PDF Intelligence — Powered by Llama 3.3 70B (Free & Open Source)

Upload any PDF → get interactive insight cards + streaming chat agent. Visitors bring their own free Groq API key — no backend costs for you.

---

Deploy in 3 minutes (Vercel — free)

# 1. Unzip and install
unzip pdf-agent-app.zip && cd pdf-agent-app
npm install

# 2. Deploy (no env vars needed — users bring their own key)
npm install -g vercel
vercel

That's it. No .env file needed. Users enter their own Groq API key directly in the app.

---

How the user experience works

When someone visits your deployed app:

1. Welcome screen — explains what the app does, feature overview
2. Get key guide — step-by-step instructions to get a free Groq key at console.groq.com
3. Setup screen — paste key, choose model, test it, hit Start
4. Upload screen — drag & drop PDF, settings accessible via ⚙️ button
5. Workspace — Insights tab (cards) + Chat tab (streaming Q&A)
6. Settings modal — change model or key mid-session without losing the PDF

Key is saved in the visitor's localStorage only — never sent to your server. You have zero API costs.

---

User flow diagram

Visit → Welcome → "Get Groq Key" guide → Paste key + pick model
  → Upload PDF → Insight Cards (click to expand, filter by topic)
                → Chat Agent (streaming, markdown, suggested prompts)
                → ⚙️ Settings (change model/key anytime)

---

Models available (all free on Groq)

Model	Speed	Quality	Best for
Llama 3.3 70B	Fast	⭐ Best	General use — recommended
Llama 3.1 8B Instant	⚡ Fastest	Good	Quick answers
Mixtral 8x7B	Fast	Great	Long documents (32k ctx)
Gemma 2 9B	Fast	Good	Google's open model
DeepSeek R1 70B	Medium	⭐ Reasoning	Complex analysis

---

Tech stack

• Next.js 14 (App Router) — framework
• Groq SDK — AI inference (user's own key)
• PDF.js — browser-side PDF parsing (no file hits server)
• SSE streaming — real-time chat responses
• react-markdown — formatted AI responses
• localStorage — key persistence (browser only)
• Vercel — deployment (free tier)

---

Groq free tier limits (per user)

• 6,000 tokens / minute
• 500 requests / day
• No credit card needed
• Sign up at console.groq.com

---

Customization

Add a model: Edit GROQ_MODELS array in components/ApiKeySetup.tsx

More pages: Change doc.text.slice(0, 12000) in InsightsPanel.tsx and ChatPanel.tsx

Card types: Edit TYPE_STYLES in InsightsPanel.tsx

Branding: Edit the app name/colors in app/globals.css CSS variables