ReAct Pattern
Master the ReAct pattern to build intelligent agents that synergize reasoning and acting
Your Progress
0 / 5 completedWhat is the Action Phase?
After reasoning about what to do, agents enter the action phase where they execute concrete operations. Actions are the bridge between thinking and doing—they interact with:
- External APIs - Search engines, databases, web services
- Computation Tools - Calculators, code interpreters, simulators
- Data Sources - Files, knowledge bases, real-time feeds
- User Interfaces - Forms, notifications, confirmations
Every action produces an observation—a result that feeds back into the next reasoning phase, creating the ReAct loop.
Interactive: Execute Actions
Click on actions to see them execute and observe the results
Click an action above to execute it
Thought → Action → Observation
💭
1. Thought
"I need to find the population of Tokyo to answer this question"
⚡
2. Action
Search(query="Tokyo population 2024")
👁️
3. Observation
Result: "Tokyo metropolitan area has approximately 37.4 million people"
Why Actions Matter in ReAct
🌍
Ground in Reality
Actions connect reasoning to real-world data, preventing pure speculation
📊
Provide Feedback
Observations validate or refute hypotheses, enabling adaptation
🔄
Enable Iteration
Each action's result informs the next thought, creating a learning loop
⚙️
Extend Capabilities
Tools and APIs give agents powers beyond language understanding alone