Thanks, still some work to do around agent.py and its handling of JSONs and tool calling and sync processes. My error 400 context size came back after I fixed agent.py in 1.6 but it should only require a minor fix in 1.7 I think I haven't looked over all of it yet. I run local LLMS for all agents.

After some work today I was able to clean up the hard fail condition for the commonly dreaded Error 400 on the Utility model context size exceeded. I added some lines to agent.py that now allows for a graceful exception capture, displays a msg to the user saying maybe its time to start a new chat or make your request smaller. It no longer grinds the program to a dead crash, forcing a new chat to be started. Now I can carry on within the same chat if I hit my context limit (rare). I could not get to work the displaying of a banner when the web utility is called. I tried for awhile added some logic that I thought would work but it didn't work even though my test debug lines were being displayed in the logs where expected. Not worth the effort to pursue. Moving on to having it do machine tasks on a remote machine! Now the real fun begins!

What are ya trying to do w/A0? Fun? work? "Get rich quick"?

lol which one?!

@Pranesh Patel Let me know what you are planning in regards to automation, thats on my big to-do list for my project. Maybe we can collab.

@Chong Zi Xuan we should chat. Maybe you can help me out. Check me out.

Neat. I run all local LLMs so cost = 0. I think most people pick models incorrectly doesn't matter cloud/local. I think a lot of people assume bigger model = better. But many don't follow the golden rule of all ai, pick a model that fits your use case. I've seen people using A0's utility model in the 120B class. It's a utility model it doesn't need that!! They incorrectly understand how A0 works. I mean you do you, but if you are paying for this stuff (cloud) I hope you are really considering your use case and finding the smallest model to meet your needs. Speed, accuracy and concurrency are driving factors. I'm no AI expert at all, not even close!! But I had that mindset when I first started, 'oh let me get the biggest one, or use up as much of the RAM as I can' nah wrong approach. What am I doing, what do I want to achieve, what models are the smallest that will do the job quickly and correctly. Which 'should' be cheaper for those on the cloud wagon. But like I said you do you! (no I'm not talking to you directly I know you are just sharing info) Cheers!