removing the base64 encoding

Signed-off-by: Francisco Javier Arceo <farceo@redhat.com>

llama_stack/providers/remote/inference/ollama/ollama.py

@@ -71,7 +71,6 @@ from llama_stack.providers.utils.inference.openai_compat import (
     process_chat_completion_stream_response,
     process_completion_response,
     process_completion_stream_response,
-    process_embedding_b64_encoded_input,
 )
 from llama_stack.providers.utils.inference.prompt_adapter import (
     chat_completion_request_to_prompt,
@@ -397,6 +396,7 @@ class OllamaInferenceAdapter(
         if model_obj.provider_resource_id is None:
             raise ValueError(f"Model {model} has no provider_resource_id set")
 
+        # Note, at the moment Ollama does not support encoding_format, dimensions, and user parameters
         params = prepare_openai_embeddings_params(
             model=model_obj.provider_resource_id,
             input=input,
@@ -404,9 +404,6 @@ class OllamaInferenceAdapter(
             dimensions=dimensions,
             user=user,
         )
-        # Note, at the moment Ollama does not support encoding_format, dimensions, and user parameters
-        # but we implement the encoding here
-        params = process_embedding_b64_encoded_input(params)
 
         response = await self.openai_client.embeddings.create(**params)
         data = b64_encode_openai_embeddings_response(response.data, encoding_format)

llama_stack/providers/utils/inference/openai_compat.py

@@ -1515,24 +1515,6 @@ def prepare_openai_embeddings_params(
     return params
 
 
-def process_embedding_b64_encoded_input(params: dict[str, Any]) -> dict[str, Any]:
-    """
-    Process the embeddings parameters to encode the input in base64 format if specified.
-        Currently implemented for ollama as base64 is not yet supported by their compatible API.
-    """
-    if params.get("encoding_format") == "base64":
-        processed_params = params.copy()
-        input = params.get("input")
-        if isinstance(input, str):
-            processed_params["input"] = base64.b64encode(input.encode()).decode()
-        elif isinstance(input, list):
-            processed_params["input"] = [base64.b64encode(i.encode()).decode() for i in input]
-    else:
-        return params
-
-    return processed_params
-
-
 def b64_encode_openai_embeddings_response(
     response_data: dict, encoding_format: str | None = "float"
 ) -> list[OpenAIEmbeddingData]:

tests/integration/inference/test_openai_embeddings.py

@@ -34,11 +34,7 @@ def skip_if_model_doesnt_support_variable_dimensions(model_id):
         pytest.skip("{model_id} does not support variable output embedding dimensions")
 
 
-@pytest.fixture(
-    params=[
-        "openai_client",
-    ]
-)
+@pytest.fixture(params=["openai_client", "llama_stack_client"])
 def compat_client(request, client_with_models):
     if request.param == "openai_client" and isinstance(client_with_models, LlamaStackAsLibraryClient):
         pytest.skip("OpenAI client tests not supported with library client")