Xavier

any news about this integration ?

Do you know what had been done and what are the missing pieces to make it happen? Le 31 mars 2014 04:41, "Michael Ekstrand" [email protected] a écrit : > @doxav...

Thank you Michael. I asked to someone of my team to go through step 1 to 4. I'll get back to you. Regards, Xavier On Mon, Mar 31, 2014 at...

A more recent approach from Anthropic: https://github.com/anthropics/ConstitutionalHarmlessnessPaper seems a better approach

At least, does any metric exist in AgentGPT to measure distance to goal ? Then we could define an exit strategy and an output format to deliver something in line...

This approach could cover the need to track status. However the main issue is the lack of "distance to goal" or value function to goal to measure progress and convergence,...

The PagedAttention mechanism from vLLM may help https://github.com/vllm-project/vllm/blob/49b26e2cec8c56594668905e853fe4af34336b05/vllm/model_executor/layers/attention.py#L16

> small pull request which just allows people to easily use their own proxies allowing for use of free versions of openai, and allows use for better management of keys...

@RayBytes Could you provide examples of those free APIs ?

I think top features should be: 1. Autonomous execution management (choose between manual confirmation vs N autonomous iterations, allow manual early stopping of autonomous iterations without exiting) 2. Open API...