make dashboard proxy configurable (#116)

* enable kwarg passthru to dash app * optionally serve app in iframe * pass tests * update docs * restructure how the dash app is created (#117) * fix imports * remove jupyter-dash from install Co-authored-by: Joe Wolfe <[email protected]>
capitalone · Aug 24, 2021 · 005eeb8 · 005eeb8
1 parent 28fd974
commit 005eeb8
Show file tree

Hide file tree

Showing 17 changed files with 345 additions and 264 deletions.
diff --git a/docs/source/dashboard.rst b/docs/source/dashboard.rst
@@ -3,20 +3,22 @@
 Dashboard
 *********
 
-The dashboard is powered by `Plotly <https://plotly.com/>`_ and
-`Dash <https://dash.plotly.com>`_ - specifically
-`JupyterDash <https://medium.com/plotly/introducing-jupyterdash-811f1f57c02e>`_.
-It provides a UI for exploring and visualizing projects stored locally
-or in S3. The dashboard can be run in the browser (no Jupyter required!) or
-within Jupyter lab or notebooks. With a little extra work, it could even be
+The dashboard is powered by `Dash <https://dash.plotly.com>`_ and `Plotly
+<https://plotly.com/>`_. It provides a UI for exploring and visualizing projects
+stored locally or in S3. The dashboard can be run in the browser or within
+Jupyter Notebook, JupyterLab or JupyterHub. With some extra work, it could be
 deployed to a static URL.
 
 Usage
 =====
 
-Install the necessary requirements with the UI extra as detailed in the
-:ref:`installation section<install>`. Then, the dashboard can invoked via the
-CLI:
+First, install the UI extra as detailed in the :ref:`installation
+section<install>`. Then decide how you'd like to run the dashboard.
+
+CLI
+~~~
+
+Run the dashboard via the rubicon_ml CLI:
 
 .. code-block:: shell
 
@@ -28,9 +30,12 @@ Use the ``--help`` flag for full usage details:
 
   rubicon_ml ui --help
 
+Programmatically
+~~~~~~~~~~~~~~~~
+
 The dashboard can also be launched directly from python code. If you're using
-in-memory persistence, this is the only way to get the dashboard to
-properly locate and load the in-memory projects and experiments.
+in-memory persistence, this is the only way to get the dashboard to properly
+locate and load the in-memory projects and experiments.
 
 .. code-block:: python
 
@@ -40,34 +45,48 @@ properly locate and load the in-memory projects and experiments.
 
 The above launches the dashboard on ``localhost:8050`` and will run from any
 Python interpreter. If you're working in a Jupyter environment, you can launch
-the dashboard inline in a notebook or in a new JupyterLab window by passing the
-``mode`` keyword argument.
+the dashboard inline in a notebook with `run_server_inline`:
 
 .. code-block:: python
 
-  Dashboard(persistence="memory", root_dir="/rubicon-root", mode="inline").run_server()
+  Dashboard(persistence="memory", root_dir="/rubicon-root").run_server_inline()
+
+If you need to configure the Dash proxy, that can also be done with a few extra params:
 
 .. code-block:: python
 
-  Dashboard(persistence="memory", root_dir="/rubicon-root", mode="jupyterlab").run_server()
+  from rubicon_ml.ui import Dashboard
+
+  dash_options = {
+      "requests_pathname_prefix": "/proxy/8050"
+  }
 
-Features
-========
+  Dashboard(
+      persistence="memory",
+      dash_options=dash_options,
+  ).run_server(proxy="http://127.0.0.1:8050::https://my-jupyterhub-server.com/")
 
-Once a project is selected within the project selection view, you'll see
-a high level grouping of its experiments. This works best if you've been
-using the built in git integration. Select a grouping to view the experiments underneath
-it and sort/filter the data directly within the experiment table. Optionally, compare
-experiments by selecting them within the experiment table and viewing the comparison
-plot below.
+More info on configuring proxies and additional ``dash_options`` can be found in the
+`Dash API Reference <https://dash.plotly.com/reference>`_.
 
-The experiment comparison chart is a
-`parallel coordinates plot <https://en.wikipedia.org/wiki/Parallel_coordinates>`_
-with each parameter and metric plotted on a standardized axis. It's useful to draw
-insights on how certain values of parameters affect key metrics.
+Features
+========
 
-For example, in the image below, you can see that true values of the ``tfidf__use_idf``
-and low values for ``clf__alpha`` correlate to high ``accuracy`` values.
+Once a project is selected within the project selection view, you'll see a high
+level grouping of its experiments. This works best if you've been using the
+built in git integration. Select a grouping to view the experiments underneath
+it and *sort/filter* the data directly within the experiment table. Optionally,
+compare experiments by selecting them within the experiment table and viewing
+the comparison plot below.
+
+The experiment comparison chart is a `parallel coordinates plot
+<https://en.wikipedia.org/wiki/Parallel_coordinates>`_ with each parameter and
+metric plotted on a standardized axis. It's useful to draw insights on how
+certain values of parameters affect key metrics.
+
+For example, in the image below, you can see that true values of the
+``tfidf__use_idf`` and low values for ``clf__alpha`` correlate to high
+``accuracy`` values.
 
