Experimental features - Amazon Q Developer

Experimental features

Amazon Q Developer CLI includes experimental features that provide advanced functionality for enhanced productivity. These features are in active development and must be explicitly enabled before use.

Important

Experimental features may change or be removed at any time. Use at your own discretion in production workflows.

Managing experimental features

Use the /experiment command to view and toggle experimental features:

/experiment

This displays an interactive menu where you can:

  • View the current status of each experiment (ON/OFF)

  • Toggle experiments by selecting them

  • View descriptions of what each experiment does

Knowledge management

The knowledge management feature provides persistent context storage and retrieval across chat sessions. Enable it with:

q settings chat.enableKnowledge true

Basic usage

Once enabled, use /knowledge commands within your chat session:

/knowledge add <name> <path>

Add files or directories to your knowledge base

/knowledge show

Display all entries in your knowledge base

/knowledge remove <identifier>

Remove entries by name, path, or context ID

/knowledge update <path>

Update existing knowledge base entry with new content

/knowledge clear

Remove all entries from your knowledge base (requires confirmation)

/knowledge status

View status of background indexing operations

Index types

Choose between two indexing approaches when adding content:

Fast (--index-type Fast)

Lexical search using BM25. Lightning-fast indexing and instant keyword-based search. Perfect for logs, configs, and large codebases.

Best (--index-type Best)

Semantic search using AI embeddings. Intelligent search that understands context and meaning. Perfect for documentation and research.

Example usage:

/knowledge add "project-docs" /path/to/docs --index-type Best
/knowledge add "log-files" /path/to/logs --index-type Fast

Agent-specific knowledge bases

Each agent maintains its own isolated knowledge base, ensuring that knowledge contexts are scoped to the specific agent you're working with. When you switch between agents, your knowledge commands automatically work with that agent's specific knowledge base.

Tangent mode

Tangent mode creates conversation checkpoints, allowing you to explore side topics without disrupting your main conversation flow. Enable it with:

q settings chat.enableTangentMode true

Using tangent mode

Once enabled, use /tangent or Ctrl+T to toggle tangent mode:

  1. Enter tangent mode: Creates a conversation checkpoint

    /tangent
    Created a conversation checkpoint (↯). Use ctrl + t or /tangent to restore the conversation later.
  2. In tangent mode: You'll see a yellow symbol in your prompt

    ↯ > What is the difference between async and sync functions?
  3. Exit tangent mode: Returns to your main conversation

    ↯ > /tangent
    Restored conversation from checkpoint (↯). - Returned to main conversation.

Best practices

Use tangent mode for:

  • Asking clarifying questions about the current topic

  • Exploring alternative approaches before deciding

  • Getting help with Q Developer CLI commands or features

  • Testing understanding of concepts

Avoid using tangent mode for completely unrelated topics or long, complex discussions.

Thinking mode

Thinking mode enables complex reasoning with step-by-step thought processes. When enabled, Amazon Q shows its reasoning process when working through complex problems or multi-step reasoning, helping you understand how conclusions are reached.

This feature is useful for debugging and learning, providing transparent decision-making processes for complex tasks.

Additional resources

For complete details about experimental features, including advanced configuration options and troubleshooting, see the supplementary Amazon Q Developer CLI documentation: