🤖 CodePilot AI Studio — Google Colab Edition

Agentic AI in Software Engineering | Module 4 Instructor: Prof. Apparsamy Perumal | Semester 5

---

What Is This?

CodePilot AI Studio builds a real AI coding assistant stage by stage in Google Colab.

Zero installation. No admin rights. Works on any browser with a free Groq API key.

Stage	Concept	Session
Stage 1	Hello Groq — talk to Llama 3 via API	✅ In Class
Stage 2	Memory — agent remembers your conversation	✅ In Class
Stage 3	Tools — intent routing to specialist skills	✅ In Class
Stage 4	RAG — long-term memory with ChromaDB	📚 Homework
Stage 5	Reflection — agent improves its own output	📚 Homework
Stage 6	LangGraph — stateful graph workflow	📚 Homework

---

⚡ Before Class — MUST DO (5 minutes)

You need a free Groq API key. Do this before class day.

Step 1 — Create a free Groq account

1. Go to: https://console.groq.com
2. Click Sign Up (top right)
3. Sign up with your Google account or email — no credit card needed

Step 2 — Create your API Key

1. After logging in, click "API Keys" in the left sidebar
2. Click "Create API Key"
3. Name it anything (e.g. codepilot)
4. Click Submit
5. Copy the key immediately — looks like: gsk_xxxxxxxxxxxxxxxxxxxx
6. Save it in Notepad or your phone Notes

⚠️ The key is shown only ONCE. If you lose it, just create a new one — takes 30 seconds.

Step 3 — Store key in Colab Secrets (saves you typing it every notebook!)

💡 Colab Secrets stores your key ONCE and all 6 notebooks find it automatically.

1. Go to https://colab.research.google.com
2. Open any notebook
3. Click the 🔑 key icon in the left sidebar
4. Click "Add new secret"
5. Name: GROQ_API_KEY ← must be exactly this
6. Value: paste your key gsk_xxxx...
7. Toggle "Notebook access" to ON
8. Done! Every notebook will now find your key automatically

Each student must use their own key — never share keys.

---

🚀 How to Open and Run a Notebook

Step 1 — Open from GitHub

1. Go to this repository on GitHub
2. Click on any .ipynb file (e.g. stage1_hello_groq/Stage1_Hello_Groq.ipynb)
3. Click the orange "Open in Colab" badge at the top of the file
4. The notebook opens directly in Google Colab

Step 2 — Run the notebook

1. Cell 1 (API Key): Run it — if you set up Colab Secrets, it loads automatically. If not, paste your key manually between the quotes.
2. Cell 2 (Setup): Run it — installs packages (~30 seconds first time)
3. Press Ctrl+F9 to run all remaining cells at once
4. Read the output below each cell
5. Try the experiments at the bottom

⏱ First run: ~1 minute (package install). All runs after that: instant.

---

🔑 Two Ways to Add Your API Key

Option A — Colab Secrets (Recommended — set up once, works everywhere)

Step	What to do
1	Click the 🔑 key icon in the Colab left sidebar
2	Click "Add new secret"
3	Name: GROQ_API_KEY
4	Value: paste your gsk_xxxx... key
5	Toggle "Notebook access" ON
Done!	All 6 notebooks load your key automatically

Option B — Paste manually (quick fallback)

If Secrets is not set up, find this line in Cell 1:

GROQ_API_KEY = "paste-your-groq-key-here"

Replace paste-your-groq-key-here with your actual key. Works immediately.

---

📁 Repository Structure

codepilot-colab/
│
├── README.md                          ← this file (start here)
│
├── stage1_hello_groq/
│   ├── Stage1_Hello_Groq.ipynb        ← Colab notebook (click Open in Colab)
│   └── Stage1_README.md               ← detailed concept guide
│
├── stage2_memory_agent/
│   ├── Stage2_Memory_Agent.ipynb
│   └── Stage2_README.md
│
├── stage3_tool_agent/
│   ├── Stage3_Tool_Agent.ipynb
│   └── Stage3_README.md
│
├── stage4_rag_explain/
│   ├── Stage4_RAG_Explain.ipynb
│   └── Stage4_README.md
│
├── stage5_reflection/
│   ├── Stage5_Reflection.ipynb
│   └── Stage5_README.md
│
└── stage6_langgraph/
    ├── Stage6_LangGraph.ipynb
    └── Stage6_README.md

---

🛠️ Tech Stack — All Free

Tool	Purpose
Google Colab	Browser-based Python — no install needed
Groq API	Fast LLM cloud service (free tier)
Llama 3.1 8B	Open-source AI model by Meta
LangChain	Python AI agent framework
ChromaDB	Vector database for RAG (Stage 4)
HuggingFace Embeddings	Text-to-vector conversion (Stage 4)
LangGraph	Graph-based workflow (Stage 6)

---

❓ Troubleshooting

"Key not found in Secrets / AuthenticationError"

Your GROQ_API_KEY is wrong or Secrets not set up. Go to https://console.groq.com → create a fresh key → add to Secrets OR paste manually.

"Notebook access is OFF in Secrets"

Click the 🔑 icon → find GROQ_API_KEY → toggle Notebook access to ON → re-run Cell 1.

"ModuleNotFoundError"

Run Cell 2 (the pip install cell) again. Sometimes Colab resets between sessions.

"Session expired"

Colab sessions expire after ~12 hours. Re-open the notebook and run from Cell 1. Your Groq key in Secrets is still there — no need to add it again.

"I lost my Groq API key"

Go to https://console.groq.com → API Keys → Create a new one → update Secrets value.

---

📊 What You Will Build

By Stage 6 you will have built a complete AI coding assistant that can:

• Talk to a fast cloud LLM (Groq + Llama 3.1)
• Remember your full conversation across multiple turns
• Detect what you want (debug / explain / review) automatically
• Retrieve relevant knowledge before answering (RAG)
• Critique and improve its own output (reflection)
• Execute a stateful workflow with loops and branching (LangGraph)

---

🎓 Credits

Built for Agentic AI Software Development — Module 4 Department of Computer Science & Engineering | Semester 5 Prof. Apparsamy Perumal

Tech: Groq · LangChain · LangGraph · ChromaDB · HuggingFace