AskVAVA

For sales & marketing teams

Customer Engagement (VAVA Engage)

Deploy an AI-powered chatbot on your website to answer customer questions, recommend products, and complete checkouts — all through conversational chat. Self-contained catalogue, Stripe payments, and customizable guardrails included.

1. What is VAVA Engage?

VAVA Engage is a customer-facing chatbot widget you embed on your own website. Unlike internal chat (VAVA Assistant for your team), Engage is designed for anonymous public visitors who don't have AskVAVA accounts.

Key Features

  • Conversational commerce — customers browse products, add to cart, and checkout via chat
  • Self-contained catalogue — manage your products directly in AskVAVA (no Shopify/Woo integration needed)
  • Stripe Connect payments — secure checkout with your own Stripe account
  • Website crawler — train the bot on your site's content automatically
  • Guardrails & branding — control topics, tone, escalation triggers, and off-hours messages
  • Abuse protection — Cloudflare Turnstile, rate limits, cost caps

2. Setup & Deployment

Step 1: Enable Engage

Navigate to Dashboard → Engage. If this is your first time, you'll see a setup wizard. Toggle Enable VAVA Engage to activate the chatbot for your organisation.

Step 2: Configure Branding

Customize how the widget appears to your customers:

  • Widget Title — e.g., "Chat with us"
  • Greeting Message — first message customers see
  • Brand Color — hex color for the chat bubble and UI accents
  • Logo URL — optional logo shown in the chat header

Step 3: Embed the Widget

Copy the embed code snippet from the Engage → Settings page. It looks like:

<script src="https://askvava.com/widget.js"></script>
<script>
  VavaEngage.init({ orgSlug: 'your-org-slug' });
</script>

Paste this before the closing </body> tag on every page where you want the chat bubble to appear. The widget is under 30 KB and loads asynchronously (won't block your site).

Step 4: Train the Bot

VAVA Engage learns from two sources:

  1. Documents & FAQs — Upload PDFs, Word docs, or add FAQ entries in Dashboard → Documents and Dashboard → FAQs. The bot uses RAG to answer questions from these.
  2. Website Crawl — Go to Engage → Knowledge, click Train from Website, and enter your sitemap URL or homepage. AskVAVA crawls up to 200 pages (depth ≤ 2), strips navigation/footer, chunks the content, and embeds it for semantic search.

3. Product Catalogue & Commerce

Adding Products

Navigate to Engage → Products. Click Add Product and fill in:

  • Name — e.g., "Wireless Headphones"
  • Description — used for semantic search (the bot matches customer queries to this)
  • Price — in cents (e.g., 4900 = $49.00)
  • Currency — SGD / USD / MYR / etc.
  • SKU (optional) — for inventory tracking
  • Stock (optional) — leave blank for unlimited; set a number to enforce limits
  • Image URL (optional) — product photo
  • Active — toggle off to hide from the catalogue without deleting

Bulk Import (CSV)

Click Import CSV to upload a spreadsheet. AskVAVA's AI Import will map columns automatically. Supported fields: name, description, price, currency, sku, stock, category.

How Customers Buy

  1. Customer asks: "Do you have wireless headphones?"
  2. Bot uses search_products tool (semantic search over product descriptions) and lists matches with prices
  3. Customer: "Add the blue one to my cart"
  4. Bot uses add_to_cart tool, shows updated cart with subtotal
  5. Customer: "Checkout"
  6. Bot collects email (for receipt), validates stock server-side, creates a Stripe Checkout Session, and sends a hosted payment link
  7. Customer clicks the link, pays via Stripe (card form hosted by Stripe, not AskVAVA), and receives an order confirmation email

Order Management

View paid orders in Engage → Orders. Export to CSV for fulfillment. Stock decrements automatically on successful payment (via Stripe webhook).

🔒 Payment Security

AskVAVA never stores card numbers. Stripe handles all payment data. You receive funds directly to your connected Stripe account (minus a 2.5% platform fee, configurable).

