about the "high-level" implementation question

[XIAOke8698]

In the "high-level" implementation of the home page, why did the AI conversation not load tools into funtioncall when I asked for the second time what the weather was like? As a result, the tool was not invoked in the second session, although I used the memory module as well.

[laelhalawani]

Your question has no information on how did you actually do it so it's impossible for us to guess what did you do wrong.
Paste your code or snippets.

Generally speaking, make sure you save llm with bound tools
llm_with_tools = llm.bind_tools(your_tools_list)

then use the bound llm for invocation `llm_with_tools.invoke(messages)``
or bind them with each call when you invoke.
Also the models can decide not to use the tools in the react. Idk what you did to be any more specific.

[XIAOke8698]

Thank you. I found the problem because my LLM has two nodes, which are forwarded by Nginx and loaded to different nodes. One node is not started in the function_call format. As a result, the tool list cannot be read.