 .. note::
     The y-axes can be dragged and rearranged to reorder the plot. And you can highlight

diff --git a/environment.yml b/environment.yml
@@ -27,7 +27,6 @@ dependencies:
   - beautifulsoup4>=4.8.2
   - dash>=1.14.0
   - dash-bootstrap-components>=0.10.6
-  - jupyter-dash
   - percy>=2.0.2
   - selenium>=3.141.0
 

diff --git a/notebooks/integrations/integration-git.ipynb b/notebooks/integrations/integration-git.ipynb
@@ -120,7 +120,7 @@
     "from rubicon_ml.ui import Dashboard\n",
     "\n",
     "\n",
-    "Dashboard(persistence=\"memory\").run_server()"
+    "Dashboard(persistence=\"memory\").run_server_inline()"
    ]
   },
   {
@@ -136,7 +136,7 @@
  ],
  "metadata": {
   "kernelspec": {
-   "display_name": "Python 3",
+   "display_name": "Python 3 (ipykernel)",
    "language": "python",
    "name": "python3"
   },
@@ -150,7 +150,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.9.4"
+   "version": "3.9.6"
   }
  },
  "nbformat": 4,

diff --git a/notebooks/integrations/integration-prefect-workflows.ipynb b/notebooks/integrations/integration-prefect-workflows.ipynb
@@ -650,13 +650,13 @@
     "from rubicon_ml.ui import Dashboard\n",
     "\n",
     "\n",
-    "Dashboard(persistence=\"filesystem\", root_dir=root_path).run_server()"
+    "Dashboard(persistence=\"filesystem\", root_dir=root_path).run_server_inline()"
    ]
   }
  ],
  "metadata": {
   "kernelspec": {
-   "display_name": "Python 3",
+   "display_name": "Python 3 (ipykernel)",
    "language": "python",
    "name": "python3"
   },
@@ -670,7 +670,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.9.4"
+   "version": "3.9.6"
   }
  },
  "nbformat": 4,

diff --git a/notebooks/integrations/integration-sklearn.ipynb b/notebooks/integrations/integration-sklearn.ipynb
@@ -343,13 +343,13 @@
    "source": [
     "from rubicon_ml.ui import Dashboard\n",
     "\n",
-    "Dashboard(persistence=\"filesystem\", root_dir=root_path).run_server()"
+    "Dashboard(persistence=\"filesystem\", root_dir=root_path).run_server_inline()"
    ]
   }
  ],
  "metadata": {
   "kernelspec": {
-   "display_name": "Python 3",
+   "display_name": "Python 3 (ipykernel)",
    "language": "python",
    "name": "python3"
   },
@@ -363,7 +363,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.9.4"
+   "version": "3.9.6"
   }
  },
  "nbformat": 4,

diff --git a/notebooks/quick-look.ipynb b/notebooks/quick-look.ipynb
@@ -235,7 +235,7 @@
     "from rubicon_ml.ui import Dashboard\n",
     "\n",
     "\n",
-    "Dashboard(persistence=\"filesystem\", root_dir=root_path).run_server()"
+    "Dashboard(persistence=\"filesystem\", root_dir=root_path).run_server_inline()"
    ]
   },
   {
@@ -386,7 +386,7 @@
  ],
  "metadata": {
   "kernelspec": {
-   "display_name": "Python 3",
+   "display_name": "Python 3 (ipykernel)",
    "language": "python",
    "name": "python3"
   },
@@ -400,7 +400,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.9.4"
+   "version": "3.9.6"
   }
  },
  "nbformat": 4,

diff --git a/rubicon_ml/ui/README.md b/rubicon_ml/ui/README.md
@@ -1,21 +1,5 @@
 ## Rubicon Dashboard
 
-The Rubicon Dashboard (powered by [Dash](https://dash.plotly.com/introduction)) offers a way to visualize your logged data within a local or shareable UI. To view the Dashboard, follow these steps:
+The Rubicon Dashboard (powered by [Dash](https://dash.plotly.com/introduction)) offers a way to visualize your logged data within a local or shareable UI.
 
-* install the required dependencies:
-
-    ```
-    pip install rubicon[ui]
-    ```
-
-* use the CLI to launch the dashboard:
-
-    ```bash
-    # usage
-    rubicon_ml ui --help
-
-    # run
-    rubicon_ml ui --root-dir "/path/to/root" --project-name "Example"
-    ```
-
-* view the dashboard at `http://127.0.0.1:8050/`
+To get started, visit the [dashboard docs](https://capitalone.github.io/rubicon-ml/dashboard.html)
diff --git a/rubicon_ml/ui/__init__.py b/rubicon_ml/ui/__init__.py
@@ -11,6 +11,6 @@ def _check_for_ui_extras():
 
 _check_for_ui_extras()
 
-from rubicon_ml.ui.dashboard import Dashboard  # noqa E402
+from rubicon_ml.ui.dashboard import Dashboard  # noqa F401
 
 __all__ = ["Dashboard"]
diff --git a/rubicon_ml/ui/app.py b/rubicon_ml/ui/app.py
diff --git a/rubicon_ml/ui/callbacks/__init__.py b/rubicon_ml/ui/callbacks/__init__.py
@@ -0,0 +1,4 @@
+from .project_explorer import set_project_explorer_callbacks
+from .project_selection import set_project_selection_callbacks
+
+__all__ = ["set_project_explorer_callbacks", "set_project_selection_callbacks"]