[DOC] Model callouts and descriptions

nbs/docs/getting-started/5_faq.ipynb

@@ -153,6 +153,25 @@
     "## Features and Capabilities"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<details>\n",
+    "    <summary>What is the input to TimeGPT?</summary>\n",
+    "\n",
+    "`TimeGPT` accepts `pandas` dataframes in [long format](https://www.theanalysisfactor.com/wide-and-long-data/#comments) with the following necessary columns: \n",
+    "\n",
+    "- `ds` (timestamp): timestamp in format `YYYY-MM-DD` or `YYYY-MM-DD HH:MM:SS`. \n",
+    "- `y` (numeric): The target variable to forecast. \n",
+    "\n",
+    "(Optionally, you can also pass a DataFrame without the `ds` column as long as it has DatetimeIndex)\n",
+    "\n",
+    "`TimeGPT` also works with [distributed dataframes](https://docs.nixtla.io/docs/tutorials-computing_at_scale) like `dask`, `spark` and `ray`. \n",
+    "\n",
+    "</details>"
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},
@@ -295,7 +314,7 @@
     "<details>\n",
     "    <summary>Does TimeGPT support polars?</summary>\n",
     "\n",
-    "As of now, `TimeGPT` supports only pandas DataFrames and does not offer support for polars. \n",
+    "As of now, `TimeGPT` does not offer support for polars. \n",
     "\n",
     "</details>"
    ]

nbs/docs/tutorials/050_training.ipynb

@@ -5,15 +5,19 @@
    "id": "6de758ee-a0d2-4b3f-acff-eed419dd17c5",
    "metadata": {},
    "source": [
-    "# Training"
-   ]
-  },
-  {
-   "cell_type": "markdown",
-   "id": "5d267032-535b-4b7b-b7d3-d2db8f673af6",
-   "metadata": {},
-   "source": [
-    "This section offers tutorials related to training a model."
+    "# Training\n",
+    "\n",
+    "This section offers tutorials related to training `TimeGPT` under specific conditions.\n",
+    "\n",
+    "### What You Will Learn\n",
+    "\n",
+    "1. **[Long Horizon Forecasting](https://docs.nixtla.io/docs/tutorials-long_horizon_forecasting)**\n",
+    "\n",
+    "    - Discover how make predictions beyond two seasonal periods or even further into the future, using `TimeGPT`'s specialized model for long horizon forecasting.\n",
+    "\n",
+    "2. **[Multiple Series Forecasting](https://docs.nixtla.io/docs/tutorials-multiple_series_forecasting)**\n",
+    "\n",
+    "    - Learn how to use `TimeGPT` to forecast multiple time series simultaneously."
    ]
   }
  ],

nbs/docs/tutorials/080_validation.ipynb

@@ -13,9 +13,17 @@
    "id": "5d267032-535b-4b7b-b7d3-d2db8f673af6",
    "metadata": {},
    "source": [
-    "One of the primary challenges in time series forecasting is the inherent uncertainty and variability over time, making it crucial to validate the accuracy and reliability of the models employed. \n",
+    "One of the primary challenges in time series forecasting is the inherent uncertainty and variability over time, making it crucial to validate the accuracy and reliability of the models employed. `TimeGPT` offers the possibility for cross-validation and historical forecasts to help you validate your predictions.\n",
     "\n",
-    "TimeGPT offers the possibility for cross-validation and historical forecasts to help you validate your predictions."
+    "### What You Will Learn\n",
+    "\n",
+    "1. **[Cross-Validation](https://docs.nixtla.io/docs/tutorials-cross_validation)**\n",
+    "\n",
+    "    - Learn how to perform time series cross-validation across different continuous windows of your data. \n",
+    "\n",
+    "2. **[Historical Forecasts](https://docs.nixtla.io/docs/tutorials-historical_forecast)**\n",
+    "\n",
+    "    - Generate in-sample forecasts to validate how `TimeGPT` would have performed in the past, providing insights into the model's accuracy. \n"
    ]
   }
  ],

