Skip to content

Visualization

One of the number one complaints when working with LLMs is that they can be a black box. Agentic applications exacerbate this problem by adding even more complexity. Railtracks aims to make it easier than ever to visualize your runs.

We support:

  • Local Visualization (no sign up required)
  • Remote Visualization (Ideal for deployed agents)

Local Development Visualization

Railtracks comes with a built-in visualization tool that runs locally with no sign up required.

Usage

Install CLI Tool
pip install 'railtracks[visual]'
Initialize UI and Start
railtracks init
railtracks viz

This will create a .railtracks directory at your project root and open the web app in your browser. Once initialised, railtracks will find that directory automatically, even if you run your agents from a subdirectory, by walking up the folder tree until it locates .railtracks.

Running from multiple directories?

Run railtracks init once from your project root (the same level as your .git folder). All subsequent agent runs across the project will resolve to that single .railtracks directory regardless of which subdirectory they are launched from.

If you need a fixed location outside your project (e.g. a shared drive or CI environment), set the RAILTRACKS_HOME environment variable to the parent directory where .railtracks should live: 

export RAILTRACKS_HOME=/path/to/my/project   # .railtracks is created inside here
RAILTRACKS_HOME always takes priority over directory traversal.

Saving State

By default, all of your runs will be saved to the .railtracks directory so you can view them locally. If you don't want that, set the flag to False:

import railtracks as rt

# set the configuration globally
rt.set_config(save_state=True)

# or by flow
flow = rt.Flow("my_flow", 
               entry_point=flow_entry,
               save_state=True)

Updating the UI

As we continue to support the local visualizer and add more features, you may choose to integrate these updated UI components into your local installation by running:

Update UI elements
railtracks update

Remote Visualization

Note

Would you be interested in observability for your agents in an all in one platform?

Please fill out the following form