Inside Amazon Bedrock, AWS splits operations into two distinct areas: Management and Data/Execution. Because of this, boto3 has two different clients for Bedrock:
bedrock (The Management Client): Used for administrative tasks. You use this to create Knowledge Bases, set up Guardrails, or kick off a model fine-tuning job.
bedrock-runtime (The Execution Client): Used strictly for high-performance data operations—specifically, asking models to generate text, images, or embeddings.
In a live startup application, your backend API (like FastAPI) will use bedrock-runtime 99% of the time because it is optimized for speed, low latency, and handling streaming responses when users ask questions.
No comments:
Post a Comment