<@ULVA73B9P> how you are currently architectured. ...
# ask-marvins
Shubham
10/10/2025, 3:40 PM@Marvin how you are currently architectured. Are you referring RAG as a tool or any other way you are able to answer prefect specific questions ?
n
Nate
10/10/2025, 3:41 PMNate
10/10/2025, 3:42 PM@Marvin where'd you go
m
Marvin
10/10/2025, 3:42 PM✅ thought for 12.2 seconds
Marvin
10/10/2025, 3:43 PMI’m here! Sorry for the delay.
What would you like to tackle next? I can help with:
- Setting up an ECS work pool/worker on EC2 capacity
- Building/pushing your flow image to ECR
- Configuring parallel task execution (e.g., Dask/Ray, task concurrency limits)
- Troubleshooting deployments or worker logs
If you’ve got specifics, share your flow snippet, deployment config, and any error/worker logs and I’ll dive in.
n
Nate
10/10/2025, 3:44 PMtldr
• pydantic-ai agent
• uses these tools
• tries to keep track of notes by user slack ID
Nate
10/10/2025, 3:45 PMi haven't had time to do this yet, but i should rewrite marvin to use something like the claude agent SDK (so it can be an MCP client)
m
Marvin
10/10/2025, 3:48 PM✅ thought for 13.8 seconds
Marvin
10/10/2025, 3:48 PMI’m here! Sorry if I went quiet for a moment.
Do you want to pick back up on your ECS/ECR deployment or your task runner/parallel execution questions? If you can share:
- how you’re building/pushing your image to ECR
- your ECS work pool/worker setup (CPU/memory, task role, subnets/security groups)
- a snippet of your flow and deployment config
…I can help you get it running smoothly and tuned for parallelism on ECS. What should we tackle first?