Training the Bot

The support bot answers customer questions from two sources you control:

FAQs — question/answer pairs you curate.
Knowledge documents — PDFs (product guides, policies, help articles) whose text the bot uses as context.

When a customer sends the first message in a support conversation, the bot finds the most relevant FAQ/knowledge, generates a reply, and either answers or escalates to a human based on your confidence settings.

Bot configuration and training are dashboard-owner/admin actions. The endpoints below use a dashboard JWT. Most teams manage this in the dashboard UI; the API is here for automation.

FAQs

FAQs are the bot’s primary knowledge. Keep questions phrased the way customers actually ask.

curl -X POST https://dev.vocantly.com/api/v1/apps/{appId}/faqs \
  -H "Authorization: Bearer DASHBOARD_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "question": "How do I reset my password?",
    "answer": "Go to Settings → Security → Reset password and follow the email link."
  }'

curl https://dev.vocantly.com/api/v1/apps/{appId}/faqs \
  -H "Authorization: Bearer DASHBOARD_TOKEN"

curl -X PATCH https://dev.vocantly.com/api/v1/apps/{appId}/faqs/{id} \
  -H "Authorization: Bearer DASHBOARD_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{ "answer": "Updated answer text." }'

curl -X DELETE https://dev.vocantly.com/api/v1/apps/{appId}/faqs/{id} \
  -H "Authorization: Bearer DASHBOARD_TOKEN"

Write one FAQ per distinct question. Several short, specific FAQs beat one long catch-all — they match customer phrasing more reliably.

Knowledge documents (PDF)

Upload PDFs to give the bot broader context than FAQs alone. The platform extracts the text and uses it when generating replies.

curl -X POST https://dev.vocantly.com/api/v1/apps/{appId}/knowledge/documents \
  -H "Authorization: Bearer DASHBOARD_TOKEN" \
  -F "file=@product-guide.pdf" \
  -F "label=Product guide"

Format: PDF, up to 15 MB per file.
Manage: GET /apps/{appId}/knowledge/documents to list, DELETE …/{documentId} to remove.

Company context

The bot also grounds answers in your company name, website, and description (set on the app). This keeps replies on-brand and lets the bot answer basic “who are you” questions without an explicit FAQ.

Tuning when the bot answers

Configure the bot per app (optionally per department):

curl -X PUT https://dev.vocantly.com/api/v1/apps/{appId}/bots/config \
  -H "Authorization: Bearer DASHBOARD_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "is_enabled": true,
    "min_confidence": 0.6,
    "auto_escalate": true,
    "escalate_message": "Let me connect you with a teammate.",
    "department_id": null
  }'

boolean

Turn the bot on or off for this app/department.

number

0–1. If the bot’s confidence is below this, it won’t answer directly.

boolean

When confidence is low, automatically hand off to a human agent.

string

Message shown to the customer when escalating.

string

Scope this config to a specific department (optional).

Handover

A bot conversation can move to a human at any time, and back:

Auto-escalation — triggered by low confidence (above).
Agent takeover — an agent claims the conversation (POST /support/:conversationId/handover/takeover).
Resume bot — an owner/admin hands it back (POST /support/:conversationId/handover/bot-resume).

Clients are notified in real time via the support event (bot_takeover / bot_released) — see Support (SDK).

AI bot overview

What the bot does and how it fits the support flow.

Support flow

End-to-end: widget → bot → routing → agent.

Get Started

Chat

Customer Support

Voice & Video

Team

Integrations

FAQs

Knowledge documents (PDF)

Company context

Tuning when the bot answers

Handover

AI bot overview

Support flow

​FAQs

​Knowledge documents (PDF)

​Company context

​Tuning when the bot answers

​Handover

AI bot overview

Support flow

FAQs

Knowledge documents (PDF)

Company context

Tuning when the bot answers

Handover