nbs/docs/tutorials/09_historical_forecast.ipynb

@@ -121,6 +121,18 @@
     ")"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "id": "dce721eb",
+   "metadata": {},
+   "source": [
+    "> 👍 Use an Azure AI endpoint\n",
+    ">\n",
+    "> To use an Azure AI endpoint, set the `base_url` argument:\n",
+    ">\n",
+    "> `nixtla_client = NixtlaClient(base_url=\"you azure ai endpoint\", api_key=\"your api_key\")`"
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
@@ -292,6 +304,22 @@
     ")"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "id": "6933cbb0",
+   "metadata": {},
+   "source": [
+    "> 📘 Available models in Azure AI\n",
+    ">\n",
+    "> If you are using an Azure AI endpoint, please be sure to set `model=\"azureai\"`:\n",
+    ">\n",
+    "> `nixtla_client.forecast(..., model=\"azureai\")`\n",
+    "> \n",
+    "> For the public API, we support two models: `timegpt-1` and `timegpt-1-long-horizon`. \n",
+    "> \n",
+    "> By default, `timegpt-1` is used. Please see [this tutorial](https://docs.nixtla.io/docs/tutorials-long_horizon_forecasting) on how and when to use `timegpt-1-long-horizon`.\n"
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,

nbs/docs/tutorials/100_uncertainty_quantification.ipynb

@@ -5,17 +5,19 @@
    "id": "6de758ee-a0d2-4b3f-acff-eed419dd17c5",
    "metadata": {},
    "source": [
-    "# Uncertainty quantification"
-   ]
-  },
-  {
-   "cell_type": "markdown",
-   "id": "5d267032-535b-4b7b-b7d3-d2db8f673af6",
-   "metadata": {},
-   "source": [
-    "In forecasting, we are often interested in a distribution of predictions rather than only a point prediction, because we want to have a notion of the uncertainty around the forecast.\n",
+    "# Uncertainty quantification\n",
+    "\n",
+    "In forecasting, it is essential to consider the full distribution of predictions rather than only a point prediction. This approach allows for a better understanding of the uncertainty surrounding the forecast. `TimeGPT` supports uncertainty quantification through quantile forecasts and prediction intervals.\n",
+    "\n",
+    "### What You Will Learn\n",
+    "\n",
+    "1. **[Quantile Forecasts](https://docs.nixtla.io/docs/tutorials-quantile_forecasts)**\n",
+    "\n",
+    "    - Learn how to compute specific quantiles of the forecast distribution using `TimeGPT`. \n",
+    "\n",
+    "2. **[Prediction Intervals](https://docs.nixtla.io/docs/tutorials-prediction_intervals)**\n",
     "\n",
-    "TimeGPT offers the possibility to quantify uncertainty using quantile forecasts and prediction intervals."
+    "    - Learn how to generate prediction intervals with `TimeGPT`, which give you a range of values that the forecast can take with a given probability. \n"
    ]
   }
  ],

nbs/docs/tutorials/10_uncertainty_quantification_with_quantile_forecasts.ipynb

@@ -131,6 +131,18 @@
     ")"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "id": "1e0c97c2",
+   "metadata": {},
+   "source": [
+    "> 👍 Use an Azure AI endpoint\n",
+    ">\n",
+    "> To use an Azure AI endpoint, set the `base_url` argument:\n",
+    ">\n",
+    "> `nixtla_client = NixtlaClient(base_url=\"you azure ai endpoint\", api_key=\"your api_key\")`"
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
@@ -404,6 +416,22 @@
     "timegpt_quantile_fcst_df.head()"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "id": "6367adc2",
+   "metadata": {},
+   "source": [
+    "> 📘 Available models in Azure AI\n",
+    ">\n",
+    "> If you are using an Azure AI endpoint, please be sure to set `model=\"azureai\"`:\n",
+    ">\n",
+    "> `nixtla_client.forecast(..., model=\"azureai\")`\n",
+    "> \n",
+    "> For the public API, we support two models: `timegpt-1` and `timegpt-1-long-horizon`. \n",
+    "> \n",
+    "> By default, `timegpt-1` is used. Please see [this tutorial](https://docs.nixtla.io/docs/tutorials-long_horizon_forecasting) on how and when to use `timegpt-1-long-horizon`."
+   ]
+  },
   {
    "cell_type": "markdown",
    "id": "f75f9fcc-24a5-44b5-b19d-8875360d7af3",

