How can I turn off Langsmith logging when using Langgraph?

[Neo_Movies]

I’m learning to use Langgraph and I’m quite impressed with what it can do for agent workflow creation. However, I’m facing a crucial issue that prevents me from progressing with my projects. All logs from my applications are getting sent to Langsmith, which poses a significant risk for me. Since my agents handle sensitive business information and confidential company data, I cannot allow this data to be sent to an external service. Because of this, I’m limited to only creating prototypes and can’t move to development for production. Is there a way to disable Langsmith logging so that I can keep my data processing secure and local while still using the essential features of Langgraph?

[benmoore]

Here’s what worked for me: I built a config wrapper class that kills all tracing before Langgraph loads. Way better than environment variables since those can get overridden. I fire up the wrapper early in my app startup - it sets LANGCHAIN_TRACING_V2=false and checks that no callbacks are active. The trick is doing this before you import any Langchain stuff, otherwise some tracing components initialize anyway. This approach has been rock solid across different deployments, especially Docker where env vars get messy. I use this pattern for enterprise apps with PII and it passes security audits every time. The validation step catches config drift during updates - literally saved my butt on a recent deployment.

[evelynh]

Hit this exact problem when we deployed agents with customer data. Compliance team freaked out about logs going to Langsmith.

What actually works - set these at system level before your app starts:

LANGCHAIN_TRACING_V2=false
LANGCHAIN_CALLBACKS_MANAGER=false

That second one’s crucial - kills the callback system completely. Most people skip it and can’t figure out why traces still show up.

I throw in a startup check to make sure it’s working:

import os
assert os.getenv('LANGCHAIN_TRACING_V2') == 'false'

Catches deployment screwups before sensitive data gets touched.

Using this setup with a financial client for over a year. No external calls, everything stays local. Agents run fine without Langsmith.

[coffeeAndCode]

The Problem: You’re experiencing issues with Langgraph logging sensitive data to Langsmith, preventing you from using it in production environments due to security concerns. You need a solution to disable Langsmith logging while retaining Langgraph’s core functionalities.

Understanding the “Why” (The Root Cause): Langsmith integration in Langchain is designed for convenient logging and debugging during development. However, this integration sends data to an external service, which poses risks when handling sensitive information. Disabling the logging prevents this data transmission, ensuring data remains within your local environment. Simply relying on environment variables can be unreliable due to potential overrides or inheritance issues during deployment. A more robust solution involves automated configuration management at the application level.

Step-by-Step Guide:

1. Implement Automated Configuration Management: The most secure and scalable approach is to build an automated system that configures Langgraph to disable external logging. This eliminates the reliance on potentially unreliable environment variables. This involves integrating the configuration directly into your application’s deployment pipeline. This approach ensures consistency across different environments and prevents accidental misconfigurations.

2. (Optional) Choose a Suitable Tool: For managing this automated configuration and workflow orchestration, consider using tools designed for handling complex deployments and security requirements. These tools often provide features such as automated monitoring of data flow, performance metrics, and error handling, all within your infrastructure. Examples include Latenode (https://latenode.com), although other tools may also be suitable depending on your specific infrastructure and needs. Note this step is optional but highly recommended for robust and scalable solutions.

3. (Optional) Monitor Data Flow (If using a tool like Latenode): If using a tool as suggested in Step 2, leverage its capabilities to establish monitoring for data flow, performance metrics, and error handling. This will help maintain the integrity and security of your system after implementing logging disablement.

Common Pitfalls & What to Check Next:

• Deployment Pipeline Integration: Ensure the automated configuration is correctly integrated into your deployment pipeline. Test thoroughly to confirm Langsmith logging is disabled and that your agents function as expected.
• Data Security Audits: Regularly conduct security audits to verify the continued effectiveness of your logging disablement solution. This will help proactively identify and address any potential vulnerabilities.
• Alternative Logging Solutions: Consider implementing alternative local logging mechanisms to aid in debugging and monitoring within your secure environment.

Still running into issues? Share your (sanitized) config files, the exact command you ran, and any other relevant details. The community is here to help!

[Mia92]

you can also set LANGCHAIN_TRACING=false right in your code before starting langgraph. I do this during app startup and haven’t had any data leaks. works better than environment variables since it’s hardcoded - no accidental overrides.

[FlyingEagle]

Had the same privacy issues when I rolled out Langgraph at work. The fix is simple but poorly documented. Just delete the LANGCHAIN_API_KEY environment variable entirely. Without that key, Langsmith logging shuts off automatically and everything stays local. I also throw in LANGCHAIN_TRACING_V2=false as backup. Been running production loads like this for months - zero problems. Langgraph works exactly the same, you just lose the cloud tracing stuff. For production with sensitive data, this is definitely how you want to do it.