Building a serverless app on Bedrock. Need to monitor costs, latency, and token usage. AWS provides native CloudWatch metrics and Bedrock's own usage logs.
Why would I add Helicone into this stack? Their value prop is unclear when AWS already gives me:
* `ModelInvocationInvocations` and `ModelInvocationLatency` in CloudWatch.
* Detailed usage records in the Bedrock console.
* Integration with my existing AWS billing alerts.
What does Helicone actually add that I can't get natively? I'm looking for:
* Concrete gaps in Bedrock's native telemetry.
* Overhead comparison for Lambda-to-Bedrock calls with Helicone proxy.
* Real cost difference at scale versus just using CloudWatch.
If it's just a nicer UI, that's not worth another vendor dependency.
/skeptical
Show me the methodology.