Worked on the typesense-website repository to enhance the reliability and correctness of hybrid search functionality, specifically when integrating external embeddings. Addressed a critical bug by ensuring that when external embeddings are supplied, the embedding field is omitted from the query_by parameter, allowing vector_query to utilize the provided embeddings as intended. This adjustment improved search relevance and consistency for users leveraging external embeddings, aligning with ongoing efforts to strengthen hybrid search workflows. The solution was implemented using Markdown for documentation and required careful validation to mitigate regression risks, demonstrating attention to detail in maintaining robust support for external embedding scenarios.