User Guide

Everything you need to know — from creating your first bot to embedding it on your website.

AI-POWERED CHATBOT PLATFORM

What is QubotAi?

QubotAi lets you build a smart, trained AI chatbot for your website — without writing a single line of code. You feed it your content, it learns from it, and your visitors get instant, accurate answers around the clock.

Whether you run a small business, an e-commerce store, a SaaS product, or a professional service — QubotAi turns your existing knowledge into a 24/7 customer-facing assistant that handles questions, collects leads, and keeps visitors engaged.

Your AI Bot
Trained on your content · Always on
Visitor: "What are your business hours?"
Bot: "We're open Monday to Saturday, 9 AM – 7 PM. You can also reach us via the contact form anytime!"

Why Your Business Needs It

Always Available

Your bot answers questions at 2 AM on a Sunday, while you sleep. No missed leads, no frustrated visitors waiting for a reply.

Captures Leads Automatically

When a visitor asks something the bot can't answer, it can collect their name and contact details — turning a dead-end into a warm lead for your team.

Cuts Support Load

Repetitive questions — pricing, hours, return policy, how-tos — get answered instantly. Your team focuses on the queries that actually need a human.

Speaks Any Language

The bot detects the visitor's language and replies in the same one automatically — no extra setup needed for a multilingual audience.

Zero Coding Required

Add the bot to any website — WordPress, Shopify, Webflow, or plain HTML — by pasting two lines of code. No developer needed.

Built-in Analytics

See what visitors are asking, how many sessions you're getting, and which bots are performing — all from a single dashboard.

Up and Running in 3 Steps

1
Create & train your bot

Upload your docs, paste your FAQs, or point it at your website. Hit Train and the bot learns your content in minutes.

2
Customise the widget

Set your brand color, choose a voice, write an opening message, and adjust the persona with a Master Prompt — all from the dashboard.

3
Embed on your website

Copy two lines of code from your bot's page and paste them into your site. Your AI chatbot is live — immediately, everywhere.

Scroll down to follow the full guide — each section walks you through one part of the process with settings explained, examples included, and tips along the way.

1

Creating a Bot

A bot is your AI-powered chatbot. You can create multiple bots — one per website, product, or use-case, depending on your plan.

  1. Go to Dashboard → My Bots
  2. Click Create New Bot
  3. Enter a Bot Name (visible to your visitors in the widget header) and an optional Description (for your own reference)
  4. Click Save — your bot is created in Draft status
A freshly created bot has no knowledge and no training. You must add contexts and train the bot before it can answer questions.
2

Bot Settings — Every Option Explained

Open a bot and click Edit to access all settings. They are grouped into four sections.

