Graph hangs in "pending state" in LangGraph Studio when using Google LLM ( ChatGoogleGenerativeAI )

[JulianAlvarezdeGiorgi]

❗️Issue: Graph hangs in "pending state" in LangGraph Studio when using Google LLM

Summary

When using a Google LLM (ChatGoogleGenerativeAI) in my graph, the graph enters a "pending state" and never resolves when submitting messages in LangGraph Studio. However, when using the same graph outside of Studio, everything works fine.

Environment

• LangGraph Studio
• create_react_agent used to build the graph
• LLM provider: Google (Gemini)
• ✅ Works fine with: Groq
• ✅ Works fine outside LangGraph Studio with Google

Code Snippet

graph = create_react_agent(llm, tools=tools, state_modifier=system_message)

Behavior Observed

• The graph runs perfectly outside LangGraph Studio, even when using Google as the LLM.
• When run inside LangGraph Studio:
  • The graph enters a "pending state".
  • Pressing “Join run” causes it to hang indefinitely and never resolves.
• Switching the LLM provider to Groq inside LangGraph Studio:
  • ✅ The graph runs as expected and completes successfully.

Screenshot: Pending State in LangGraph Studio

Below is a screenshot showing the issue — the graph is stuck in a pending state, and the "Join run" button appears but doesn't resolve the run, it keeps running forever after pressing it:

Troubleshooting Done

• ✅ Tested with Groq in Studio → Works fine.
• ✅ Tested with Google outside Studio → Works fine.
• ❌ Google LLM inside Studio → Still hangs in pending state.

Hypothesis

This issue seems related to compatibility between LangGraph Studio’s execution environment and the Google LLM (Gemini). Possible causes might include:

• A tool call or step that never resolves due to differences in Studio's runtime.
• LangGraph Studio failing to acknowledge completion of a tool or LLM call when using Google.

Expected Behavior

The graph should execute normally inside LangGraph Studio, just as it does outside, regardless of the LLM provider (Google or Groq).

---

Any help or insights would be greatly appreciated! 🙏

[hinthornw]

Are there errors in your terminal?

[JulianAlvarezdeGiorgi]

Hi, thank you for replying!
There's no error displayed — the graph is stuck in a pending state and keeps running at the agent node (I also confirmed this in LangSmith).

[hinthornw]

Please share a minimal reproducible example so we can help.

[JulianAlvarezdeGiorgi]

Sure! This is not what I'm implementing, but I tested this minimal example and it shows the same behavior. It seems like the issue might be related to how LangGraph Studio invokes the LLM, maybe?

Enviroment and Dependencies

• Python version: 3.13.2
• langchain: 0.3.22
• langgraph: 0.3.25
• langchain-google-genai: 2.1.2
• langchain-groq: 0.3.2

Code

from dotenv import load_dotenv

# Load environment variables from the .env file
load_dotenv()

from langchain_google_genai import ChatGoogleGenerativeAI
from langchain_groq import ChatGroq

if True:
    llm = ChatGoogleGenerativeAI(model='gemini-2.0-flash', temperature=0)
else:
    llm = ChatGroq(model='llama-3.3-70b-versatile', temperature=0)

from typing import Literal

from langchain_core.tools import tool


@tool
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 = [get_weather]


# Define the graph

from langgraph.prebuilt import create_react_agent

graph = create_react_agent(llm, tools=tools)

[hinthornw]

The "continue" issues is usually because an error is thrown in the graph - i'd check the server logs and/or langsmith traces to see it

[RayyanMasood]

Same issue