ReAct Prompting Animate

← Back to topic

🎞 ANIMATE · Mode 3 of 5 · +100 XP— ReAct Prompting

💭Think

⚡Act

👁️Observe

✅Conclude

🎯 Next: Test →

Step 1 / 4

Speed

What's happening

Reason about the task

Thought: identify what information is needed

💡 Analogy

Like a detective thinking "I need to check the alibi" before calling the witness.

The model first thinks about what it needs to do before taking an action.

Console Output

> Thought: "I need to find the current population of Tokyo."
> Next: plan an action

●react-prompting.ts

Reason about the taskL3–3

1// ReAct = Reasoning + Acting interleaved

2const reactPrompt = `

3Thought: I need to find the current population of Tokyo.

4Action: search("Tokyo population 2024")

5Observation: Tokyo population is approximately 13.96 million.

7Thought: Now I need to compare it with Shanghai.

8Action: search("Shanghai population 2024")

9Observation: Shanghai population is approximately 24.87 million.

11Thought: I can now answer the question.

12Action: finish("Shanghai is larger: 24.87M vs Tokyo's 13.96M")

13`;

15// Each "Action" is dispatched to a real tool:

16async function executeReAct(task: string) {

17 let trace = task;

18 while (!trace.includes('Action: finish')) {

19 const step = await llm.complete(trace);

20 trace += step;

21 const action = parseAction(step);

22 const obs = await tools[action.name](action.args);

23 trace += `\nObservation: ${obs}\n`;

24 }

25}

25 lines

Steps · 4

What's happening

Reason about the task

Thought: identify what information is needed

💡 Analogy

Like a detective thinking "I need to check the alibi" before calling the witness.

The model first thinks about what it needs to do before taking an action.

Console Output

> Thought: "I need to find the current population of Tokyo."
> Next: plan an action

Was this page helpful?

Discussion

← Back to topic

🎞 ANIMATE · Mode 3 of 5 · +100 XP— ReAct Prompting

💭Think

⚡Act

👁️Observe

✅Conclude

🎯 Next: Test →

Step 1 / 4

Speed

What's happening

Reason about the task

Thought: identify what information is needed

💡 Analogy

Like a detective thinking "I need to check the alibi" before calling the witness.

The model first thinks about what it needs to do before taking an action.

Console Output

> Thought: "I need to find the current population of Tokyo."
> Next: plan an action

●react-prompting.ts

Reason about the taskL3–3

1// ReAct = Reasoning + Acting interleaved

2const reactPrompt = `

3Thought: I need to find the current population of Tokyo.

4Action: search("Tokyo population 2024")

5Observation: Tokyo population is approximately 13.96 million.

7Thought: Now I need to compare it with Shanghai.

8Action: search("Shanghai population 2024")

9Observation: Shanghai population is approximately 24.87 million.

11Thought: I can now answer the question.

12Action: finish("Shanghai is larger: 24.87M vs Tokyo's 13.96M")

13`;

15// Each "Action" is dispatched to a real tool:

16async function executeReAct(task: string) {

17 let trace = task;

18 while (!trace.includes('Action: finish')) {

19 const step = await llm.complete(trace);

20 trace += step;

21 const action = parseAction(step);

22 const obs = await tools[action.name](action.args);

23 trace += `\nObservation: ${obs}\n`;

24 }

25}

25 lines

Steps · 4

What's happening

Reason about the task

Thought: identify what information is needed

💡 Analogy

Like a detective thinking "I need to check the alibi" before calling the witness.

The model first thinks about what it needs to do before taking an action.

Console Output

> Thought: "I need to find the current population of Tokyo."
> Next: plan an action