Basic Info
Bot Name The display name shown in the chat widget header. Keep it short and friendly — visitors will see this.
Description Internal note for yourself. Visitors never see this. Useful when you manage multiple bots.
Master Prompt The single most important setting. This is the instruction you give the AI before every conversation — it defines its personality, tone, task, and what to do when it doesn't know something. See Section 7 for examples and best practices.
Widget Prompt Strategy Controls how the masterPrompt passed via the embed snippet (see Section 6) interacts with this Master Prompt.
  • Override — The widget prompt completely replaces this Master Prompt (default).
  • Append — The widget prompt is added to the end of this Master Prompt.
Widget Settings
Bot Color Primary color of the chat button, header, and user message bubbles. Use a hex value that matches your brand. Default: #4f46e5 (indigo).
Bot Size Controls the dimensions of the chat window popup.
  • Small — 300 × 400 px (compact, suits sidebars)
  • Medium — 370 × 520 px (default, works on most sites)
  • Large — 480 × 650 px (spacious, better for complex Q&A)
Position Where the floating chat button appears on the screen. Options: Right Bottom (most common), Left Bottom, or Center Bottom.
Initial Message The first message the bot automatically shows when a visitor opens the chat. Set something welcoming and action-oriented. Example: "Hi! I'm here to help. Ask me anything about our products."
Widget Theme Controls the color scheme of the chat window body.
  • Light — white chat background, light grey message area (default). Works on most websites.
  • Dark — near-black chat background with dark bubbles and a styled dark scrollbar. Ideal for dark-mode websites.
The header, chat button, and visitor message bubbles always use your brand color regardless of theme. You can also override this per-page via the embed snippet — see Section 6.
Branding
Remove "Powered by QubotAi" When toggled on, the small "Powered by QubotAi" link at the bottom of the chat widget is hidden. Useful for white-label deployments or when you want a fully branded experience.
Chat Limits
What counts as 1 exchange?
One visitor message + one bot reply = 1 exchange. This is the unit used for both the per-session limit below and your plan's monthly allowance.
Max exchanges per session Limits how many exchanges a visitor can have in a single session. Leave blank for unlimited. Useful for controlling costs on high-traffic sites.
Limit reached message The message displayed to the visitor once they hit the exchange limit. Use this to redirect them — e.g., "You've reached the limit for this session. For more help, email us at support@example.com."
Lock input after limit When on: the chat input is disabled after the limit is reached — the visitor cannot type further. When off: the visitor can still type but the AI will not respond; messages are stored as logs.
Reset session after (minutes) Once a session hits the limit, it will automatically reset after this many minutes — giving the visitor a fresh session. Leave blank to never reset. Example: 60 means the session resets after 1 hour, allowing the visitor to chat again.
Conversation Summary Plan feature

Controls automatic AI summarization of conversations after a visitor goes idle. Manual summarization from the conversation detail page is always available regardless of this setting.

Auto-summarize conversations When toggled on, any session that has been idle for 10 minutes is automatically summarized by AI in the background. The summary appears in the conversation detail view and in your Sessions CSV export. No action is needed from you — it runs on a schedule. If new messages arrive after a summary is created, the session will be summarized again the next time it goes idle.
Manual vs Auto: You can always click Summarize on any conversation detail page regardless of this toggle. The toggle only controls the automated background run. If no new messages have arrived since the last summary, the button is disabled — there is nothing new to summarize.
File & Image Collection Plan feature

Allows your widget to accept file and image uploads from visitors. Uploaded files are stored securely and are not processed by AI or read as knowledge — they are simply collected for your team to review and download.

Enable file & image uploads When on, a 📎 paperclip button appears next to the chat input in your widget. Visitors can click it to select a file. After uploading, the bot automatically replies acknowledging receipt and letting the visitor know your team will review it.
Allowed file types Images (JPG, PNG, GIF, WebP), PDF, Word (DOC, DOCX), Excel (XLS, XLSX), CSV, and plain text (TXT). Other types are rejected with a friendly message.
File size limit Maximum 10 MB per file. Files exceeding this are rejected client-side before upload.
Viewing & downloading files Open any conversation in Chat History. File uploads appear as a downloadable bubble in the conversation thread — click the bubble to download the file. Files are stored against the session so you always know which visitor sent what.
This feature is useful for collecting CVs, support screenshots, signed forms, or any document a visitor would otherwise email you. Combine it with the lead-capture Master Prompt to collect contact details alongside the file.
Files are stored in your server's storage. Periodically review and clear old uploads if storage space is a concern. The AI never reads or learns from uploaded files.
3

Adding Knowledge Contexts

Contexts are the knowledge sources your bot learns from. Without contexts, your bot can only use its Master Prompt. Add at least one context before training.

Open your bot → click Add Context. You can add multiple contexts of different types to the same bot.

Context Types
Plain Text

Directly paste or type any text — FAQs, product descriptions, policies, or anything you want the bot to know. Ready immediately — no processing needed.

Best for: Short FAQs, quick facts, manual content entry.

Website URL

Provide a URL and QubotAi scrapes the content automatically. Three scraping modes:

  • Single Page — scrapes only the URL you enter.
  • Follow Links (Crawl) — starts at your URL and follows internal links up to the depth you set (1–5 levels). Crawls up to 100 pages max.
  • Sitemap — provide your sitemap.xml URL; scrapes all listed pages (up to 100).

