Consecutive Tool Calls Failing to Populate Name and ID in LangGraph react agent while using MistralAI/Mistral-Large

[nikhil23011996]

Description:
Encountering an issue when using LangGraph React Agent with the MistralAI/Mistral-Large model, particularly during scenarios involving consecutive tool calls alongside astream_events for handling asynchronous events. The tool name and ID are correctly populated for the first call but fail to populate during the second consecutive call, causing incomplete or incorrect execution.

Minimum viable code:

WATSONX_URL = "https://us-south.ml.cloud.ibm.com"
WATSONX_PROJECT_ID = "XXXXXXX"
WATSONX_APIKEY = "XXXXXXX"

from langchain_ibm import ChatWatsonx

Initialize LLM
model = ChatWatsonx(model_id = 'mistralai/mistral-large', #'meta-llama/llama-3-405b-instruct', , #'ibm/granite-13b-instruct-v2', #'ibm/granite-3-8b-instruct', #'mistralai/mixtral-8x7b-instruct-v01', #'
params = {
'decoding_method': 'greedy',
'max_new_tokens': 4000,
'min_new_tokens': 0,
'temperature': 0,
},
project_id = WATSONX_PROJECT_ID,
url=WATSONX_URL,
apikey=WATSONX_APIKEY )

from langchain_core.tools import tool
from typing import Literal

@tool("search")
def search(query: str):
"""Call to surf the web for capital of countries"""
# This is a placeholder, but don't tell the LLM that...
return ["capital of america is washington D.C."]

@tool("get_weather")
def get_weather(city: Literal["nyc", "sf"]):
"""Use this to get weather information."""
if city == "nyc":
return "It might be cloudy in nyc"
elif city == "sf":
return "It's always sunny in sf"
else:
raise AssertionError("Unknown city")

tools = [search, get_weather]

model = model.bind_tools(tools)

Define the graph

from langgraph.prebuilt import create_react_agent

graph = create_react_agent(model, tools=tools)
from langchain_core.messages import HumanMessage, AIMessage, ToolMessage

buffer = ""
async for event in graph.astream_events(
{"messages": [HumanMessage(content=' what is the weather in the NY')]},
{"recursion_limit": 10},
version="v2"
):
event_type = event["event"]

if event_type == "on_chat_model_stream":
    # Handle the main content streaming from the language model
    chunk = event["data"]["chunk"]
    content = chunk.content if hasattr(chunk, "content") else str(chunk)
    buffer += content
    if content:
        # Stream the content chunk immediately
        print (buffer)

Using query which will call multiple tools

async for event in graph.astream_events(
{"messages": [HumanMessage(content='what is the weather in the NY and capital of america')]},
{"recursion_limit": 10},
version="v2"
):
event_type = event["event"]
print(event)
if event_type == "on_chat_model_stream":
# Handle the main content streaming from the language model
chunk = event["data"]["chunk"]
content = chunk.content if hasattr(chunk, "content") else str(chunk)
buffer += content
if content:
# Stream the content chunk immediately
print (buffer)

Error screenshot