U+0CA0 (KAṄNAḌA LETTER ṬṬA) [aka "disapproving eyes"] does not always render properly

[jamisonbryant]

What happened?

When prompting o1-mini to jostle my memory of the "disapproving eyes" ASCII art thingy that was popular on the internet a while back, I noticed that it got the right answer but the characters for the eyes were rendered inconsistently. I included sample prompts and screenshots below.

It's odd to me that it renders properly sometimes, but not other times, e.g. in the same sentence. I would expect it if renders once correctly that character should render correctly all the time.

Version Information

ghcr.io/danny-avila/librechat-dev                                        latest                     af6927decdf4   2 weeks ago     918MB
ghcr.io/danny-avila/librechat-rag-api-dev-lite                           latest                     49a785a50986   2 weeks ago     1.28GB

Steps to Reproduce

Prompt 1:

what was that one "eyes" ascii art you used to see on reddit a lot? It looked like disapproving eyes

Response:

Prompt 2 (getting more detail for this bug report):

what is the technical name for the character that defines the eyes?

Response:

What browsers are you seeing the problem on?

Brave Version 1.77.100 Chromium: 135.0.7049.100 (Official Build) (arm64)

Relevant log output

Raw logs:

chat-meilisearch  | 2025-04-25T13:03:23.680871Z  INFO HTTP request{method=GET host="meilisearch:7700" route=/indexes/convos/documents/51024b41-21fb-4fac-9269-3a90f5715cdd query_parameters= user_agent=node status_code=200}: meilisearch: close time.busy=18.9ms time.idle=8.26ms
chat-meilisearch  | 2025-04-25T13:03:23.691585Z  INFO HTTP request{method=POST host="meilisearch:7700" route=/indexes/messages/documents query_parameters= user_agent=node status_code=202}: meilisearch: close time.busy=17.5ms time.idle=37.1ms
chat-meilisearch  | 2025-04-25T13:03:24.588909Z  INFO index_scheduler::batch: document indexing done indexing_result=DocumentAdditionResult { indexed_documents: 1, number_of_documents: 295 } processed_in=891.891001ms
chat-meilisearch  | 2025-04-25T13:03:24.636935Z  INFO index_scheduler: A batch of tasks was successfully completed with 1 successful tasks and 0 failed tasks.
chat-meilisearch  | 2025-04-25T13:03:46.419669Z  INFO HTTP request{method=POST host="meilisearch:7700" route=/indexes/messages/documents query_parameters= user_agent=node status_code=202}: meilisearch: close time.busy=3.96ms time.idle=6.14ms
chat-meilisearch  | 2025-04-25T13:03:46.435131Z  INFO HTTP request{method=POST host="meilisearch:7700" route=/indexes/messages/documents query_parameters= user_agent=node status_code=202}: meilisearch: close time.busy=1.67ms time.idle=7.58ms
chat-meilisearch  | 2025-04-25T13:03:46.471347Z  INFO HTTP request{method=GET host="meilisearch:7700" route=/indexes/convos/documents/51024b41-21fb-4fac-9269-3a90f5715cdd query_parameters= user_agent=node status_code=200}: meilisearch: close time.busy=541µs time.idle=2.02ms
chat-meilisearch  | 2025-04-25T13:03:46.662515Z  INFO HTTP request{method=POST host="meilisearch:7700" route=/indexes/messages/documents query_parameters= user_agent=node status_code=202}: meilisearch: close time.busy=3.12ms time.idle=12.9ms
chat-meilisearch  | 2025-04-25T13:03:47.027225Z  INFO index_scheduler::batch: document indexing done indexing_result=DocumentAdditionResult { indexed_documents: 1, number_of_documents: 295 } processed_in=607.766626ms
chat-meilisearch  | 2025-04-25T13:03:47.039686Z  INFO index_scheduler: A batch of tasks was successfully completed with 1 successful tasks and 0 failed tasks.
chat-meilisearch  | 2025-04-25T13:03:47.155674Z  INFO index_scheduler::batch: document indexing done indexing_result=DocumentAdditionResult { indexed_documents: 2, number_of_documents: 296 } processed_in=110.171ms
chat-meilisearch  | 2025-04-25T13:03:47.245476Z  INFO index_scheduler: A batch of tasks was successfully completed with 2 successful tasks and 0 failed tasks.


I don't think there was an actual error, just unexpected behavior.

Screenshots

No response

Code of Conduct

• I agree to follow this project's Code of Conduct

[danny-avila]

It's likely because the LLM is enclosing the character with inline-code.

See my example, it renders correctly: https://demo.librechat.ai/share/c217qBPwFCKJxsB_GlyT2

[jamisonbryant]

You're right that appears to be the differentiator. So it's a problem with the model, then?

Thanks for being so responsive. This is my first bug report for this repo.