Best for: Entire websites, documentation portals, knowledge bases.

PDF File

Upload any PDF file (max 20 MB). QubotAi extracts the text content automatically.

Best for: Product manuals, brochures, reports, legal documents.

Word Document (.docx)

Upload a .doc or .docx file (max 20 MB). Text is extracted and processed automatically.

Best for: Internal documentation, onboarding materials, SOPs.

CSV File

Upload a .csv or plain text file (max 20 MB). Rows are extracted as structured text the bot can reference.

Best for: Product catalogues, pricing tables, structured data.

Markdown / TXT

Upload a .md or .txt file (max 20 MB). Great for developer documentation or any plain-text content.

Best for: README files, developer docs, changelogs.

Context Processing States

File and website contexts must be processed before training. You will see one of these statuses on your bot's context list:

Pending Queued for processing. Usually updates within a minute.
Processing Currently extracting content. Do not train until this finishes.
Ready Content extracted successfully. The context is ready to be trained on.
Failed Processing failed (e.g. blocked website, corrupt file). Fix the source and click Reprocess.
Plain Text contexts are always immediately ready — they skip the processing queue entirely. If you need something available right away, use Plain Text.
4

Training the Bot

Training converts all your contexts into searchable AI knowledge (vector embeddings). The bot uses these at query time to find the most relevant chunks and include them in its response.

How to Train
  1. Make sure all your contexts are in Ready status
  2. Open the bot page and click Train Bot
  3. The bot status changes to Training…
  4. Once complete, status becomes Ready — your bot can now answer questions
Training counts against your monthly training runs limit. Each click of "Train Bot" uses one run, regardless of how many contexts you have.
When to Retrain

You must retrain whenever you want the bot to reflect changes. Retrain when:

  • You add a new context (any type)
  • You edit an existing Plain Text context
  • You re-upload a file on an existing context
  • You change the website URL or crawl mode on an existing context
  • Your source website content changed and you want the bot to reflect the latest content (reprocess the context first, then retrain)
  • You deleted a context and want to remove its knowledge from the bot
Changing the Master Prompt, widget colors, size, position, or chat limits does not require retraining — those settings take effect immediately on save.
Reprocessing a Context

For website and file-based contexts, you can click Reprocess on the context to re-fetch/re-extract its content (without adding a new context). This is useful when:

  • Your website content was updated and you want the bot to learn the new version
  • A context previously failed and you want to retry it after fixing the source

After reprocessing completes (Ready), click Train Bot to apply the new content.

5

Trying the Bot

Before embedding the bot on your live website, test it directly from your dashboard using the Try Bot button — available on a bot's detail page once its status is Ready.

The Try Bot page opens a live, full-featured chat interface powered by the same AI engine your visitors will use. Use it to:

  • Verify the bot answers questions correctly from your contexts
  • Test edge cases — questions your content doesn't cover
  • Check the tone and behavior defined by your Master Prompt
  • Spot-check after a retrain to confirm changes took effect
Messages sent in "Try Bot" are real API calls and count toward your plan's monthly exchange limit.
6

Embedding on Your Website

Copy the two-line snippet from your bot's detail page and paste it just before the </body> closing tag on any page of your website.

<script>window.QubotConfig = { apiKey: "YOUR_API_KEY" };</script> <script src="https://qubotai.in/widget.js"></script>

Replace YOUR_API_KEY with the API Key from your bot's settings page. Each bot has a unique key — do not mix them up.

The snippet works on any platform — WordPress, Shopify, Webflow, Wix, custom HTML, or any framework.

Advanced: Page-Specific Master Prompt

Add a masterPrompt key to provide additional instructions on a specific page. Depending on your Widget Prompt Strategy setting (see Section 2), this will either replace your default Master Prompt or be appended to it.

