Gemini API returns 500 with special characters

[johnnyheineken]

Description of the bug:

This exact code:

text = "čku“ s hodnotami ANO a NE 	+ „"
model = genai.GenerativeModel("gemini-1.5-flash")
result = model.generate_content(
    f"Can you summarize this text?\n{text}"
)
print(f"{result.text=}")

returns

InternalServerError: 500 An internal error has occurred. Please retry or report in https://developers.generativeai.google/guide/troubleshooting

Other malfunctioning strings:

'muláře \t- chybně vyplněný formulář \t- ji'

The problem seems to be with

U+0009 : <control> CHARACTER TABULATION [TAB] {horizontal tabulation (HT); tab}

Actual vs expected behavior:

Actual:
Returns 500

Expected:
Using Czech letters & Czech punctuation will not return 500 and the text will be summarised using Gemini, as requested

Any other information you'd like to share?

Windows 10,

sys.version_info(major=3, minor=9, micro=13, releaselevel='final', serial=0)

google-ai-generativelanguage             0.6.10
google-api-core                          2.17.1
google-api-python-client                 2.119.0
google-auth                              2.25.2
google-auth-httplib2                     0.2.0
google-cloud-aiplatform                  1.49.0
google-cloud-bigquery                    3.17.2
google-cloud-core                        2.4.1
google-cloud-resource-manager            1.12.2
google-cloud-storage                     2.14.0
google-crc32c                            1.5.0
google-generativeai                      0.8.2
google-resumable-media                   2.7.0
google-search-results                    2.4.2
googleapis-common-protos                 1.62.0
grpc-google-iam-v1                       0.13.0
langchain-google-genai                   0.0.6

[Hamza-nabil]

It does work with gemini-1.5-flash-002 the last stable version of Gemini 1.5 Flash (see Model versions and lifecycle).

code :

text = "čku“ s hodnotami ANO a NE 	+ „"
model = genai.GenerativeModel("gemini-1.5-flash-002")
result = model.generate_content(
    f"Can you summarize this text?\n{text}"
)
print(f"{result.text}")

Output :

The text "čku“ s hodnotami ANO a NE" translates from Czech to English as  "check" with values YES and NO.  It describes a binary check or a boolean variable.

[johnnyheineken]

But the docs you sent are saying that I also should be using 002, according to this:

However I confirm that using "gemini-1.5-flash-002" works.

I also pinpointed the issue to U+0009 : <control> CHARACTER TABULATION [TAB] {horizontal tabulation (HT); tab}

[Hamza-nabil]

What was the issue with U+0009 : <control> CHARACTER TABULATION [TAB] {horizontal tabulation (HT); tab} ?

[gmKeshari]

Hi @johnnyheineken . Sometimes we see Internal Server Error (500) because there are too many requests coming, which results in an unexpected error on Google's side. You can refer to this Troubleshooting doc "https://ai.google.dev/gemini-api/docs/troubleshooting?lang=python" I recommend temporarily switching to another model (e.g. from Gemini 1.5 Pro to Gemini 1.5 Flash) and see if it works Or wait a bit and retry your request.