Construct Neptune URLs

You can construct Neptune app URLs based on run or experiment properties.

Constructing experiment URL

If a new run becomes the head of an experiment, Neptune automatically switches to visualizing the new run in the web app.

To construct a URL that opens the run for a given experiment:

1. Navigate to the app section that the URL should open.

2. In the URL, set runIdentificationKey to the experiment name:

   ...neptune.ai/{Workspace}/{Project}/runs/...&runIdentificationKey={ExperimentName}&type=experiment

   {ExperimentName} is the name passed to the experiment_name argument at run creation. It's displayed in the Name column of the runs table.

Example: https://scale.neptune.ai/o/neptune/org/LLM-training-example/runs/details?viewId=standard-view&runIdentificationKey=long_YFP0TJYE_d_0.99999_f_0.15_n_100&type=experiment

Constructing run URL

To construct a URL that opens a particular run, use the following schema:

...neptune.ai/{Workspace}/{Project}/-/run/?customId={CustomRunID}

{CustomRunID} is the identifier you pass to run_id at initialization. It's stored in the sys/custom_run_id attribute.

Example: https://scale.neptune.ai/o/neptune/org/LLM-training-example/-/run/?customId=QFEKX7LGHW

Point URL to specific tab

To target a specific tab in the focused run view, append:

&detailsTab={TabName}

where {TabName} is one of the following:

• metadata → All metadata tab
• charts → Charts tab
• dashboard → Dashboards tab

Example: https://scale.neptune.ai/o/neptune/org/LLM-training-example/-/run/?customId=QFEKX7LGHW&detailsTab=charts

Constructing URL for filtered view

To create a URL that leads you to a filtered table view, use the query parameter:

...neptune.ai/{Workspace}/{Project}/runs/table?viewId=standard-view&query=(...)

For example, to see all runs with the trial tag:

query=((%60sys%2Ftags%60%3AstringSet%20CONTAINS%20%22trial%22))

To see all runs with either the trial or script tag:

query=((%60sys%2Ftags%60%3AstringSet%20CONTAINS%20%22trial%22)%20OR%20(%60sys%2Ftags%60%3AstringSet%20CONTAINS%20%22script%22))

To see all runs with both the trial and script tags:

query=((%60sys%2Ftags%60%3AstringSet%20CONTAINS%20%22trial%22)%20AND%20(%60sys%2Ftags%60%3AstringSet%20CONTAINS%20%22script%22))

Example link: https://scale.neptune.ai/o/neptune/org/LLM-training-example/runs/table?viewId=standard-view&query=((%60sys%2Ftags%60%3AstringSet%20CONTAINS%20%22large%22))

Constructing URL in code

If you have the custom run ID set to run_id and the project path set to the NEPTUNE_PROJECT environment variable, you can generate the URL with the following Python string:

f"https://scale.neptune.ai/{os.getenv('NEPTUNE_PROJECT')}/-/run/?customId={run_id}"

Get URL to latest run

To get and print the URL of the latest run in your project, you need the neptune-fetcher package. To install it:

pip install -U neptune-fetcher

Then, to fetch the latest run:

from neptune_fetcher import ReadOnlyProject

project = ReadOnlyProject()
latest_run_id = project.fetch_runs()["sys/custom_run_id"][0]
print(latest_run_id)

To apply a filter, use fetch_runs_df(). For example, to get the latest run created by yourself:

from neptune_fetcher import ReadOnlyProject

project = ReadOnlyProject()
latest_run_id = project.fetch_runs_df(
    columns=["sys/custom_run_id"],
    owners=["YOUR_NEPTUNE_USERNAME"],  # replace with your username
)["sys/custom_run_id"].values[0]

print(latest_run_id)

To get the URL for your run:

import os
from random import random
from neptune_scale import Run
from neptune_fetcher import ReadOnlyProject

project = ReadOnlyProject()

run = Run(run_id="PX30SG002")
# Once the run has synchronized, you can fetch its ID from the server
latest_run_id = project.fetch_runs()["sys/custom_run_id"][0]  # returns PX30SG002

print(
    f"https://scale.neptune.ai/{os.getenv('NEPTUNE_PROJECT')}/-/run/?customId={latest_run_id}"
)

Result

https://scale.neptune.ai/team-alpha/project-x/-/run/?customId=PX30SG002