<script> window.QubotConfig = { apiKey: "YOUR_API_KEY", masterPrompt: "You are a sales assistant for our Pricing page. Focus only on plan features and pricing. Be persuasive but honest." }; </script> <script src="https://qubotai.in/widget.js"></script>
Advanced: Force Dark Mode on a Specific Page

You can override the Widget Theme setting from your bot dashboard on a per-page basis using the theme key. Useful if only one page on your site has a dark background.

<script> window.QubotConfig = { apiKey: "YOUR_API_KEY", theme: "dark" }; </script> <script src="https://qubotai.in/widget.js"></script>
React / Next.js Integration

Create a small component that injects the script once on mount:

import { useEffect } from 'react'; export default function ChatWidget({ apiKey, masterPrompt }) { useEffect(() => { window.QubotConfig = { apiKey, masterPrompt: masterPrompt || null }; const script = document.createElement('script'); script.src = "https://qubotai.in/widget.js"; script.async = true; document.body.appendChild(script); return () => { document.body.removeChild(script); }; }, [apiKey]); return null; } // Usage: // <ChatWidget apiKey="YOUR_API_KEY" /> // <ChatWidget apiKey="YOUR_API_KEY" masterPrompt="Custom prompt for this page" />

Mastering the Master Prompt

The Master Prompt is the single instruction that defines how your bot behaves. Your training data provides the facts; the Master Prompt provides the personality, rules, and fallback behavior.

Example 1: Support Bot with Lead Capture

Best when you want to capture contact details for questions the bot can't answer.

"You are a support assistant for Acme Corp. Answer strictly using the provided knowledge. If the answer is not in the knowledge base, say you don't have that information right now, then politely ask for the user's name and email or mobile number so a team member can follow up within 24 hours."
Example 2: Friendly Brand Persona

Best for lifestyle, creative, or consumer brands.

"You are 'Aria', a cheerful and enthusiastic guide for StyleHouse. Use friendly language, include an occasional emoji, and keep answers short (2–3 sentences max). If you can't answer from the knowledge base, suggest the user browse our website or reach out on Instagram @stylehouse."
Example 3: Strict Professional / Legal Mode

Best for legal, medical, or compliance-heavy contexts.

"You are a professional documentation assistant. Answer only from the provided official documentation. If the information is not found, respond exactly: 'I apologize, but that information is not available in our official documentation.' Do not infer or speculate. Keep answers factual and concise."
Example 4: Sales-Focused Page Prompt (via widget override)

Override the default prompt only on your pricing or landing page.

"You are a helpful sales assistant on our Pricing page. Your goal is to help visitors choose the right plan. Highlight the value of paid plans, address objections honestly, and encourage visitors to start a free trial or contact us if unsure. Keep it conversational."
What Makes a Good Master Prompt?
  • Define the role clearly — "You are a support assistant for [Company Name]" sets the stage instantly.
  • Set the tone — formal, friendly, concise, technical — the AI will match it.
  • Handle the unknown case — always tell the bot what to do when it doesn't know the answer (capture a lead, point to another resource, etc.).
  • Set a length expectation — "Keep answers under 3 sentences" prevents verbose replies that lose visitors.
  • Don't over-restrict — very long prompts with too many rules can confuse the model. Focus on the most important 3–4 instructions.
8

Chat Analytics

Go to Analytics in the sidebar to get a birds-eye view of how all your bots are performing. Data is aggregated across every bot you own.

Summary Cards
Total Messages

Total number of individual exchanges (user message + bot reply) across all your bots, all time.

Total Sessions

Number of unique visitor sessions that have chatted with any of your bots. Each browser session is counted once.

Active Bots

Count of your bots that are in Ready status and have received at least one message.

Avg. Messages / Session

Total messages ÷ total sessions. A higher number means visitors are having longer, more engaged conversations.

Limit-Hit Sessions

Number of sessions where a visitor hit the per-session exchange limit you configured. Useful for tuning your Chat Limits setting.

Activity Chart — Last 30 Days

A dual-line chart showing sessions and messages per day over the past 30 days. Gaps on days with no activity are shown as zero — not omitted — so you can clearly see quiet vs. busy periods.

