Integrate with your agents

Supported Frameworks

It does not matter how you are using DSPy, nanomon suits you!

Step 1: Create a project in nanomon

Once you have your nanomon running (Self-Hosted or Cloud-Hosted), start by creating a project:

Open nanomon UI home page
Create a new Project
Copy the project Client ID and Secret ID (you’ll need this to proper link the agent/flow with this project)

Step 2: Set the project credentials

Connect your agent/flow with the nanomon project you’ve just created in Step 1. To do that set the following environment variables in your DSPy project:

NANOMON_CLIENT_ID=<your_project_client_id>
NANOMON_SECRET_ID=<your_project_secret_id>

Step 3: Set the API Url (self-hosted only)

We know that when running locally we all want to have control about what are the ports to be used, why and when, because of that we allow you to change the nanomon API Url simply by setting, in your DSPy project the environment variable

NANOMON_API_URL=http://localhost:8000/api/v1 # Or the one you want to

This way the agents will report to the nanomon api that is running in that url you defined.

Final Step: Configure your flow

As the final step you just have to configure your agent/flow to send monitoring data to your nanomon project. All you have to do is to:

Install nanomon

pip install nanomon

Add this code snippet to your code

nanomon_context = NanomonRunContext(default_tags=["dev", "nanomon-tester"]) # add whatever tags you want
configure_nanomon(nanomon_context._sink)
lm = nanomon_context.instrument_lm(lm) # with lm being your regular DSPy lm

# Configure DSPy
dspy.configure(lm=lm)

# Wrap it
with nanomon_context.run(
    tags=["my-agent-flow"], # an example of a tag of your choice
    metadata={"user-id": "User ID 12345"} # just an example of how you can use it
):

# Run your lang graph (or simply your DSPy Predict or ReAct directly)
final_state = await graph.ainvoke(initial_state)

Getting started

Deployment types