Okay, maybe I'm using this wrong, but the auto-classify feature keeps messing up my project tags. I'm trying to organize some pretty specific AWS resources (like a NAT gateway setup for a private subnet), and Hailuo keeps tagging it under generic "Networking" or even "Compute" instead of "VPC" or "Cost Management."
It's frustrating because I rely on those tags for reports. For example:
```hcl
resource "aws_nat_gateway" "main" {
allocation_id = aws_eip.nat.id
subnet_id = aws_subnet.public.id
tags = {
Project = "core-infra"
CostCenter = "network-ops" # Hailuo often misses this context
}
}
```
When I review the auto-classified suggestions, it proposes tags like "Gateway" (too broad) or ignores the cost-saving intent of the architecture. Has anyone else run into this with niche cloud resources? How do you handle it? Do you just turn auto-classify off?
You're definitely not alone. The underlying issue I've observed is that these auto-classification systems are often trained on broader public datasets, not the internal taxonomy of a specific engineering organization. Your NAT gateway is a perfect example: it's a concrete implementation detail that sits at the intersection of networking, security, and cost management - context the model likely lacks.
For anything beyond a very generic classification, I've found it necessary to turn the feature off and rely on a documented tagging schema. We enforce this at the pipeline level: a deployment fails if resources don't include tags from our controlled dictionary. It's more upfront work, but it eliminates the reporting ambiguity you're experiencing.
Have you considered pushing your tagging schema back into Hailuo via custom categories? Some teams use that to gradually train the classifier on their specific domain, though results have been mixed in my experience.
null