Top Bots

A ranked table of your bots sorted by total message count. Shows message count and unique session count per bot — useful when you have multiple bots and want to see which ones are getting the most traffic.

Recent Sessions

The 5 most recently active sessions across all bots, with the last message preview and timestamp. Click any session to jump straight to its full conversation in Chat History.

9

Chat History

Go to Chat History in the sidebar to browse every conversation your bots have had. You can filter by bot and paginate through sessions.

Session List

Each row in the list represents one visitor session and shows:

  • Bot name — which bot handled the conversation
  • Started at — when the visitor first sent a message
  • Last message at — when the conversation last had activity
  • Message count — total exchanges in that session
  • Last message preview — the visitor's most recent question
  • Limit hit badge — shown if the session reached the exchange limit
Session Detail View

Click any session to open the full conversation transcript, displayed in the same chat-bubble format the visitor saw. You can also see:

  • The complete back-and-forth between the visitor and the bot
  • The AI-generated summary (if one has been created — see Section 10)
  • A button to generate or regenerate the summary
10

AI Chat Summarization

On any session detail page, click Generate Summary to have the AI produce a structured summary of that conversation in seconds.

What the Summary Contains

The summary is structured into four sections using Markdown:

What the user was asking about A brief description of the visitor's overall intent or topic.
Key points discussed A bullet list of the main topics, questions, and answers that came up in the chat.
Resolution status Whether the visitor's query was resolved, left unanswered, or partially addressed.
Contact details shared Any name, email, or phone number the visitor mentioned during the conversation — extracted automatically.
You can regenerate a summary at any time — for example, if the session continued after the first summary was created. Each click overwrites the previous summary.
Summaries are also included in the Sessions CSV export (see Section 11), so you can review them in bulk without opening each session individually.
11

Chat Export

Go to Export Data in the sidebar to download your bot's conversation data as CSV files. You select a bot and an optional date range, then download one or both export types.

Export Types
Conversations Export

One row per individual message exchange. Downloads as .csv. Contains:

Session First 8 characters of the session token — enough to group messages from the same visitor together.
Date The date the message was sent (YYYY-MM-DD).
Time The time the message was sent (HH:MM:SS).
User Message The exact text the visitor typed.
Bot Response The full reply the bot sent back.
Limit Reached Yes or No — flags messages that were sent after the session exchange limit was hit.
Sessions Export

One row per visitor session. Downloads as .csv. Contains:

Session First 8 characters of the session token.
Started At Date and time the session began (YYYY-MM-DD HH:MM:SS).
Last Activity Date and time of the last message in the session.
Total Exchanges Number of back-and-forth messages in the session.
Summary The AI-generated summary for this session, if one has been created (see Section 10). Empty if not yet summarized.
Date Range Filter

Both exports accept an optional From and To date. Leave both blank to export all data. When a range is set, the filename includes the dates for easy archiving — e.g. my-bot-conversations-2025-01-01-to-2025-01-31.csv.

Use the Sessions Export + AI Summaries combination as a lightweight CRM — generate summaries for sessions where visitors shared contact details, then export to get a CSV of leads with context about what they were asking about.
13

Interact Mode Plan feature

Interact Mode turns your text chatbot into a fully voice-driven conversation. Visitors can speak to the bot, hear its reply spoken aloud, and interrupt it mid-sentence — just like a real conversation.

