Agents From Scratch

Three implementations of the same travel planning AI agent, each built with a different approach:

File	Approach
agent_python.py	Plain Python — manual agent loop, no framework
agent_langchain.py	LangChain — framework handles the loop
agent_pydantic_ai.py	Pydantic AI — minimal, decorator-based

Each agent can search the web (via Serper) and answer travel planning questions.

---

1. Create a Virtual Environment

python3 -m venv .venv

Activate it:

# macOS / Linux
source .venv/bin/activate

# Windows
.venv\Scripts\activate

---

2. Install Dependencies

pip install -r requirements.txt

---

3. Set Up API Keys

You need two API keys:

• GROQ_API_KEY — free LLM inference. Get one at console.groq.com
• SERPER_API_KEY — Google search API. Get one at serper.dev

Open each Python file and replace the placeholder values near the top (under STEP 0 — Configuration):

os.environ["GROQ_API_KEY"]   = "your-groq-api-key"
os.environ["SERPER_API_KEY"] = "your-serper-api-key"

---

4. Run the Agents

Plain Python agent:

python agent_python.py

LangChain agent:

python agent_langchain.py

Pydantic AI agent:

python agent_pydantic_ai.py