Skip to content

03 — Agentic Systems

End-to-end system designs that combine multiple architectural and design patterns into complete, production-like agentic applications. Each system is implemented in all four frameworks.

What Makes These Different from 02-Architectures

02-Architectures 03-Agentic-Systems
Demonstrates one pattern in isolation Combines multiple patterns into a complete system
Minimal, focused examples Full workflows with realistic agents and tasks
Shows HOW a pattern works Shows HOW to build a system that solves a real problem

Systems

# System Patterns Used Complexity
01-Research-Assistant Multi-topic research → synthesis → critique → report Orchestrator-Subagent + Parallel + Reflexion ★★★
02-Document-Processor Classify → extract → validate → route + HITL gate Pipeline + Conditional Routing + HITL ★★★
03-Autonomous-Task-Planner Goal decomposition → execute → monitor → replan Plan-and-Execute + Hierarchical + Feedback Loop ★★★
04-Code-Review-System Parallel specialized reviews → aggregate → prioritized report Parallel Fan-out + Aggregation + Reflexion ★★★

Framework Implementations

Each system is implemented in all four frameworks:

Framework File Notes
LangChain LangChain/system.py + system.ipynb LCEL composition, AgentExecutor
LangGraph LangGraph/system.py + system.ipynb State graph, native parallelism, checkpointing
CrewAI CrewAI/system.py + system.ipynb Role-based agents, crew orchestration
ADK ADK/system.py + system.ipynb Google Cloud-native, SequentialAgent / ParallelAgent

Setup

cd setup/
pip install -r requirements.txt
cp .env.example .env
# Add your GOOGLE_API_KEY to .env

How to Read These Examples

  1. Read the system's README.md first — it explains the architecture, agent roles, and data flow
  2. Run the LangGraph version for the clearest state visibility (state transitions are explicit)
  3. Compare the CrewAI version to see the role-based approach to the same problem
  4. The ADK version shows Google Cloud-native patterns