nbs/docs/tutorials/11_uncertainty_quantification_with_prediction_intervals.ipynb

@@ -131,6 +131,18 @@
     ")"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "id": "4ec82f80",
+   "metadata": {},
+   "source": [
+    "> 👍 Use an Azure AI endpoint\n",
+    ">\n",
+    "> To use an Azure AI endpoint, set the `base_url` argument:\n",
+    ">\n",
+    "> `nixtla_client = NixtlaClient(base_url=\"you azure ai endpoint\", api_key=\"your api_key\")`"
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
@@ -377,6 +389,22 @@
     "timegpt_fcst_pred_int_df.head()"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "id": "e7cefb13",
+   "metadata": {},
+   "source": [
+    "> 📘 Available models in Azure AI\n",
+    ">\n",
+    "> If you are using an Azure AI endpoint, please be sure to set `model=\"azureai\"`:\n",
+    ">\n",
+    "> `nixtla_client.forecast(..., model=\"azureai\")`\n",
+    "> \n",
+    "> For the public API, we support two models: `timegpt-1` and `timegpt-1-long-horizon`. \n",
+    "> \n",
+    "> By default, `timegpt-1` is used. Please see [this tutorial](https://docs.nixtla.io/docs/tutorials-long_horizon_forecasting) on how and when to use `timegpt-1-long-horizon`."
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,

nbs/docs/tutorials/120_special_topics.ipynb

@@ -13,7 +13,27 @@
    "id": "5d267032-535b-4b7b-b7d3-d2db8f673af6",
    "metadata": {},
    "source": [
-    "Here you can find tutorials for dealing with special topics during forecasting."
+    "`TimeGPT` is a robust foundation model for time series forecasting, with advanced capabilities such as hierarchical and bounded forecasts. To fully leverage the power of `TimeGPT`, there are specific situations that require special consideration, such as dealing with irregular timestamps or handling datasets with missing values.\n",
+    "\n",
+    "In this section, we will cover these special topics.\n",
+    "\n",
+    "### What You Will Learn\n",
+    "\n",
+    "1. **[Irregular Timestamps](https://docs.nixtla.io/docs/tutorials-irregular_timestamps)**\n",
+    "\n",
+    "    - Learn how to deal with irregular timestamps for correct usage of `TimeGPT`.\n",
+    "\n",
+    "2. **[Bounded Forecasts](https://docs.nixtla.io/docs/tutorials-bounded_forecasts)**\n",
+    "\n",
+    "    - Explore `TimeGPT`'s capability to make forecasts within a specified range, ideal for applications where outcomes are bounded.\n",
+    "\n",
+    "3. **[Hierarchical Forecasts](https://docs.nixtla.io/docs/tutorials-hierarchical_forecasting)**\n",
+    "\n",
+    "    - Understand how to use `TimeGPT` to make coherent predictions at various levels of aggregation.\n",
+    "\n",
+    "4. **[Missing Values](https://docs.nixtla.io/docs/tutorials-missing_values)**\n",
+    "\n",
+    "    - Learn how to address missing values within your time series data effectively using `TimeGPT`."
    ]
   }
  ],

