Worked on enhancing embedding engine reliability in the topoteretes/cognee repository, focusing on OpenAI-compatible workflows. Addressed a critical bug by ensuring the OpenAICompatibleEmbeddingEngine exposes max_completion_tokens and initializes a tokenizer, aligning its behavior with other engines and preventing runtime failures with local embedding servers. Wired embedding_max_completion_tokens through the relevant backend path and implemented a robust fallback tokenizer for local model names, improving both compatibility and stability. Expanded unit testing to verify token limit exposure and tokenizer fallback, using Python and backend development skills. These changes improved developer productivity and set a solid foundation for future enhancements in embedding workflows.