Just reposting my messages from the last 2 days in hopes that someone responds.... not sure why the agents just breaks down, connections stop working and stop producing the desired output (becomes something the agent wants to produce over time instead of what I am loking for...) Any help is appreciated, it's been alot of time and credits to reach this point. So again, any support is appreciate. If support isnt available, knowing that it also helpful
Kevin B., getting the agent step to perform exactly as we want can prove to be a bit challenging in terms of prompting as well as limitations on the technical side of things. I do would recommend building on rails as much as you can and only outsource things to the agent step when it is absolutely necessary. I have is the past, run into somewhat of a similar error as you described in your second point. To that, I would recommend you use the set variable action and prompt it to extract the exact things (in markdown preferably with examples) and pass it along to the next step and use that variable's output for the final message. I have attached an example for your reference.
I seems to have resolve the problem, but I feel there were two problems. 1. The scraping of twitter data via api just stopped working, no issues with api, seems resolved now (used grok to make a prompt to fix it). 2. The output that I am looking for in very specific, and what has happened many time is that the agent provides an overview instead of the specific information I need (link, quote, etc). I am just trying to understand how and why the process that;s working, suddenly isnt. Expressed different, I've found Lindy very helpful in doing alot of tasks, but it would be more helpful if the processes that work working continues to work as I develop new agents