save_image not working in python-kaleido v1

[ZacharyHu0]

> library(plotly)
+ fig <- plot_ly(z = ~volcano) %>% add_surface()
+ save_image(fig, "surface-plot.svg")

Error in `py_get_attr()`:
! AttributeError: module 'kaleido' has no attribute 'scopes'
Run `reticulate::py_last_error()` for details.
Hide Traceback
    ▆
 1. └─plotly::save_image(fig, "surface-plot.svg")
 2.   └─plotly::kaleido()
 3.     ├─kaleido$scopes
 4.     └─reticulate:::`$.python.builtin.module`(kaleido, "scopes")
 5.       └─reticulate::py_get_attr(x, name, FALSE)

I'm running the latest master branch version 4868a68f installed by:
remotes::install_github("plotly/plotly")

R=4.4.3
Python=3.12.12

I've noticed the merge
Add support for kaleido v1.0 (save_image()); fix CI #2447

It seems the merge #2447 didn't work as expected, since "kaleido.scopes.plotly has been removed in v1."

[ZacharyHu0]

here's my Conda list:

# packages in environment at /Volumes/Data/miniconda3/envs/r-reticulate:
#
# Name                     Version          Build                Channel
_openmp_mutex              4.5              7_kmp_llvm           conda-forge
bzip2                      1.0.8            hd037594_8           conda-forge
ca-certificates            2026.1.4         hbd8a1cb_0           conda-forge
choreographer              1.2.1            pyhcf101f3_0         conda-forge
colorama                   0.4.6            pyhd8ed1ab_1         conda-forge
exceptiongroup             1.3.1            pyhd8ed1ab_0         conda-forge
icu                        78.2             h38cb7af_0           conda-forge
iniconfig                  2.3.0            pyhd8ed1ab_0         conda-forge
libblas                    3.11.0           5_h51639a9_openblas  conda-forge
libcblas                   3.11.0           5_hb0561ab_openblas  conda-forge
libcxx                     21.1.8           hf598326_0           conda-forge
libexpat                   2.7.3            haf25636_0           conda-forge
libffi                     3.5.2            he5f378a_0           conda-forge
libgcc                     15.2.0           hcbb3090_16          conda-forge
libgfortran                15.2.0           h07b0088_16          conda-forge
libgfortran5               15.2.0           hdae7583_16          conda-forge
liblapack                  3.11.0           5_hd9741b5_openblas  conda-forge
liblzma                    5.8.2            h8088a28_0           conda-forge
libopenblas                0.3.30           openmp_ha158390_4    conda-forge
libsqlite                  3.51.2           h1ae2325_0           conda-forge
libzlib                    1.3.1            h8359307_2           conda-forge
llvm-openmp                21.1.8           h4a912ad_0           conda-forge
logistro                   2.0.1            pyhcf101f3_0         conda-forge
narwhals                   2.15.0           pyhcf101f3_0         conda-forge
ncurses                    6.5              h5e97a16_3           conda-forge
numpy                      2.4.1            py312he281c53_0      conda-forge
openssl                    3.6.0            h5503f6c_0           conda-forge
orjson                     3.11.5           py312h6f167cc_0      conda-forge
packaging                  26.0             pyhcf101f3_0         conda-forge
pip                        25.3             pyh8b19718_0         conda-forge
plotly                     6.0.0            py_0                 plotly
pluggy                     1.6.0            pyhf9edf01_1         conda-forge
pygments                   2.19.2           pyhd8ed1ab_0         conda-forge
pytest                     9.0.2            pyhcf101f3_0         conda-forge
pytest-timeout             2.4.0            pyhd8ed1ab_0         conda-forge
python                     3.12.12          h18782d2_1_cpython   conda-forge
python-kaleido             1.2.0            pyhcf101f3_0         conda-forge
python_abi                 3.12             8_cp312              conda-forge
readline                   8.3              h46df422_0           conda-forge
setuptools                 80.10.1          pyh332efcf_0         conda-forge
simplejson                 3.20.2           py312h4409184_1      conda-forge
tk                         8.6.13           h892fb3f_3           conda-forge
tomli                      2.4.0            pyhcf101f3_0         conda-forge
typing_extensions          4.15.0           pyhcf101f3_0         conda-forge
tzdata                     2025c            hc9c84f9_1           conda-forge
wheel                      0.46.3           pyhd8ed1ab_0         conda-forge

[cpsievert]

Thanks for reporting this! I've pushed a fix in commit b1103b2.

The issue was that the code was using tryNULL(kaleido$scopes) to detect whether kaleido v1 was installed, but accessing a non-existent Python attribute can throw an AttributeError that doesn't always get caught properly by R's tryCatch. The fix uses reticulate::py_has_attr() instead, which safely checks for attribute existence without triggering the error.

Would you mind reinstalling from GitHub and letting me know if this resolves the issue for you?

remotes::install_github("plotly/plotly.R")

Thanks again for the detailed report!

[dschulz13]

I had the same issue as @ZacharyHu0. The fix solves this error; however, on my Windows laptop I do now instead get the error

Error in py_run_string_impl(code, local, convert) : 
    File "<string>", line 1
    import json; fig = json.load(open('C:\Users\DOMINI~1\AppData\Local\Temp\Rtmpa49Ihy\file26a4639245df.json'))
                                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
SyntaxError: (unicode error) 'unicodeescape' codec can't decode bytes in position 2-3: truncated \UXXXXXXXX escape

when running save_image(). It seems to me like the temporary file path is not adjusted properly to use / instead of \ when being passed to py_run_string() within newKaleidoScope().

Edit:
As a suggestion, the line 138 in the file kaleido.R should read normalizePath(tmp_json, winslash = "/").

[cpsievert]

Thanks for reporting. Would you like to contribute a PR and verify the fix?

[dschulz13]

I would love to, but unfortunately I do not have the time currently.

[ZacharyHu0]

Would you mind reinstalling from GitHub and letting me know if this resolves the issue for you?

remotes::install_github("plotly/plotly.R")

I can confirm the fix works smoothly.
Thank you for your contribution to this community !

my Conda environment for reference:

r-plotly           4.12.0.9000           [f7acae4]
kaleido           1.2.0