Requires HTTPS. Browsers only grant microphone access on secure origins. Make sure your website is served over https:// before enabling this feature.
How to Enable
  1. Your plan must include the Interact Mode feature. Check your plan details or contact support.
  2. Open the bot you want to enable it on and click Edit.
  3. Under the Interact Mode section, toggle Enable Interact Mode on and choose a TTS Voice for the bot.
  4. Click Save. The microphone button will now appear in the widget for all visitors.
TTS Voice Options

Choose the voice the bot speaks with. Available voices:

alloy Neutral and balanced. Works well for general-purpose bots.
echo Measured and authoritative. Good for professional or informational bots.
fable Expressive and storytelling in tone. Suits creative or brand-led experiences.
onyx Deep and confident. Great for sales or support bots that need gravitas.
nova Warm and friendly female voice. The default — works on almost every use-case.
shimmer Soft and gentle female voice. Great for lifestyle, wellness, or customer-care bots.
How It Works for the Visitor
  1. The visitor clicks the microphone button in the chat widget.
  2. The browser asks for microphone permission — the visitor must allow it.
  3. The bot enters listening mode (pulsing mic icon). The visitor speaks naturally.
  4. After a brief pause the bot processes the speech, fetches a reply, and speaks it aloud (speaker wave icon).
  5. Once the bot finishes speaking it automatically returns to listening, ready for the next question — no button press needed.
  6. To end voice mode, the visitor clicks the button that appears while Interact Mode is active.
Barge-In — Interrupting the Bot

Visitors do not have to wait for the bot to finish speaking. If they start talking while the bot is mid-reply, the bot detects the interruption, stops speaking immediately, and begins listening. This makes conversations feel natural and fluid.

Barge-in uses the microphone to detect sustained voice from the visitor. It filters out the bot's own audio so the bot does not interrupt itself — only a real human voice triggers it.
Language Support

The bot automatically detects the language of each spoken message and replies in that same language. Speak in English, get an English reply. Speak in Hindi, get a Hindi reply — no configuration needed.

Voice Mode Responses

In Interact Mode the bot keeps its replies short and conversational — typically one sentence — so they sound natural when spoken aloud. Bullet points, lists, and markdown formatting are suppressed automatically in voice mode.

Each voice exchange (spoken question + spoken reply) counts toward your plan's monthly message limit, the same as a regular text exchange.
14

Conversation Report Emails Plan feature

Get an automatic email digest of all conversations your bots have had — delivered to your registered email address at whatever frequency you choose. Each email summarises every session in 2–3 plain-English sentences so you can stay on top of what your visitors are asking without opening the dashboard.

How to Enable
  1. Go to Profile (top-right menu → My Profile).
  2. Scroll to the Notification Preferences section.
  3. Use the Conversation Report Email dropdown to pick your frequency.
  4. Click Update Profile to save.
This section is visible to all users but is disabled for free plan accounts. Upgrade to a paid plan to unlock it.
Frequency Options
Hourly A report is sent once every hour covering the conversations that happened in the last 60 minutes. Best for high-traffic bots where you want near real-time awareness.
Every 4 hours Four reports per day — sent once every 4 hours. A good balance between staying updated and inbox noise.
Every 6 hours Three reports per day, spaced 6 hours apart.
Every 12 hours Two reports per day — a morning digest and an evening one. Works well for most businesses.
Once a day One daily email covering all conversations from the past 24 hours. The lowest-noise option — ideal if you just want a morning overview.
What the Email Contains
Stats bar Total conversations in the period, total messages exchanged, and the start time of the window covered by the report.
Conversation cards One card per conversation — showing the bot name, how many messages were exchanged, the time it started, and a 2–3 sentence plain-English summary of what the visitor needed and whether it was resolved.
View All link If there were more than 10 conversations in the period, the email shows the first 10 and includes a View All Conversations button that takes you directly to your Chat History page for the full list.
How Summaries Are Generated

The report relies on AI-generated conversation summaries. Here is what happens automatically before each email is sent:

  1. The system fetches all sessions that started since your last report.
  2. Any session that does not yet have a summary is automatically summarised by AI at send time — you do not need to do anything manually.
  3. Each full summary is then condensed to 2–3 sentences for the email digest.
If a reporting window had zero conversations, no email is sent. The clock resets and the next window starts fresh — you will not receive empty reports.
Reports are sent to your registered email address. If you change your email in Profile, future reports will go to the new address. Make sure your email is correct before enabling this feature.

Ready to start building?

Create Your Bot Now