Unable to connect to Lang Smith Server to the local dev (Urgent)

[kshadow07]

Checked other resources

• This is a bug, not a usage question.
• I added a clear and descriptive title that summarizes this issue.
• I used the GitHub search to find a similar question and didn't find it.
• I am sure that this is a bug in LangChain rather than my code.
• The bug is not resolved by updating to the latest stable version of LangChain (or the specific integration package).
• This is not related to the langchain-community package.
• I posted a self-contained, minimal, reproducible example. A maintainer can copy it and run it AS IS.

Package (Required)

• langchain
• langchain-openai
• langchain-anthropic
• langchain-classic
• langchain-core
• langchain-cli
• langchain-model-profiles
• langchain-tests
• langchain-text-splitters
• langchain-chroma
• langchain-deepseek
• langchain-exa
• langchain-fireworks
• langchain-groq
• langchain-huggingface
• langchain-mistralai
• langchain-nomic
• langchain-ollama
• langchain-perplexity
• langchain-prompty
• langchain-qdrant
• langchain-xai
• Other / not sure / general

Example Code (Python)

from ast import mod
import os
import asyncio
from typing import Literal
from dotenv import load_dotenv
from deepagents import create_deep_agent
from langchain.chat_models import init_chat_model

# Load environment
load_dotenv()

universal_coder = {
    "name": "atomic_coder",
    "description": "Executes ONE atomic coding task. Blind to the larger project context.",
    "system_prompt": """You are an atomic code executor. You receive ONE specific coding instruction.
     use tool write_files to write in files
     use tool edit_file to edit in files 
     use tool glob to know where is the location to edit
     use tool grep to edit files also

YOUR CONSTRAINTS:
- You do NOT know what project this is for
- You do NOT know what other tasks are being done
- You ONLY execute the exact instruction given
- Output clean, working code with no commentary

EXECUTION RULES:
1. If asked to "Write function X" → Write ONLY that function
2. If asked to "Create file Y.html" → Create ONLY that file
3. If asked to "Fix bug Z" → Fix ONLY that bug
4. Use write_file to save output
5. Keep responses under 150 lines

LANGUAGE SUPPORT:
- Detect language from instruction (Python, JavaScript, HTML/CSS, etc.)
- Use modern syntax and best practices
- Add minimal comments only for complex logic

DO NOT:
- Ask clarifying questions
- Suggest improvements to the overall architecture
- Explain the "why" - just execute the "what"
""",
    "model": "openai:gpt-4o-mini"  # Cheap model as per MAKER
}

# Model configuration
model = init_chat_model(
    model="openai:gpt-5-mini",
    streaming=False  # Disable streaming to avoid verification issues
)

manager_system_prompt = """ You are a Coding Project Manager using the MAKER framework.

YOUR WORKFLOW:
1. DECOMPOSE: Break user's request into 10-50 ATOMIC tasks using 'write_todos'
   - Each task = 1 function OR 1 file OR 1 bug fix
   - Make tasks PARALLEL-SAFE (no dependencies between tasks in same batch)
   
2. DELEGATE: Call 'task(name="atomic_coder")' for EACH todo
   - Pass ONLY the specific instruction (e.g., "Write validateEmail() in JS")
   - DO NOT pass project context
   
3. PARALLELIZE: Group independent tasks and delegate them simultaneously
   - Example: All HTML components can run in parallel
   - Then all JS functions can run in parallel
   - Then integration/testing runs sequentially

4. COMPILE: After all tasks complete, review the file system and assemble results

DECOMPOSITION EXAMPLE:
User: "Build a calculator app"

Bad (Monolithic):
❌ "Build calculator app" → atomic_coder

Good (MAKER):
✅ Phase 1 (Parallel):
   - "Create calculator.html with button grid"
   - "Create styles.css with calculator UI"
   
✅ Phase 2 (Parallel):
   - "Write add(a,b) function in calc.js"
   - "Write subtract(a,b) function in calc.js"  
   - "Write multiply(a,b) function in calc.js"
   - "Write divide(a,b) function in calc.js"
   
