High-quality training data is essential for developing effective AI agents. 

, allowing you to curate collections of data specifically for agent training. By tagging examples as "good" or "bad," you help the agent learn from both 

For instance, when reviewing agent outputs, you might find responses that meet your criteria. Tagging these as "good" includes them in the training data as positive examples. If an output doesn't meet your standards, tagging it as "bad" adds it as a negative example, teaching the agent to avoid similar mistakes in the future.

By building datasets with both positive and negative examples, you can improve your agents' performance based on 

 and your specific guidelines.​​​​‌﻿‍﻿​‍​‍‌‍﻿﻿‌﻿​‍‌‍‍‌‌‍‌﻿‌‍‍‌‌‍﻿‍​‍​‍​﻿‍‍​‍​‍‌﻿​﻿‌‍​‌‌‍﻿‍‌‍‍‌‌﻿‌​‌﻿‍‌​‍﻿‍‌‍‍‌‌‍﻿﻿​‍​‍​‍﻿​​‍​‍‌‍‍​‌﻿​‍‌‍‌‌‌‍‌‍​‍​‍​﻿‍‍​‍​‍​‍﻿﻿‌﻿​﻿‌﻿‌​‌﻿‌‌‌‍‌​‌‍‍‌‌‍﻿﻿​‍﻿﻿‌‍‍‌‌‍﻿‍‌﻿‌​‌‍‌‌‌‍﻿‍‌﻿‌​​‍﻿﻿‌‍‌‌‌‍‌​‌‍‍‌‌﻿‌​​‍﻿﻿‌‍﻿‌‌‍﻿﻿‌‍‌​‌‍‌‌​﻿﻿‌‌﻿​​‌﻿​‍‌‍‌‌‌﻿​﻿‌‍‌‌‌‍﻿‍‌﻿‌​‌‍​‌‌﻿‌​‌‍‍‌‌‍﻿﻿‌‍﻿‍​﻿‍﻿‌‍‍‌‌‍‌​​﻿﻿‌​﻿​‌​﻿​​​﻿‌‍‌‍‌‌‌‍‌​​﻿‌‌​﻿​‍​﻿​‍​‍﻿‌​﻿​﻿​﻿​﻿‌‍​‌‌‍‌‌​‍﻿‌​﻿‌​‌‍‌​‌‍‌‍‌‍​‌​‍﻿‌‌‍​‌​﻿​​‌‍‌​​﻿​﻿​‍﻿‌‌‍‌​​﻿​‍‌‍‌‍​﻿‌‍​﻿‌‌​﻿‌‍‌‍‌‌‌‍‌‍​﻿‌​​﻿​​​﻿‌‍​﻿​‌​﻿‍﻿‌﻿‌​‌﻿‍‌‌﻿​​‌‍‌‌​﻿﻿‌‌‍​﻿‌‍‍​‌‍​‌‌‍﻿‍‌‍‌﻿‌‍‌‌‌‍﻿​‌‍﻿﻿‌‍‌﻿‌​‍‌‌﻿‌​‌‍‌‌‌‍﻿‌​﻿‍﻿‌﻿​​‌‍​‌‌﻿‌​‌‍‍​​﻿﻿‌‌‍﻿​‌‍﻿﻿‌‍‌﻿​‍‌‌​﻿‌‌‌​​‍‌‌﻿﻿‌‍‍﻿‌‍‌‌‌﻿‍‌​‍‌‌​﻿​﻿‌​‌​​‍‌‌​﻿​﻿‌​‌​​‍‌‌​﻿​‍​﻿​‍‌‍​‌​﻿‌‍‌‍​‌​﻿​​​﻿​﻿‌‍‌‌​﻿‌‍​﻿​‍‌‍​‌‌‍​‍‌‍​‍​﻿​‍​‍‌‌​﻿​‍​﻿​‍​‍‌‌​﻿‌‌‌​‌​​‍﻿‍‌‍﻿‌‌‍‌​‌﻿‍​​﻿﻿﻿‌‍​‍‌‍​‌‌﻿​﻿‌‍‌‌‌‌‌‌‌﻿​‍‌‍﻿​​﻿﻿‌​‍‌‌​﻿​‍‌​‌‍‌﻿​﻿‌﻿‌​‌﻿‌‌‌‍‌​‌‍‍‌‌‍﻿﻿​‍‌‍‌‍‍‌‌‍‌​​﻿﻿‌​﻿​‌​﻿​​​﻿‌‍‌‍‌‌‌‍‌​​﻿‌‌​﻿​‍​﻿​‍​‍﻿‌​﻿​﻿​﻿​﻿‌‍​‌‌‍‌‌​‍﻿‌​﻿‌​‌‍‌​‌‍‌‍‌‍​‌​‍﻿‌‌‍​‌​﻿​​‌‍‌​​﻿​﻿​‍﻿‌‌‍‌​​﻿​‍‌‍‌‍​﻿‌‍​﻿‌‌​﻿‌‍‌‍‌‌‌‍‌‍​﻿‌​​﻿​​​﻿‌‍​﻿​‌​‍‌‍‌﻿‌​‌﻿‍‌‌﻿​​‌‍‌‌​﻿﻿‌‌‍​﻿‌‍‍​‌‍​‌‌‍﻿‍‌‍‌﻿‌‍‌‌‌‍﻿​‌‍﻿﻿‌‍‌﻿‌​‍‌‌﻿‌​‌‍‌‌‌‍﻿‌​‍‌‍‌﻿​​‌‍​‌‌﻿‌​‌‍‍​​﻿﻿‌‌‍﻿​‌‍﻿﻿‌‍‌﻿​‍‌‌​﻿‌‌‌​​‍‌‌﻿﻿‌‍‍﻿‌‍‌‌‌﻿‍‌​‍‌‌​﻿​﻿‌​‌​​‍‌‌​﻿​﻿‌​‌​​‍‌‌​﻿​‍​﻿​‍‌‍​‌​﻿‌‍‌‍​‌​﻿​​​﻿​﻿‌‍‌‌​﻿‌‍​﻿​‍‌‍​‌‌‍​‍‌‍​‍​﻿​‍​‍‌‌​﻿​‍​﻿​‍​‍‌‌​﻿‌‌‌​‌​​‍﻿‍‌‍﻿‌‌‍‌​‌﻿‍​​‍​‍‌﻿﻿‌

Memorang changelog

EdFlow: Enhancing Agent Training with Curated Datasets