Bug Description

In the file cognee/infrastructure/databases/vector/embeddings/LiteLLMEmbeddingEngine.py, the init method correctly receives and stores the dimensions parameter (corresponding to cognee's EMBEDDING_DIMENSIONS configuration, default 3072), which is used for get_vector_size() return value and mock mode vector generation ([0.0] * self.dimensions).
However, in the embed_text() method's actual embedding branch, the call to litellm.aembedding() only passes model, input, api_key, api_base, and api_version parameters, without passing the dimensions parameter. This causes LiteLLM to use the model's default embedding dimension (e.g., 3072 for OpenAI text-embedding-3-large), ignoring cognee's custom configuration.

Steps to Reproduce

1. Configure cognee's EMBEDDING_DIMENSIONS=512.
2. Initialize LiteLLMEmbeddingEngine and call embed_text("test").
3. Check the returned vector length: expected 512, but actual is model default (e.g., 3072).

Expected Behavior

litellm.aembedding() should receive the dimensions=self.dimensions parameter (supported by models like OpenAI), ensuring the output dimension matches cognee's configuration.

Actual Behavior

Add dimensions=self.dimensions to the litellm.aembedding() call in the embed_text() method:

response = await litellm.aembedding(
    model=self.model,
    input=texts,
    api_key=self.api_key,
    api_base=self.api_base,
    api_version=self.api_version,
    dimensions=self.dimensions  # Add this line
)

Environment

• OS: [macOS 15.6.1]
• Python version: [3.12.11]
• Cognee version: [0.5.1]
• LLM Provider: [OpenAI]
• Database: [Neo4j,qdrant]

Logs/Error Messages

Additional Context

No response

Pre-submission Checklist

• I have searched existing issues to ensure this bug hasn't been reported already
• I have provided a clear and detailed description of the bug
• I have included steps to reproduce the issue
• I have included my environment details