Bedrock Provider Example
This example demonstrates how to use the Bedrock LLM with different model providers, including support for Nova models and inference profiles.
Features
- Automatic provider detection from model ID
- Explicit provider specification for edge cases
- Support for Nova models (e.g.,
amazon.nova-lite-v1:0)
- Support for inference profiles (e.g.,
us.amazon.nova-lite-v1:0)
Prerequisites
- AWS credentials configured (via environment variables or AWS credentials file)
- Access to the Bedrock models you want to use
Usage
# Using default Titan model
go run main.go
# Using Nova model
go run main.go -model "amazon.nova-lite-v1:0"
# Using inference profile
go run main.go -model "us.amazon.nova-lite-v1:0"
# Using Anthropic model with explicit provider (for edge cases)
go run main.go -model "us.anthropic.claude-3-7-sonnet-20250219-v1:0" -provider "anthropic"
# Custom prompt
go run main.go -prompt "What is the capital of France?"
# Verbose output
go run main.go -verbose
Environment Variables
Set these environment variables before running:
export AWS_ACCESS_KEY_ID=your_key_id
export AWS_SECRET_ACCESS_KEY=your_secret_key
export AWS_REGION=us-east-1
Supported Providers
The Bedrock integration automatically detects the provider from the model ID:
- Nova: Models containing
.nova- (e.g., amazon.nova-lite-v1:0, us.amazon.nova-pro-v1:0)
- Anthropic: Models containing
anthropic (e.g., anthropic.claude-3-sonnet-20240229-v1:0)
- Amazon: Models containing
amazon (excluding Nova)
- Meta: Models containing
meta (e.g., meta.llama3-1-405b-instruct-v1:0)
- Cohere: Models containing
cohere (e.g., cohere.command-r-plus-v1:0)
- AI21: Models containing
ai21 (e.g., ai21.jamba-1-5-large-v1:0)
Model Provider Option
For cases where automatic detection doesn't work correctly (e.g., custom inference endpoints), you can explicitly specify the provider:
llm, err := bedrock.New(
bedrock.WithModel("custom.endpoint.model-id"),
bedrock.WithModelProvider("anthropic"), // Explicitly set provider
)