Is your feature request related to a problem? Please describe.
For users who send the LLM request events to a specific model, It is expected that the auto-instrumented trace spans should provide fine granularity of LLM metadata including Token Spent, Model Name/Provider, Prompt Completion Reason, etc, by adhering to LLM span semantic convention submitted by LLM WG.
Describe the solution you'd like
I can contribute and start with AWS Bedrock by updating aws-sdk-2.2 to add BedrockRuntime as a new supported type in AwsSdkRequestType. And then mondify AwsSdkExperimentalAttributesExtractor to capture LLM specific attribute from Bedrock API request and response for the following attributes.
Model Id/Name
Prompt Token
Output Token
Model Provider
Completion Reason
Top_p
Top_K
Temperature
Max_tokens
Bedrock supports multiple LLM models, will start with the following
Is your feature request related to a problem? Please describe.
For users who send the LLM request events to a specific model, It is expected that the auto-instrumented trace spans should provide fine granularity of LLM metadata including Token Spent, Model Name/Provider, Prompt Completion Reason, etc, by adhering to LLM span semantic convention submitted by LLM WG.
Describe the solution you'd like
I can contribute and start with AWS Bedrock by updating aws-sdk-2.2 to add
BedrockRuntime
as a new supported type in AwsSdkRequestType. And then mondify AwsSdkExperimentalAttributesExtractor to capture LLM specific attribute from Bedrock API request and response for the following attributes.Bedrock supports multiple LLM models, will start with the following
Describe alternatives you've considered
No response
Additional context
No response