[P1] Add Chatbot Suggested Prompts / Quick Action Buttons

[flexykrn]

Bug Description

The AI documentation assistant (RAG chatbot) presents users with a blank text input on first open. New users don't know what they can ask or where to start. This creates friction and reduces engagement. Competitor chatbots and modern AI interfaces use suggested prompts to guide users and demonstrate capabilities.

Competitor Benchmark

• Intercom: 'What can I help you with?' + 3 suggested topics
• Crisp: Quick replies as clickable chips above input
• ChatGPT: Conversation starters for new chats
• Claude: Suggested prompts based on context
• Stripe AI: 'Get started' quick actions (refunds, invoices, disputes)

Affected Users

• First-time visitors discovering the chatbot
• Users unsure what questions the chatbot can answer
• Developers looking for quick entry points to common tasks

Blast Radius

• Primary: All chatbot users (estimated 30%+ of doc visitors)
• Secondary: New developers who need guidance on where to start
• Impact: Higher engagement, faster time-to-answer, better onboarding

Root Cause Analysis

Chatbot UI opens with empty state. No onboarding flow or capability hints. Users must guess what the chatbot knows about.

Proposed Solution

Option A: Static Quick Actions (Recommended)

Show 4-6 clickable chips above input:

Quick Actions:
[Deploy my first contract] [Set up XDC wallet]
[Run a validator node] [Connect to testnet]
[Token standards] [Cross-chain bridge]

Option B: Context-Aware Suggestions

Change suggestions based on current page:

• On /smartcontract/: 'How do I verify my contract?'
• On /subnet/: 'How do I deploy a subnet?'
• On /api/: 'What are the RPC endpoints?'

Additional Fixes Required

• Design chip/button component for chatbot UI
• Define initial set of 6-8 quick actions
• Track click-through rates on each prompt
• A/B test static vs context-aware suggestions
• Mobile: horizontal scrollable chips

Acceptance Criteria

• Chatbot shows suggested prompts on first open
• Prompts are clickable (not just text)
• Clicking a prompt submits it as user message
• Context-aware on relevant pages (if implemented)
• Mobile responsive (horizontal scroll or grid)
• Dismissible / hideable by user
• Analytics tracking on prompt usage

Raw Context

Current chatbot opens to empty input. Users type 'hi' or 'hello' instead of useful questions. Quick actions would demonstrate value immediately and guide users to common tasks.