Agentcore Memory Throttling issue #160
Description
Describe the bug
I am trying to deploy a Google ADK agent into Agentcore Runtime connected with Agentcore Memory. My application is quite basic and will return weather information based on the city we request and for now i have all weather data hard-coded and i do not rely on any external API.
Since Google ADK do not have a native integration for Agentcore Memory, i created a custom class to integrate with Agentcore memory.
While testing the feature i intermittently received below throttling exception at runtime.
{"error": "Runtime error: An error occurred (ThrottledException) when calling the CreateEvent operation (reached max retries: 4): Rate exceeded.", "type": "runtime_error", "details": "An error occurred (ThrottledException) when calling the CreateEvent operation (reached max retries: 4): Rate exceeded."}
To Reproduce
Throttling exception occur in random while testing in local
Expected behavior
I should not be receiving throttling exception
Desktop (please complete the following information):
- OS: Linux [Ubuntu WSL on Windows]
- Browser : NA as i was testing with Agentcore starter toolkit
- Version : NA
Additional context
I checked this link https://docs.aws.amazon.com/bedrock-agentcore/latest/devguide/bedrock-agentcore-limits.html#memory-limits to understand the quotas and limit of Agentcore Memory and i noticed below limit which seems to be more restrictive.
Maximum CreateEvent requests per actor, per session, including conversational payloads: 0.25 req/s
which means 1 event per 4 seconds per session which seems to be more restrictive.
How can we work with this restriction? I can add some exponential backoff logic but this will increase the overall response time of this conversational agent. I feel like for real production usecases this can become a problem. Is there any best practices available on how to work with such restrictions in high throughput scenarios?