ScaiBot

ScaiBot is the chat product on top of ScaiGrid. You configure a bot — its model, tone, knowledge base, escalation rules — in the admin UI, then embed it on any web page with a single <script> tag. Visitors chat with it; you see transcripts, analytics, and route hard cases to humans.

It is built on top of ScaiGrid's inference, accounting, and identity layers, so everything a bot consumes is metered, budgeted, and audited the same way any other inference call is.

When to use it#

• You want a chatbot on your website and don't want to integrate an LLM API yourself.
• You have product docs, FAQs, or handbooks you want the bot to answer from (RAG).
• You need to escalate to a human, an email, a Slack channel, or a webhook when the bot doesn't know the answer.
• You want conversation analytics — completion rate, escalation rate, average turns, tokens per conversation.

If you only need a raw chat completion API, you don't need ScaiBot — call ScaiGrid's /v1/inference/chat directly.

What you get#

• Embeddable widget. Single <script> tag, no iframe gymnastics, persists across reloads.
• Knowledge ingestion. Upload PDFs, Word docs, Markdown, HTML — they get chunked, embedded, and indexed.
• Grounded answers. Every response cites the knowledge source it came from.
• Tone configuration. Set formality, verbosity, empathy, vocabulary, scope — without writing a system prompt.
• Escalation rules. Keyword, intent, sentiment, confidence, or explicit-request triggers; email/webhook/Slack/queue actions.
• Conversation store. Every chat is persisted server-side and retrievable.
• Analytics. Per-bot completion, escalation, and token metrics.

Two-minute mental model#

You manage three nouns and one verb:

• A Bot is a configured assistant. A tenant owns many bots.
• A Conversation is one visitor's chat session with one bot.
• A Knowledge Base is what the bot is allowed to read from.
• And the verb: a visitor chats, which means sending messages and getting responses.

The widget on your page is just a thin client over the chat verb.

Where to go next#

• Quickstart — get a bot running and embedded in five minutes.
• Architecture — how the widget, API, and ScaiGrid fit together.
• Knowledge and RAG — how documents become grounded answers.
• Escalation — when and how the bot hands off.
• API reference — every endpoint, request, response.
• Embed a bot on your site — full walkthrough.

ScaiBot's module ID inside ScaiGrid is scaibot; its API is mounted at /v1/modules/scaibot/.