This is a small command-line utility written in Go to make API requests to the AWS Bedrock chat API for Claude.
The tool takes an optional model as input and gives a Charm.sh's huh based interface which lets you query Claude.
On exit it will show tokens consumed and timing metrics.
To use the CLI tool, you need to have the your AWS environment variables setup as usual (e.g. a working AWS_PROFILE) that allows for bedrock:InvokeModel permission.
Also you'll need to ask AWS to allow you to use these models, basically just go to AWS Bedrock and request access to all their models.
It's free (getting access) and usually doesn't take very long (minutes/hours rather than days).
$ export AWS_PROFILE=bedrock
$ ./chat
The --model flag can be used to specify a different model (default is anthropic.claude-v2, try anthropic.claude-3-sonnet-20240229-v1:0 for more expensive chat, see ):
$ ./chat --model "anthropic.claude-3-sonnet-20240229-v1:0"
To build the CLI tool, you can either use your local golang installation (needs 1.20 or higher), or have Docker build it.
Run the following command to build the Docker image:
$ docker build -t chat:latest .
After the build is complete, you can run the CLI tool using the Docker image:
$ docker run --rm -e AWS_PROFILE=<yourprofile> -v ~/.aws:/root/.aws:ro chat:latest
Note that the AWS credentials need to be present in some form, if you have setup a local profile this will do. See https://docs.aws.amazon.com/cli/latest/userguide/cli-configure-files.html for details.
E.g. alternatively you can provide the AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY and AWS_DEFAULT_REGION variables or simply run it from e.g. EC2 with a proper instance profile.
Anyway, above can also be done using make:
make build-docker docker-run
Run the following command to build the chat binary:
$ make build-local
After building you can run it directly:
$ ./chat
This is just a pet project to tinker with Charm and Claude.