Thank you for making this great system available open source!
So, it could print out periodically like every time it exceeds another X dollars or cents, or after every API call, some estimated API cost analytics, possibly including all-time total expense, total expense for this run, and expense of last API call, as a 3-tuple or something.
The ability to halt before a certain cost limit has been reached.
I would please really like to see some of the redacted features put back into this or a forked repo, because not everyone who has a valid use of those features has the ability to reimplement those features.
Integration with an existing to-do list system or API. I have about 200,000 tasks recorded in a planning system.
The ability to utilize a local LLM instead of GPT-4, for cost and infosec reasons.
Some kind of FIFO assurances, so that tasks are processed in sequence.
Some Jason/AgentSpeak(L) event loop architectural features (if any are missing): see:
Thank you for making this great system available open source!
So, it could print out periodically like every time it exceeds another X dollars or cents, or after every API call, some estimated API cost analytics, possibly including all-time total expense, total expense for this run, and expense of last API call, as a 3-tuple or something.
The ability to halt before a certain cost limit has been reached.
I would please really like to see some of the redacted features put back into this or a forked repo, because not everyone who has a valid use of those features has the ability to reimplement those features.
Integration with an existing to-do list system or API. I have about 200,000 tasks recorded in a planning system.
The ability to utilize a local LLM instead of GPT-4, for cost and infosec reasons.
Some kind of FIFO assurances, so that tasks are processed in sequence.
Some Jason/AgentSpeak(L) event loop architectural features (if any are missing): see:
https://www.mdpi.com/electronics/electronics-10-02136/article_deploy/html/images/electronics-10-02136-g004.png
https://www.wiley.com/en-us/Programming+Multi+Agent+Systems+in+AgentSpeak+using+Jason-p-9780470061831
The possibility for a human to enter and exit the loop, via some kind of dialogue system.