My ADHD cannot handle no timestamps 0:00 - Introduction and Welcome 0:11 - Introduction to Amazon Bedrock 0:48 - Overview of Amazon Bedrock Agents 1:21 - Brief Introduction to StratusGrid 2:30 - Stratosphere: Cost Optimization Tool 3:50 - Detailed Explanation of Amazon Bedrock Agent Workflow 9:08 - Example of Retrieving Data with Bedrock Agents 11:02 - Setting Up and Testing Bedrock Agents 25:01 - Creating a New Bedrock Agent from Scratch 34:01 - Testing and Debugging Bedrock Agent 47:00 - Implementing and Validating Lambda Functions 56:00 - Limitations and Additional Features of Bedrock Agents 57:03 - Conclusion and Final Thoughts
Great video Trevor! If you ever do a follow-up on how to use a KB with an Agent that'd be great. FYI, I'm working on a recommendation engine that gets data from a Bedrock KB and wanting to have it as a frontpage bot. For newbies, it is a bit confusing how to set this up. A prompt flow allowed me to clean the kb results via a lambda but a prompt flow doesn't give me the ability to subsequently drill down on the resultset: it has no memory. An agent has memory but how do I guide it along with the need to clean up the results from a kb via lambda etc. I'm going in circle.... Anyway, I love your videos, thank you for posting them! 😊
Thanks for your kind comment, and the video idea! Have you checked out the new Prompt Flows feature? That might accomplish what you're trying to do easily, combining Knowledge Bases and Agents. It's in preview, and was just announced a month ago. I just learned about it a few minutes ago, when I logged into the Bedrock console!
And now I see your specific comment about Prompt Flows .... "doesn't give me the ability to subsequently drill down on the resultset". Hmmmm, I would have to think about how to solve that. I am pretty new to LLMs and Bedrock in particular! If you're ever interested in chatting and brainstorming about the topic, let me know.
@@codeinrust Hi! How to reconcile the concept of a Bedrock Agent (for memory) and the use of prompt flows... I don't think I can! My use case is a presale bot recommendation engine. So far, getting the results via Lambda from a bedrock kb works, but I really hope I can find out how to get the bot to drill down/filter results previously obtained when the user provides more precision. It is a fun adventure that is for sure.
My ADHD cannot handle no timestamps
0:00 - Introduction and Welcome
0:11 - Introduction to Amazon Bedrock
0:48 - Overview of Amazon Bedrock Agents
1:21 - Brief Introduction to StratusGrid
2:30 - Stratosphere: Cost Optimization Tool
3:50 - Detailed Explanation of Amazon Bedrock Agent Workflow
9:08 - Example of Retrieving Data with Bedrock Agents
11:02 - Setting Up and Testing Bedrock Agents
25:01 - Creating a New Bedrock Agent from Scratch
34:01 - Testing and Debugging Bedrock Agent
47:00 - Implementing and Validating Lambda Functions
56:00 - Limitations and Additional Features of Bedrock Agents
57:03 - Conclusion and Final Thoughts
Great video Trevor! If you ever do a follow-up on how to use a KB with an Agent that'd be great.
FYI, I'm working on a recommendation engine that gets data from a Bedrock KB and wanting to have it as a frontpage bot. For newbies, it is a bit confusing how to set this up. A prompt flow allowed me to clean the kb results via a lambda but a prompt flow doesn't give me the ability to subsequently drill down on the resultset: it has no memory. An agent has memory but how do I guide it along with the need to clean up the results from a kb via lambda etc. I'm going in circle....
Anyway, I love your videos, thank you for posting them! 😊
Thanks for your kind comment, and the video idea! Have you checked out the new Prompt Flows feature? That might accomplish what you're trying to do easily, combining Knowledge Bases and Agents. It's in preview, and was just announced a month ago. I just learned about it a few minutes ago, when I logged into the Bedrock console!
Actually, I now see you commented 11 days ago on the AWS video about Prompt Flows. So, apparently you already know about it. 😉
And now I see your specific comment about Prompt Flows .... "doesn't give me the ability to subsequently drill down on the resultset". Hmmmm, I would have to think about how to solve that. I am pretty new to LLMs and Bedrock in particular! If you're ever interested in chatting and brainstorming about the topic, let me know.
@@codeinrust Hi! How to reconcile the concept of a Bedrock Agent (for memory) and the use of prompt flows... I don't think I can! My use case is a presale bot recommendation engine. So far, getting the results via Lambda from a bedrock kb works, but I really hope I can find out how to get the bot to drill down/filter results previously obtained when the user provides more precision. It is a fun adventure that is for sure.
How can I timely invoke agent after a certain, after every 10 min for that too I have to declare the lambda function .