revised the paper draft

paper/paper.bib

@@ -125,3 +125,11 @@ @article{fu2023comparative
   year={2023},
   publisher={The International Archives of the Photogrammetry, Remote Sensing and Spatial~…}
 }
+
+@Article{csardi2006igraph,
+  title={The igraph software package for complex network research},
+  author={Csardi, G. and Nepusz, T.},
+  journal={InterJournal, Complex Systems},
+  pages={1695},
+  year={2006}
+}

paper/paper.md

@@ -11,20 +11,25 @@ authors:
   - name: Xiaokang Fu
     orcid: 0000-0002-3396-6720
     affiliation: "1"
+  - name: Devika, Kakkar
+    orcid:
+    affiliation: "2"
 affiliations:
   - name: "State Key Laboratory of Information Engineering in Surveying, Mapping and Remote Sensing, Wuhan University, 129 Luoyu Road, Wuhan, China"
     index: 1
+  - name: "Center for Geographic Analysis, Harvard University, 1737 Cambridge Street, MA, USA"
+    index: 2
 date: 19 November 2025
 bibliography: paper.bib
 ---
 
 # Summary
 
-`georouting` is a Python package that provides a unified interface to multiple web routing services, currently including Google Maps, Bing Maps, the open-source OSRM server [@luxen-vetter-2011], Esri services, Baidu Maps, and OpenStreetMap-based routing through OSMnx [@boeingModelingAnalyzingUrban2025]. It enables researchers and practitioners to compute travel routes, travel times, and travel distances between large numbers of origin–destination (OD) pairs using a consistent and high-level API. The package is designed to integrate naturally with the scientific Python ecosystem by returning results as `pandas` DataFrames [@mckinney2011pandas] and `geopandas` GeoDataFrames [@kelsey_jordahl_2020_3946761], and by exposing convenient methods for map-based visualization of routes and accessibility patterns. In many applied domains—such as transport planning, health-care accessibility, logistics, and urban analytics—accurate travel-time and distance estimates are needed to replace simple Euclidean distances or crude network approximations. `georouting` lowers the barrier to using production-grade web routing services by handling provider-specific request formats, batching strategies, and response parsing behind a single API. The package is available on both PyPI and conda-forge, comes with automated tests and continuous integration, and includes AI-oriented documentation (via `llms.txt` [@jeremy_howard_2024]) to make it easier for users to get support from large language models. The package is open source (MIT license) and available at https://github.com/wybert/georouting.
+`georouting` is a Python package that provides a unified interface to multiple web routing services and open-source local routing solutions, currently including Google Maps, Bing Maps, Esri services, Baidu Maps, TomTom Maps, Here Maps, Mapbox, OpenRouteService, an OSRM local/demo server [@luxen-vetter-2011] with helper functions to auto-build and start Docker or Singularity backends (suitable for HPC environments), and routing through OSMnx [@boeingModelingAnalyzingUrban2025] and iGraph [@csardi2006igraph]. It enables researchers and practitioners to compute travel routes, travel times, and travel distances between large numbers of origin–destination (OD) pairs using a consistent and high-level API. The package is designed to integrate naturally with the scientific Python ecosystem by returning results as `pandas` DataFrames [@mckinney2011pandas] and `geopandas` GeoDataFrames [@kelsey_jordahl_2020_3946761], and by exposing convenient methods for map-based visualization of routes and accessibility patterns. In many applied domains—such as transport planning, health-care accessibility, logistics, and urban analytics—accurate travel-time and distance estimates are needed to replace simple Euclidean distances or crude network approximations. `georouting` lowers the barrier to using production-grade web routing services by handling provider-specific request formats, batching strategies, and response parsing behind a single API. For OSRM, it includes Geofabrik region utilities (with optional size lookup) and a one-shot helper that downloads extracts, builds, and launches a local service with the chosen profile (car/foot/bicycle) via Docker or Singularity/Apptainer. The package is available on both PyPI and conda-forge, comes with automated tests and continuous integration, and includes AI-oriented documentation (via `llms.txt` [@jeremy_howard_2024]) to make it easier for users to get support from large language models. The package is open source (MIT license) and available at https://github.com/wybert/georouting.
 
