[Bug]: String data: stripped from entire content in streamed Gemini responses

[lokflam]

What happened?

When streaming responses from Gemini models, the string data: is unexpectedly removed from the entire content, not just from the SSE message prefix.

Potential cause:

The issue is likely cause by from this line of code, which appears to remove all instances of data: from the streaming chunks:
https://github.com/BerriAI/litellm/blob/v1.59.10/litellm/llms/vertex_ai/gemini/vertex_and_google_ai_studio_gemini.py#L1434

Steps to reproduce:

1. Non-streaming response

   from litellm import completion
   import os
   
   response = completion(
       model="vertex_ai/gemini-2.0-flash-exp",
       messages=[{"role": "user", "content": "print `somedata: abc` in yaml"}],
   )
   
   print(response.choices[0].message.content)

   Output

   somedata: abc

2. Streaming response

   from litellm import completion, stream_chunk_builder
   import os
   
   response = completion(
       model="vertex_ai/gemini-2.0-flash-exp",
       messages=[{"role": "user", "content": "print `somedata: abc` in yaml"}],
       stream=True,
   )
   
   chunks = []
   for chunk in response:
       chunks.append(chunk)
   
   print(stream_chunk_builder(chunks).choices[0].message.content)

   Output

   some abc

Relevant log output

Are you a ML Ops Team?

No

What LiteLLM version are you on ?

v1.59.10

Twitter / LinkedIn details

No response