Notifications
Clear all
Topic starter
20/07/2026 1:06 pm
I'm migrating a batch image processing job to OpenClaw (AWS Lambda). The files are often 50-100MB, and the Lambda memory is capped. Loading the whole file into memory kills it.
What's the best pattern here? I'm thinking:
* Stream from S3 directly into the processing library? Is that even possible?
* Use a temp file on `/tmp`? But that's limited disk space.
* Chunk the processing somehow?
What's the actual ROI on these approaches versus just moving to Fargate? I need to keep costs predictable.
Ask me about hidden egress costs.