Occasionally I encounter a build action that cannot be completed by Replit because it requires manual work such as reading a pdf to extract data. This is where I often use ChatGPT in agent mode, which is designed to carry out tasks for you. I ask the Replit agent to draft a prompt for ChatGPT to complete the task and then paste it in to that agent. It’s an amazingly useful was to break through an issue like this.
I’m not familiar with all the AI services out there, and there might be another one that you use that can complete the task as well. I’m interested, have others used this type of process to extract data?
1 Like
Yes. What is better than one or two agents running in tandem in Replit? 5 agents running in ChatGpt or Claude or Gemini. There are some issues that need a neutral observer to see the probleem and resolve it without breaking your working app. At times replit agent takes a wrong turn and cascades numerous issues through your working app. The only way back is not always a roll back. Because, the agent makes mistakes in coding at times a rollback want fix. That is why you need to have the Replit architect check what the Replit agent codes on every session. Make small changes then check to see the result. Test the new feature one at a time.
2 Likes