✅ Phase 3 (Sequential):
   - "Write handleButtonClick() to connect UI to logic"
   - "Test all functions with edge cases"

TASK SIZING:
- ✅ GOOD: "Write a login form with email and password fields"
- ✅ GOOD: "Create handleSubmit() function for the login form"
- ❌ BAD: "Build the entire authentication system"

PARALLEL EXECUTION RULES:
- Files in different directories → Parallel
- Functions in same file → Sequential (to avoid conflicts)
- Independent components → Parallel
- Dependent tasks → Sequential

OUTPUT FORMAT:
When done, provide:
1. Task Summary: "Completed X tasks in Y batches"
2. File Tree: Show created files
3. Usage Instructions: How to run the code
"""


# ============================================
# CREATE AGENT
# ============================================

agent = create_deep_agent(
    model=model,
    subagents=[universal_coder],
    system_prompt=manager_system_prompt,
)
result = agent.invoke({"messages": [{"role": "user", "content": "CREATE a very beautiful todo list app"}]})

# Print the agent's response
print(result["messages"][-1].content)
print(f"✓ Agent initialized with {len(agent.get_graph().nodes)} nodes")

Error Message and Stack Trace (if applicable)

Connection failed. Ensure your server is running at this endpoint.
This happens in the langsmith studio

Description

even though the server is running

Even Though i have used latest version of everything this issue is still there please fix it

System Info

System Information

OS: Windows
OS Version: 10.0.26100
Python Version: 3.12.6 (tags/v3.12.6:a4a2d2b, Sep 6 2024, 20:11:23) [MSC v.1940 64 bit (AMD64)]

Package Information

langchain_core: 1.0.0
langchain: 1.0.2
langsmith: 0.4.37
langchain_anthropic: 1.0.0
langchain_mcp_adapters: 0.1.11
langchain_ollama: 1.0.0
langchain_openai: 1.0.1
langchain_pinecone: 0.0.1
langgraph_api: 0.4.46
langgraph_cli: 0.4.4
langgraph_runtime_inmem: 0.14.1
langgraph_sdk: 0.2.9

Optional packages not installed

langserve

Other Dependencies

anthropic: 0.71.0
blockbuster: 1.5.25
claude-agent-sdk: Installed. No version info available.
click: 8.3.0
cloudpickle: 3.1.1
cryptography: 44.0.3
grpcio: 1.76.0
grpcio-tools: 1.76.0
httpx: 0.28.1
jsonpatch: 1.33
jsonschema-rs: 0.29.1
langchain-aws: Installed. No version info available.
langchain-community: Installed. No version info available.
langchain-deepseek: Installed. No version info available.
langchain-fireworks: Installed. No version info available.
langchain-google-genai: Installed. No version info available.
langchain-google-vertexai: Installed. No version info available.
langchain-groq: Installed. No version info available.
langchain-huggingface: Installed. No version info available.
langchain-mistralai: Installed. No version info available.
langchain-perplexity: Installed. No version info available.
langchain-together: Installed. No version info available.
langchain-xai: Installed. No version info available.
langgraph: 1.0.1
langgraph-checkpoint: 3.0.0
langsmith-pyo3: Installed. No version info available.
mcp: 1.18.0
numpy: 2.3.4
ollama: 0.6.0
openai: 2.6.0
openai-agents: Installed. No version info available.
opentelemetry-api: 1.38.0
opentelemetry-exporter-otlp-proto-http: 1.38.0
opentelemetry-sdk: 1.38.0
orjson: 3.11.3
packaging: 25.0
pinecone-client: 3.2.2
protobuf: 6.33.0
pydantic: 2.12.3
pyjwt: 2.10.1
pytest: Installed. No version info available.
python-dotenv: 1.1.1
pyyaml: 6.0.3
requests: 2.32.5
requests-toolbelt: 1.0.0
rich: 14.2.0
simsimd: 3.9.0
sse-starlette: 2.1.3
starlette: 0.48.0
structlog: 25.4.0
tenacity: 9.1.2
tiktoken: 0.12.0
truststore: 0.10.4
typing-extensions: 4.15.0
uvicorn: 0.38.0
vcrpy: Installed. No version info available.
watchfiles: 1.1.1
zstandard: 0.25.0

