ExecBrief - A RAG App powered by Bedrock

To build my Retrieval Augmented Generation (RAG) application using Amazon Bedrock to ask questions about video or audio recordings as the source, you can follow these steps:

1. Upload the video transcripts to Amazon S3: Create an s3 bucket, and create input and output sub folders. Upload your recordings to input folder under your newly created s3 bucket.
2. Using Amazon Transcribe create a Transcription job, give Input data location as your input sub folder under your s3 bucket. Set Output data location as the output sub folder under your s3 bucket. Run the job. Your transcripts will be places under S3 bucket output folder. You can notice metadata file as well.
3. Create a Knowledge Base in Amazon Bedrock: Go to the Amazon Bedrock console and create a new Knowledge Base. During the creation process, you'll need to specify the S3 bucket output folder location where you have the meeting transcripts. Amazon Bedrock will index the text content and create vector embeddings for efficient retrieval. you can pick Titan Embeddings G1 - Textv1.2 for example for embedding.
4. To generate responses, you can pick model of your choice, for example i picked Claude Instant 1.2 Text Model. You can test the responses, and the knowledge base on bedrock playground.
5. Finally you can Integrate your knowledge base into your application as is or add it to agents.

Please note the app will allow you to ask questions about any YouTube video as well. If video is lengthy and you don't have time to watch full video, here's a solution!