reserveai 1.0.0rc3

Python SDK for insureAI's ReserveAI API

ReserveAI SDK

By insureAI

📔 Table of Contents

• 📔 Table of Contents
  • 🌟 About the Project
  • 🎯 Features
  • 🧰 Install
  • 👀 Getting Started
  • ⚠️ License
  • 🤝 Contact
  • 💎 Acknowledgements and Citations

🌟 About the Project

The ReserveAI SDK (by insureAI) provides exclusive access to a powerful new approach for actuarial reserving. Designed to seamlessly integrate with ReserveAI and move beyond the limitations of traditional, static analysis, this toolkit introduces a systematic framework to address the core challenges of modern reserving: eliminating subjectivity, managing volatility, and providing a robust, defensible basis for your ultimate liability estimates.

At its core, the SDK allows you to effortlessly back-test thousands of model variations against your own data. Our proprietary evaluation engine assesses each model on multiple dimensions of performance. This provides a holistic and uniquely insightful view of model quality that conventional methods fail to capture. This SDK is designed for actuaries and risk managers who need to conduct reserve reviews with greater confidence and efficiency.

To get started, we recommend visiting our documentation:

• The Tutorials provide step-by-step examples on using the SDK.
• The FAQ has quick answers to common questions.
• The Methodologies which may be accessed after contacting [e-mail].

🎯 Features

• Run industry-standard actuarial methods: Run reserving methods such as Cape Cod, Chain Ladder, and Bornhuetter–Ferguson (with more methods in active development).

• Triangle data utilities: Convenient representation of triangle datasets:

  • Switch between long or wide (triangle) views.
  • Plot standard reserving visuals.
  • Run quick diagnostics.

• Data science friendly: Full interoperability with Pandas and Polars: input data and results can be seamlessly converted to and from dataframes. Thus, for more complex workflows, the full Python data science ecosystem can be brought to bear by extracting the underlying dataframes.

• Automated model exploration: Each reserving run evaluates hundreds or thousands of model variations. The models are returned for review, comparison, and deeper analysis.

• Rich model diagnostics: Inspect detailed outputs for chosen models, including triangle projections, cubes, and performance metrics, to support robust reserve reviews.

• Server-side computation: All reserving runs execute on ReserveAI’s secure servers; the SDK provides a lightweight yet seamless interface for managing runs and fetching results through the ReserveAI REST API.

• Multiple-line multiple-target support: Run multiple reserving methods on multiple lines of business and multiple target variables simultaneously.

🧰 Install

The SDK is available via PyPI, and hence can be installed using your preferred package manager:

• Using pip:

  pip install reserveai[all]
  

• Using poetry:

  poetry add reserveai[all]
  

• Using uv:

  uv add reserveai[all]
  

There are several optional dependencies that can be installed using the [optional] extras syntax:

• pandas: Installs the pandas dependency (e.g., uv add reserveai[pandas]).
• polars: Installs the polars dependency.
• plotly: Installs the plotly dependency.

These can be combined, e.g., reserveai[polars,plotly]. Use reserveai[all] to install all optional dependencies.

👀 Getting Started

To use the SDK against the ReserveAI API, you'll need ReserveAI client credentials. To obtain these, please contact us at [e-mail]. Don't have credentials yet? Not a problem — the notebooks in the tutorials and recipes section that show how to get started with the SDK are already pre-rendered!

⚠️ License

The ReserveAI SDK is distributed under the Apache License 2.0; see LICENSE.txt for more information. (That license applies only to the SDK code; it does not grant any rights to insureAI's proprietary APIs, algorithms, or platforms.)

The use of the API (via the SDK, or directly) is subject to the API Terms of Use.

🤝 Contact

We are not actively accepting contributions at this time. However, if you encounter a bug, have a question, or wish to report a security vulnerability, please reach out to us at [e-mail].

💎 Acknowledgements and Citations

Full data attributions are available in The Actuary and IBNR Techniques: A Machine Learning Approach by Balona and Richman.