[kshadow07]

This is current version of the dependencies

"deepagents>=0.2.7",
"langchain-openai>=1.0.3",
"langgraph>=1.0.3",
"langgraph-cli[inmem]>=0.4.7",
"python-dotenv>=1.2.1",
"openai>=1.40.0",
"openai-agents>=0.3.0",
"python-dotenv>=1.0.1",
"pytest>=8.0.0",

[AnirbansarkarS]

This appears to be an issue with the LangSmith/LangGraph integration and the create_deep_agent function from the deepagents package. Here are several solutions to try:

Solution 1: Check LangSmith Configuration

import os
from dotenv import load_dotenv

# Load environment
load_dotenv()

# Ensure LangSmith is properly configured
os.environ["LANGCHAIN_TRACING_V2"] = "true"
os.environ["LANGCHAIN_ENDPOINT"] = "https://api.smith.langchain.com"
os.environ["LANGCHAIN_PROJECT"] = "your-project-name"  # Set a project name

# Your existing code continues...

Solution 2: Use LangGraph Directly (Alternative to deepagents)

import os
from dotenv import load_dotenv
from langchain_openai import ChatOpenAI
from langgraph.checkpoint.sqlite import SqliteSaver
from langgraph.prebuilt import create_react_agent

# Load environment
load_dotenv()

# Initialize model
model = ChatOpenAI(model="gpt-4o-mini", temperature=0)

# Create agent using LangGraph directly
memory = SqliteSaver.from_conn_string(":memory:")
agent = create_react_agent(model, tools=[], checkpointer=memory)

# Use the agent
result = agent.invoke({"messages": [{"role": "user", "content": "CREATE a very beautiful todo list app"}]})
print(result["messages"][-1].content)

Solution 3: Fix the Model Initialization

The issue might be with the model name and initialization:

from langchain_openai import ChatOpenAI
from dotenv import load_dotenv
import os

load_dotenv()

# Fix 1: Use correct model name
model = ChatOpenAI(
    model="gpt-4o-mini",  # Fixed model name
    temperature=0,
    streaming=False
)

# Fix 2: Check if deepagents is the issue
try:
    from deepagents import create_deep_agent
    
    agent = create_deep_agent(
        model=model,
        subagents=[],  # Start with empty subagents
        system_prompt="You are a helpful assistant",
    )
    
    result = agent.invoke({"messages": [{"role": "user", "content": "CREATE a simple todo list app"}]})
    print(result["messages"][-1].content)
    
except Exception as e:
    print(f"DeepAgents error: {e}")
    # Fallback to basic LangChain
    from langchain.schema import HumanMessage
    messages = [HumanMessage(content="CREATE a simple todo list app")]
    result = model.invoke(messages)
    print(result.content)

Solution 4: Complete Working Example (Minimal)

import os
from dotenv import load_dotenv
from langchain_openai import ChatOpenAI
from langchain.schema import HumanMessage

# Load environment
load_dotenv()

# Set LangSmith configuration
os.environ["LANGCHAIN_TRACING_V2"] = "false"  # Disable temporarily to test

# Initialize model with correct parameters
model = ChatOpenAI(
    model="gpt-4o-mini",
    temperature=0.7,
    max_tokens=1000,
    timeout=30
)

try:
    # Simple direct invocation
    messages = [HumanMessage(content="CREATE a very beautiful todo list app with HTML, CSS and JavaScript")]
    result = model.invoke(messages)
    print("✅ Success!")
    print(result.content)
    
except Exception as e:
    print(f"❌ Error: {e}")
    print("Trying alternative approach...")
    