 # Statement of Need
 
-There is a growing need in the health-care, geospatial, and data-science communities for tools that can compute realistic travel distances and travel times at scale. There are many tools available including web-based routing services like Google Maps and Bing/Azure Maps and open source routing engines like OSRM and pgRouting [@pgrouting_2025]. The researchers often need to compare the routing results from different providers or switch between providers due to data-coverage, accuracy, pricing, or licensing constraints. However, each tool has its own workflow and often limited built-in visualization support, making it difficult to cross verify results or switch between providers without significant refactoring. `georouting` aims to address this gap by providing a single, extensible, Pythonic, and AI-friendly routing interface that abstracts over different tool-specific details while still exposing tool-specific strengths. Users can change routing methods by switching a class (e.g., `GoogleRouter`, `BingRouter`, `OSRMRouter`, `BaiduRouter`, `EsriRouter`, or an `OSMnx`-based router) without refactoring their analysis pipeline. Common operations—such as computing a route between two coordinates, building a full OD distance matrix, or computing distances only for specific OD pairs—are implemented in a backend-agnostic way. It also provides a `plot_route()` method to visualize the routing results on an interactive map. We provide llms.txt files to make it easier for users to get AI-powered help when working with the package using vibe coding tools like Cursor or Claude Code. The package is intended for geospatial researchers, transportation and health geographers, operations researchers, and data scientists who need reproducible and scalable routing and accessibility analyses from Python. `georouting` has been used in the rapidroute project [@rapidroute], enabling researchers to compute drive times for millions of OD pairs at the national scale within minutes on high-performance computing systems, as well as in a study evaluating drive-time estimation methods on geospatial big data [@fu2023comparative].
+There is a growing need in the health-care, geospatial, and data-science communities for tools that can compute realistic travel distances and travel times at scale. There are many tools available including web-based routing services like Google Maps, Bing/Azure Maps, Esri, Baidu, TomTom, Here, Mapbox, OpenRouteService, and open source routing engines like OSRM and pgRouting [@pgrouting_2025]. Researchers often need to compare results from different providers or switch providers due to coverage, accuracy, pricing, or licensing constraints. However, each tool has its own workflow and often limited built-in visualization support, making it difficult to cross verify results or switch providers without significant refactoring. `georouting` addresses this gap by providing a single, extensible, Pythonic, and AI-friendly routing interface that abstracts over tool-specific details while still exposing tool-specific strengths. Users can change routing methods by switching a class (e.g., `GoogleRouter`, `OSRMRouter`, `EsriRouter`, `MapboxRouter`, `TomTomRouter`, `HereRouter`, `OpenRouteServiceRouter`, or an `OSMnx`-based router) without refactoring their analysis pipeline. Common operations—such as computing a route between two coordinates, building a full OD distance matrix, or computing distances only for specific OD pairs—are implemented in a backend-agnostic way. It also provides a `plot_route()` method to visualize the routing results on an interactive map. We provide llms.txt files to make it easier for users to get AI-powered help when working with the package using vibe coding tools like Cursor or Claude Code. For users who need a local engine, `georouting` ships helper functions to fetch Geofabrik extracts (with optional size metadata), generate an OSRM Dockerfile, and auto-build/run an OSRM backend with the desired profile (car/foot/bicycle) and region via Docker or Singularity/Apptainer (suited to HPC environments). The package is intended for geospatial researchers, transportation and health geographers, operations researchers, and data scientists who need reproducible and scalable routing and accessibility analyses from Python. `georouting` has been used in the rapidroute project [@rapidroute], enabling researchers to compute drive times for millions of OD pairs at the national scale within minutes on high-performance computing systems, as well as in a study evaluating drive-time estimation methods on geospatial big data [@fu2023comparative].
 
 # Georouting Functionality and Design
 
@@ -39,7 +44,7 @@ The underlying route objects (e.g., `GoogleRoute`, `BingRoute`, `OSRMRoute`) are
 
 ![Figure.\label{fig:ploting}](Screenshot1.png){ width=60% }
 
-To support large OD problems, `georouting` implements helper utilities that partition OD pairs into batches that respect provider-specific API limits while avoiding redundant calls. These utilities work with vectorized data structures (e.g., Numpy arrays or `pandas` DataFrames) and are designed to minimize the number of remote API requests. The package is written in pure Python and integrates closely with widely used geospatial libraries such as `geopandas` and `osmnx`, enabling workflows that combine web-based routing with open street network analysis.
+To support large OD problems, `georouting` implements helper utilities that partition OD pairs into batches that respect provider-specific API limits while avoiding redundant calls. These utilities work with vectorized data structures (e.g., Numpy arrays or `pandas` DataFrames) and are designed to minimize the number of remote API requests. For users needing on-prem/HPC execution, `georouting` ships OSRM helpers that can download Geofabrik extracts, build images, and launch local OSRM backends via Docker or Singularity/Apptainer—allowing high-throughput OD computation without relying on external APIs. The package is written in pure Python and integrates closely with widely used geospatial libraries such as `geopandas` and `osmnx`, enabling workflows that combine web-based routing with open street network analysis.
 
 The project follows standard Python packaging practices, ships with an automated test suite that covers core routing and batching logic (including OSRM-based tests that do not require API keys), and is continuously tested on multiple Python versions using GitHub Actions. Documentation is built with MkDocs, including narrative tutorials, usage examples, and API reference pages generated from docstrings. In addition, the project maintains `llms.txt` and `llms-full.txt` files to provide machine-readable documentation for large language models, making it easier for users to obtain high-quality automated assistance.