nbs/docs/tutorials/12_irregular_timestamps.ipynb

@@ -118,6 +118,18 @@
     ")"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "id": "d3e09434",
+   "metadata": {},
+   "source": [
+    "> 👍 Use an Azure AI endpoint\n",
+    ">\n",
+    "> To use an Azure AI endpoint, set the `base_url` argument:\n",
+    ">\n",
+    "> `nixtla_client = NixtlaClient(base_url=\"you azure ai endpoint\", api_key=\"your api_key\")`"
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
@@ -454,6 +466,22 @@
     ")"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "id": "09f50f5b",
+   "metadata": {},
+   "source": [
+    "> 📘 Available models in Azure AI\n",
+    ">\n",
+    "> If you are using an Azure AI endpoint, please be sure to set `model=\"azureai\"`:\n",
+    ">\n",
+    "> `nixtla_client.forecast(..., model=\"azureai\")`\n",
+    "> \n",
+    "> For the public API, we support two models: `timegpt-1` and `timegpt-1-long-horizon`. \n",
+    "> \n",
+    "> By default, `timegpt-1` is used. Please see [this tutorial](https://docs.nixtla.io/docs/tutorials-long_horizon_forecasting) on how and when to use `timegpt-1-long-horizon`."
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,

nbs/docs/tutorials/13_bounded_forecasts.ipynb

@@ -127,6 +127,18 @@
     ")"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "id": "a9c7b28a",
+   "metadata": {},
+   "source": [
+    "> 👍 Use an Azure AI endpoint\n",
+    ">\n",
+    "> To use an Azure AI endpoint, set the `base_url` argument:\n",
+    ">\n",
+    "> `nixtla_client = NixtlaClient(base_url=\"you azure ai endpoint\", api_key=\"your api_key\")`"
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
@@ -393,6 +405,22 @@
     "timegpt_fcst_with_transform = nixtla_client.forecast(df=df_transformed, h=10, freq='Y', level=[80, 90, 99.5])"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "id": "b6067300",
+   "metadata": {},
+   "source": [
+    "> 📘 Available models in Azure AI\n",
+    ">\n",
+    "> If you are using an Azure AI endpoint, please be sure to set `model=\"azureai\"`:\n",
+    ">\n",
+    "> `nixtla_client.forecast(..., model=\"azureai\")`\n",
+    "> \n",
+    "> For the public API, we support two models: `timegpt-1` and `timegpt-1-long-horizon`. \n",
+    "> \n",
+    "> By default, `timegpt-1` is used. Please see [this tutorial](https://docs.nixtla.io/docs/tutorials-long_horizon_forecasting) on how and when to use `timegpt-1-long-horizon`."
+   ]
+  },
   {
    "cell_type": "markdown",
    "id": "967836fa",

nbs/docs/tutorials/14_hierarchical_forecasting.ipynb

@@ -127,6 +127,18 @@
     ")"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "id": "aacbea74",
+   "metadata": {},
+   "source": [
+    "> 👍 Use an Azure AI endpoint\n",
+    ">\n",
+    "> To use an Azure AI endpoint, set the `base_url` argument:\n",
+    ">\n",
+    "> `nixtla_client = NixtlaClient(base_url=\"you azure ai endpoint\", api_key=\"your api_key\")`"
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
@@ -572,6 +584,22 @@
     "timegpt_fcst = nixtla_client.forecast(df=Y_train_df, h=8, freq='QS', add_history=True)"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "id": "1f0db1ab",
+   "metadata": {},
+   "source": [
+    "> 📘 Available models in Azure AI\n",
+    ">\n",
+    "> If you are using an Azure AI endpoint, please be sure to set `model=\"azureai\"`:\n",
+    ">\n",
+    "> `nixtla_client.forecast(..., model=\"azureai\")`\n",
+    "> \n",
+    "> For the public API, we support two models: `timegpt-1` and `timegpt-1-long-horizon`. \n",
+    "> \n",
+    "> By default, `timegpt-1` is used. Please see [this tutorial](https://docs.nixtla.io/docs/tutorials-long_horizon_forecasting) on how and when to use `timegpt-1-long-horizon`."
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,