Chatbot responds slowly when fetching data from connect action API or MCP

General Questions
, ,
1

Hello everyone, has anyone tried using Pickaxe to create support agent chatbots that interact with connect actions very slowly, causing a poor user experience?

In the token lengths configuration section, there are parameters like max output length, max input length, memory buffer - do I need to select specific parameters or make special customizations to make my chatbot respond faster?

Or are there any tips or solutions to make the bot respond quickly to questions when connecting tools via API or MCP?

I have tested with ChatGPT API, Claude API, Grok API and all give relatively slow response results. From the API side, I have shortened the response content but I still see that the time to check and receive responses back is very long.

2

Hi, @chipvn this could be caused by a host of different reasons.

Prompt engineering quality is a factor. Both the Role prompt as well as your trigger action prompts can affect how well your actions interact.

The context window/server speed of the underlying LLM you chose for your Pickaxe may also be a factor.

If you’re using Make.com and there are several modules within your scenario, that may also be causing a delayed output due to data processing load within Make.

To know for sure, if you can provide us with screenshots with some accompanying notes, or better yet, a Loom video showing internal workings of your Pickaxe, that can help community members better diagnose the issue and assist you.