4. Guardrails & Safety

Configuring Guardrails

Navigate to Engage → SettingsGuardrails tab. You can control:

  • Allowed Topics — free-text description of what the bot should discuss (e.g., "electronics, shipping, returns")
  • Blocked Phrases — list of keywords that trigger the refusal message (e.g., "politics", "religion")
  • Refusal Message — what the bot says when asked off-topic (default: "I can only help with topics related to this business.")
  • Tone — friendly / professional / casual / formal (injected into system prompt)
  • Languages — enable Mandarin / Malay / etc. (GPT-4 supports multilingual, but accuracy varies)

Escalation to Human

Set an Escalation Trigger phrase (e.g., "speak to human", "agent"). When detected, the bot collects the customer's email and sends a notification to your Escalation Email address. Your team can follow up manually.

Off-Hours Mode

Define Off-Hours Start/End times (e.g., 6 PM – 9 AM). During these hours, the bot skips AI processing and immediately replies with your Off-Hours Message (e.g., "We're currently closed. Leave your email and we'll get back to you.").

Rate Limits & Abuse Protection

Engage enforces multi-layer rate limits to prevent abuse and control costs:

  • Per-IP: 120 messages / hour (prevents distributed bot floods)
  • Per-session: 30 messages / hour, 200 messages / session (org-configurable)
  • Per-org daily: token + message caps (set in Settings)
  • Cloudflare Turnstile: on first message (blocks automated scrapers)

5. Analytics & Insights

Conversation Logs

View all customer chat sessions in Engage → Conversations. Each session shows:

  • Timestamp, IP, user agent, referrer
  • Full message transcript (user + bot)
  • Products searched / added to cart
  • Order ID if checkout completed

Note: Message content is automatically anonymised after 90 days (PDPA compliance) — timestamps and session metadata are preserved for analytics.

Conversion Tracking

The Engage → Dashboard shows:

  • Total sessions, messages sent, average session duration
  • Conversion rate (sessions → paid orders)
  • Top products by search volume + cart adds
  • Revenue by day/week/month

Exporting Data

Export conversation logs and order history to CSV for further analysis in Excel / Google Sheets / BI tools.

6. AI Features

Semantic Product Search

When you add products, AskVAVA automatically embeds the description with OpenAI text-embedding-3-small. Customer queries are embedded on-the-fly and matched via cosine similarity (pgvector). This means:

  • "Do you have earbuds?" matches "Wireless Bluetooth Headphones" even if "earbuds" isn't in the title
  • "Gifts under $50" filters by price semantically

Smart Recommendations

The bot can proactively suggest products based on:

  • Conversation context (if they mention a category)
  • Recent/popular items (when no context is available)

(Future: collaborative filtering based on "customers who bought X also bought Y".)

Intent Detection

GPT-4 detects customer intent (browsing / buying / support / escalation) and chooses the right tool automatically — no rigid menu buttons required.

7. Stripe Connect Setup

Connecting Your Stripe Account

Before enabling commerce, you need to connect a Stripe account. Navigate to Dashboard → Settings → Payments and click Connect Stripe. You'll be redirected to Stripe to authorize AskVAVA.

How Payments Work

  1. Customer completes checkout in chat → bot creates a Stripe Checkout Session on your Stripe account (via Stripe Connect "destination charge")
  2. Customer pays on Stripe's hosted checkout page (card data never touches AskVAVA's servers)
  3. Stripe sends a webhook to AskVAVA → order marked as paid, stock decremented, confirmation email sent
  4. Funds land in your Stripe account (minus Stripe's processing fee + AskVAVA's 2.5% platform fee)

Platform Fee

AskVAVA charges a 2.5% platform fee on gross order value (default). You can override this per-org in Engage → Settings → Commerce.

💳 Test Mode

Use Stripe's test cards (4242 4242 4242 4242) during development. Switch to live mode in production by connecting a live Stripe account.