AI-Agent-Flow

A hands-on collection of working AI agent examples across seven frameworks — LangChain, CrewAI, LangGraph, Pydantic AI, AutoGen, Agno, and smolagents.

The Problem With Learning Agent Frameworks

Every major AI agent framework has its own mental model, API style, and tradeoffs. Reading docs only gets you so far — you need runnable, real-world examples to understand when to reach for CrewAI versus LangGraph, or why type-safe agents with Pydantic AI matter for production systems. Most tutorials cover hello-world; few show frameworks doing actual work.

Working Examples, Not Toy Demos

Each module in this repo solves a real problem using a specific framework. They run end-to-end, connect to live APIs, and demonstrate patterns you'd actually use in production: multi-agent collaboration, RAG with memory, conditional workflow routing, structured output validation, email alerting, and web scraping with deduplication.

Example: Stock Alert System with CrewAI

Two agents collaborate — a Researcher fetches current and previous prices from Yahoo Finance, and an Analyst calculates the change and sends an email alert if a threshold is crossed:

# Configure your stocks and alert threshold
vim src/crewai_stock_alert_system/config.py

# Run the multi-agent crew
python src/crewai_stock_alert_system/run_stock_alert.py

The Researcher and Analyst agents hand off context automatically. If AAPL drops more than 2%, an email goes out.

Usage

Prerequisites

• Python 3.10+
• uv package manager

Setup

# Install dependencies
uv sync

# Copy and fill in your API keys
cp .env_sample .env

Required environment variables (see .env_sample):

Variable	Purpose
OPENAI_API_KEY	Required by all agents
WEATHER_API_KEY	OpenWeatherMap (RAG agent)
EMAIL_SENDER	Gmail address (stock alerts)
EMAIL_PASSWORD	Gmail app password (stock alerts)
EMAIL_RECEIVER	Alert recipient (stock alerts)

Run Any Example

Agent	Framework	What It Does	Command
Reasoning transparency	Agno	Shows GPT thinking step-by-step	python src/agno_hello/hello_agno.py
Multi-agent dev team	AutoGen	CodeGen + Tester agents write and critique code	python src/autogen_dev_team/create_sorting_algorithm.py
Stock alert system	CrewAI	Monitors prices and sends email alerts	python src/crewai_stock_alert_system/run_stock_alert.py
Weather RAG agent	LangChain	Conversational Q&A over live forecast data	python src/langchain_rag_agent/rag_agent.py
Branching workflow	LangGraph	Routes inputs to research, analysis, or escalation	python src/langgraph_branching_agent/run_branching_agent.py
News analyzer	Pydantic AI	Sentiment, topics, and scoring with type-safe outputs	python src/pydantic_ai_example/news_analyzer.py
News scraper	Pydantic AI	Fetches RSS feeds and stores validated articles in SQLite	python src/type_safe_news_agent/run_news_agent.py
Web search agent	smolagents	Answers questions using live web search	python src/smolagents_hello/hello_smolagents.py

Note: The news analyzer reads from the database created by the news scraper. Run run_news_agent.py first.

Project Layout

src/
├── agno_hello/                  # Agno reasoning example
├── autogen_dev_team/            # AutoGen multi-agent code collaboration
├── crewai_stock_alert_system/   # CrewAI stock monitoring with email
├── langchain_rag_agent/         # LangChain RAG with FAISS + conversation memory
├── langgraph_branching_agent/   # LangGraph conditional routing
├── pydantic_ai_example/         # Pydantic AI news analysis
├── smolagents_hello/            # smolagents web search
└── type_safe_news_agent/        # Pydantic AI news scraper + SQLite
docs/                            # Per-framework writeups