Skip to content

README.md

Latest commit

 

History

History
147 lines (105 loc) · 4.83 KB

README.md

File metadata and controls

147 lines (105 loc) · 4.83 KB

🧠 AI Meeting Intelligence Copilot — Kit README

Part of the Lamatic AgentKit · kits/embed/chat

Deploy with Vercel

An AI-powered Next.js app that converts raw meeting transcripts into structured insights (summary, action items, risks, next steps, follow-up email) and delivers them to Slack — all through a Lamatic chat widget embedded in the UI.

🗂️ Kit Structure

kits/embed/chat/
├── app/
│   ├── page.js                    # Landing page — Server Component
│   ├── layout.js                  # Root layout (Geist font + Vercel Analytics)
│   ├── globals.css                # Tailwind v4 theme + CSS variables
│   └── Screenshots/               # Demo screenshots
│       ├── 1.png
│       ├── fromLamatic-Running.png
│       ├── FromwebPage-With Followup mail-Running.png
│       └── Slack_integrated-Summarizer.png
├── components/
│   ├── LamaticChat.js             # Widget lifecycle — mounts root div + script
│   ├── HeroActions.jsx            # Interactive hero buttons (Client Component)
│   ├── TranscriptPlayground.jsx   # Textarea + Analyze button
│   └── ui/                        # shadcn/ui primitives
├── flows/
│   └── embedded-chatbot-chatbot/  # Exported Lamatic flow JSON
├── .env.local                     # ← create this (see below)
└── package.json

🚀 Local Setup

Prerequisites

  • Node.js 18+
  • A Lamatic account with your meeting intelligence flow deployed
  • A Slack incoming webhook URL (configured in your Lamatic flow)

1. Install dependencies

cd kits/embed/chat
npm install

2. Create .env.local

NEXT_PUBLIC_LAMATIC_PROJECT_ID=your_project_id
NEXT_PUBLIC_LAMATIC_FLOW_ID=your_flow_id
NEXT_PUBLIC_LAMATIC_API_URL=https://your-project.lamatic.dev

All three values are available in Lamatic Studio → Project Settings → Embed Widget.

3. Run

npm run dev
# Open http://localhost:3000

☁️ Deploy to Vercel

# Option A — Vercel CLI
vercel --root kits/embed/chat

# Option B — click the button at the top of this file

Add the same three NEXT_PUBLIC_* env vars in the Vercel dashboard.

After deploying, add your Vercel production URL to the Allowed Domains list in your Lamatic Chat Trigger node, then redeploy the flow.

🔧 Lamatic Flow Configuration

Required nodes (in order)

# Node Purpose
1 Chat Trigger Receives widget messages; whitelist your domain here
2 Generate Text LLM call — extracts summary, action items, risks, next steps, email
3 Generate JSON Structures the LLM output for Slack formatting
4a Slack API Sends formatted card to your Slack channel
4b Chat Response Streams the result back to the widget UI

Domain whitelisting (important)

In the Chat Trigger node config, add * for development or your exact domain for production. Without this, the widget returns 400 on every message.

🖼️ Screenshots

Web App Lamatic Studio
Web Studio
Slack Output Landing Page
Slack Landing

🔑 How the widget integration works

page.js (Server Component)
  └── <LamaticChat />  (Client Component)
        │
        │  On mount (useEffect):
        ├── Creates <div id="lamatic-chat-root"
        │     data-api-url="..."
        │     data-flow-id="..."
        │     data-project-id="..."
        │   /> and appends to document.body
        │
        └── Injects <script type="module"
              src="https://widget.lamatic.ai/chat-v2?projectId=...">
              The widget's React app mounts into #lamatic-chat-root,
              fetches chatConfig, creates an IndexedDB session, and
              is ready to send messages.

Why bootstrap on mount (not on button click)? The widget needs ~500 ms to fetch chatConfig and create a session. Bootstrapping immediately on page load means the widget is fully ready before the user clicks "Open Copilot", preventing the "unexpected error" on the first message send.

📜 License

MIT — see LICENSE.

Built by Vijayshree Vaibhav for the Lamatic AgentKit Challenge.