the bot is recorded with the article it cited and whether it deflected. That signal complements article ratings by telling you which articles users could not find through search at all, which is information the rating widget cannot capture.