This is a Visual Studio Code (VS Code) extension which allows developers to easily explore and experiment with large language models (LLMs) available in Amazon Bedrock.
This is an open source project and we would love for you to be involved. To contribute, please refer to CONTRIBUTING.md.
You can install this extension from the VS Code Extension Marketplace
This extension uses your local AWS credentials to invoke the Amazon Bedrock service. By default, the extension will use the default AWS profile. If you want to change the profile, you can update the bedrockPlayground.profileName setting in your User/Workspace settings.
The extension will also use the us-west-2 region by default. To change the region, you can update the bedrockPlayground.region setting.
⚠️ By using this tool, you understand that it uses Amazon Bedrock's InvokeModel API, which will incur AWS charges. The charges you incur from using Amazon Bedrock will be your responsibility. Please review the latest Amazon Bedrock pricing here before using this tool.
The extension currently has 2 features:
The playground environment provides a user interface for exploring and experimenting with LLMs in Amazon Bedrock. To open the playground environment, use the Command Palette and run the command Bedrock Playground: Open Playground.
You can adjust the inference parameters for each model by configuring your User/Workspace Settings.
You can use the Bedrock Playground: Generate command to use a LLM from Amazon Bedrock to generate text. You can configure which model to use by updating the setting bedrockPlayground.generate.modelId.
To generate text, just run the command Bedrock Playground: Generate using the Command Palette. The output will be placed in a new Markdown file.
You can also provide context from an active text editor when using the Bedrock Playground: Generate command. To provide context, simply select the code or text in your active text editor, run the command using the Command Palette, and input your request.
The Bedrock Playground: Generate command uses prompt templates to customize the text it generates. To get the best results from your model, you may want to update the prompt templates. The default templates are designed for Anthropic's Claude model. If you are using a different model, tweak the templates in the bedrockPlayground.generate.promptTemplates based on the model provider's guidance on prompting.
To customize the template used when there is no context, you should update the generate template. This template string must contain a {REQUEST} placeholder which will be substituted with the user's request. For example:
"{REQUEST}\nPlease place your response in <response></response> XML tags."
To customize the prompt template that is used when context is provided, you should update the generateWithContext template. This template must contain both a {REQUEST} placeholder as well as a {CONTEXT} placeholder. For example:
"Use the context wrapped in <context></context> tags to respond to a user's request.\nThe user's request will be wrapped in <request></request> tags.\n<context>{CONTEXT}</context>\n<request>{REQUEST}</request>\nPlease place your response in <response></response> tags."
