Post
176
Agents & MCP Hackathon Day 3 & 4
Yesterday I actually forgot to post an update. Sorry for that! Basically, I fixed some critical bugs and worked on some new ideas.
Today, I changed the search agent from being a participant in the discussion to a helper for AI models that support function calling. So while the search agent "sits" on the table, it only reacts if Mistral or Llama need to search the web or Wikipedia.
Additionally, I reduced the length of the replies and adjusted the context each model gets to get better discussions from them. I think this really improves the results.
I would love for you all to check it out, but right now my space is stuck in rebuilding. Will report back once that's fixed. In the meantime, let me know what you all are building in the comments.
https://huggingface.co/spaces/azettl/consilium_mcp
Yesterday I actually forgot to post an update. Sorry for that! Basically, I fixed some critical bugs and worked on some new ideas.
Today, I changed the search agent from being a participant in the discussion to a helper for AI models that support function calling. So while the search agent "sits" on the table, it only reacts if Mistral or Llama need to search the web or Wikipedia.
Additionally, I reduced the length of the replies and adjusted the context each model gets to get better discussions from them. I think this really improves the results.
I would love for you all to check it out, but right now my space is stuck in rebuilding. Will report back once that's fixed. In the meantime, let me know what you all are building in the comments.
https://huggingface.co/spaces/azettl/consilium_mcp