AI Agent Use Cases
Views —

AI Agent Use Cases

Added today

44All 189

Prompt to Prepare a Shopping Cart from a Recipe Photo

The author reports that Grok Bot receives a recipe photo, organizes the ingredients and quantities, and creates a shopping list by excluding ingredients already available at home.

Grok Bot
Life & Hobbies

1–20 of 189

Learning & CareersWalkthrough

Alternating Pro and Con Arguments on a Single Topic and Comparing Them

The pro and con roles respond to each other's previous arguments, and the final evaluator role summarizes the key points from both sides.

LangChain

How it works

  1. Insert the debate topic in --topic and set the number of rounds with --rounds. The code limits the number of rounds to between 1 and 4.
  2. The pro role makes a statement, then the con role responds to that content. In the next round, each side rebuts the previous opponent's argument.
  3. The evaluator role gathers the statements from both sides, identifies the strongest arguments, assigns scores, and provides an overall opinion. The results can be read in the terminal.

Results

This is a debate practice example for examining opposing arguments. The final scores and winner are outputs from the evaluator role and are not results of fact-checking.

Notes

The expert names and credentials in the code are fictional role settings. There is no step for searching external sources or verifying the cited evidence of arguments.

Source

GitHub Original post ↗

Example prompt · LangChain

Set up a tool to review the pro and con arguments on [debate topic]. Read the speaking order and evaluation code from agents/20-multi-agent-debate, and prepare the Python packages and API settings. Set the number of debate rounds to [1~4].

Have the two roles read and respond to each other's previous arguments, and have the final role organize each argument, rebuttal, and remaining issues. Do not display names or credentials as if they are real experts. Do not state that fact-checking features are already included if not present in the original text, and leave any factual claims not connected to the [verification materials] you provided as items to be checked.

Test with topics that have no clear answer and with topics that include incorrect premises. Check that roles do not switch, do not repeat the same statements, and that the final summary accurately reflects the previous debate content. The output should include a record of statements by round, a table matching arguments and rebuttals, and a list of items to be fact-checked. Do not present the model's assigned winner or score as an objective determination of truth.