    # Alternative: Check API key and connectivity
    from openai import OpenAI
    client = OpenAI(api_key=os.getenv('OPENAI_API_KEY'))
    
    try:
        response = client.chat.completions.create(
            model="gpt-4o-mini",
            messages=[{"role": "user", "content": "Say hello"}],
            max_tokens=10
        )
        print("✅ OpenAI API connectivity test passed")
    except Exception as api_error:
        print(f"❌ OpenAI API error: {api_error}")

Solution 5: Debug Steps

Add this debug script to identify the exact issue:

import os
from dotenv import load_dotenv

load_dotenv()

print("=== Debug Information ===")
print(f"OpenAI API Key present: {bool(os.getenv('OPENAI_API_KEY'))}")
print(f"LangChain Tracing: {os.getenv('LANGCHAIN_TRACING_V2', 'Not set')}")

# Test basic imports
try:
    from langchain_openai import ChatOpenAI
    print("✅ langchain_openai import successful")
except ImportError as e:
    print(f"❌ langchain_openai import failed: {e}")

try:
    from deepagents import create_deep_agent
    print("✅ deepagents import successful")
except ImportError as e:
    print(f"❌ deepagents import failed: {e}")

# Test basic OpenAI connectivity
try:
    from openai import OpenAI
    client = OpenAI()
    models = client.models.list()
    print("✅ OpenAI connectivity test passed")
except Exception as e:
    print(f"❌ OpenAI connectivity test failed: {e}")

Solution 6: Update Dependencies

Create a fresh requirements.txt:

langchain-core==1.0.0
langchain-openai==1.0.3
langchain-community==0.2.4
langgraph==1.0.3
openai==1.40.0
python-dotenv==1.0.1
langsmith==0.4.37

Install with:

pip install -r requirements.txt

Immediate Fix

For immediate testing, use this simplified version that bypasses the deepagents issue:

import os
from dotenv import load_dotenv
from langchain_openai import ChatOpenAI
from langchain.schema import HumanMessage

load_dotenv()

# Disable LangSmith temporarily
os.environ["LANGCHAIN_TRACING_V2"] = "false"

model = ChatOpenAI(model="gpt-4o-mini", temperature=0.9)
response = model.invoke([HumanMessage(content="CREATE a very beautiful todo list app")])
print(response.content)

The main issues appear to be:

1. Incorrect model name (openai:gpt-5-mini doesn't exist)
2. Potential deepagents compatibility issues
3. LangSmith configuration problems

Try Solution 4 first for immediate results, then work through the others to identify the specific root cause.

[MatteoGaetzner]

Were you able to reproduce the error @AnirbansarkarS?
After allowing local network access on my browser (Brave) I was able to observe the agent on https://smith.langchain.com/studio.
I'm using the example code without modifications and the 'current version of dependencies'.

[AnirbansarkarS]

The problem seems to be that LangSmith Studio is trying to automatically open in the browser, but there are connectivity/permission issues preventing it.

The good news is that your Brave browser workaround (allowing local network access) confirms this is the core issue. The code is actually working fine - it's just the LangSmith UI auto-launch that's failing.

For other users facing this, here are the solutions:

Browser Fix: Like you did, allow local network access in Brave/chrome flags

Code Fix: Disable LangSmith auto-opening by setting LANGCHAIN_TRACING_V2=false

Manual Access: Just manually go to https://smith.langchain.com/studio after running the code

The agent is still executing properly in the background - the browser popup is just for visualization. If you disable the auto-open, the code runs without any errors.

[MatteoGaetzner]

Not related to the bug but note that you should set LANGCHAIN_TRACING=true instead of LANGCHAIN_TRACING_V2=true (https://docs.langchain.com/langsmith/observability-quickstart).

[kshadow07]

this model is there openai:gpt-5-mini i mean i was able to run it in the cli but for testing it in the langsmith it is not connecting seems like an version error or main langchain-api error

[kshadow07]

Also my main project that i built was on deepagents itself so again changing things to Langgraph would be a bad idea