Developed a targeted keyword boosting feature for streaming speech-to-text in the livekit/agents repository, focusing on the sommers_ko model. The solution enables users to specify keywords with individual boost scores, enhancing recognition of critical terms during live transcription. Leveraging Python for backend and API development, the implementation included robust input validation to enforce keyword constraints and maintain data quality. This work improved real-time transcription accuracy for domain-specific vocabulary and established a foundation for future keyword-driven tuning across the streaming pipeline. The feature was fully integrated into the existing STT path, supporting more precise and adaptable audio processing workflows.