[BUG] Conversational Agent Unable To Determine What Tool To Use ?

[DebajitKumarPhukan]

Describe the bug
I tried a Local LLM (LLama2) as the LLM backing the Conversational Agent. The Conversational Agent has 2 tools attached: Retrieval QA Chain Tool and a Serper. I asked for today's date to the agent for which it failed miserably and returned some old date. When I specifically asked to use Server then it managed to search the internet to find the date but failed in json parsing.

The above behavior is reproducible in any local LLM or hugging face LLM.

To Reproduce
Steps to reproduce the behavior:

1. Load the attached Chatflow.
2. Run any local LLM and use the API as base URL.
3. Set Up your free Serper account and configure the API key in the ChatFlow.
4. Ask What is today's date ?
5. See error

Expected behavior

1. The conversational agent shouldn't respond back with any "parse error".
2. There should be a provision to describe about the "Serper" tool so that we can provide the description to the conversational agent to better understand when to use.

Screenshots

Flow
If applicable, add exported flow in order to help replicating the problem.
Local LLM Chatflow - Issues.json

Setup

• Installation [yarn start]
• Flowise Version [1.4.1]
• OS: [Windows]
• Browser [chrome]

Additional context
Please find attached logs for your reference which will help you to find the areas when the error is thrown.
Local LLM Error Logs.log

[DebajitKumarPhukan]

@HenryHengZJ Please merge this PR when you get sometime which will resolve this issue (default older version of Langchain had issues).
https://github.com/FlowiseAI/Flowise/pull/1255

[HenryHengZJ]

should be resolved after langchain version upgrade - #1281