Your First Conversation¶

This guide walks you through your first interaction with Amplifier, helping you understand how to communicate effectively with the AI assistant and what to expect.

Starting a Session¶

Launch Amplifier from your terminal:

amplifier run

You'll see the Amplifier prompt appear:

User:

This means Amplifier is ready to receive your instructions. You're now in an interactive session where you can have a natural conversation with the AI assistant.

Your First Prompt¶

Start with something simple to get a feel for how Amplifier works. Here are some great first prompts:

Example 1: Ask About Your Project¶

What files are in the current directory?

Amplifier will use the glob or bash tool to explore your workspace and provide a structured summary of what it finds.

Example 2: Request Code Analysis¶

Read the main.py file and explain what it does

The assistant will use the read_file tool to access the file, then provide a detailed explanation of the code's functionality.

Example 3: Simple Code Generation¶

Create a Python function that calculates the factorial of a number

Amplifier will write the code and can save it to a file if you specify a location.

Example 4: Multi-Step Task¶

Find all TODO comments in my Python files and create a summary

This demonstrates Amplifier's ability to chain multiple operations: searching files, reading content, and synthesizing information.

Understanding Responses¶

Amplifier responses typically include several components:

1. Thinking Process (Sometimes Visible)¶

The assistant may show its reasoning before taking action. This helps you understand its approach to solving your request.

2. Tool Usage¶

You'll see when Amplifier uses tools to accomplish tasks:

[Using read_file: ./src/main.py]
[Using bash: pytest tests/]

These indicators show what actions are being taken on your behalf.

3. Results and Explanations¶

After using tools, Amplifier provides: - Summaries: Condensed information about what was found or done - Analysis: Interpretation of results - Code blocks: Formatted code with syntax highlighting - Recommendations: Suggestions for next steps

4. Error Handling¶

If something goes wrong, Amplifier will: - Explain what happened - Suggest fixes or alternatives - Often retry with a different approach automatically

Tool Usage¶

Amplifier has access to various tools that extend its capabilities. Understanding these helps you make better requests.

File Operations¶

Reading files:

Read the config.yaml file

Writing files:

Create a new file called utils.py with a helper function

Editing files:

Update the README.md to include installation instructions

Code Quality¶

Checking Python code:

Check this Python file for errors and style issues

Amplifier uses ruff for formatting/linting and pyright for type checking.

Search Operations¶

Finding files:

Find all JavaScript files in the src directory

Searching content:

Search for all functions named "calculate" in Python files

Command Execution¶

Running tests:

Run the test suite

Installing packages:

Install the requests library

Git operations:

Show me the git status

Complex Tasks with Sub-Agents¶

For complex, multi-step tasks, Amplifier can delegate work to specialized agents:

Review the code I just wrote and suggest improvements

This might spawn a code-reviewer agent that works autonomously.

Best Practices for Prompts¶

Be Specific¶

❌ Vague: "Fix the code" ✅ Specific: "Fix the syntax error in main.py on line 42"

Provide Context¶

❌ Minimal: "Add a test" ✅ Contextual: "Add a unit test for the calculate_total function that checks edge cases like negative numbers and zero"

Break Down Complex Requests¶

For very complex tasks, consider breaking them into steps:

1. First, analyze the current authentication system
2. Then suggest improvements for security
3. Finally, implement the changes

Or let Amplifier handle it all at once:

Analyze the auth system, suggest security improvements, and implement them

Iterate and Refine¶

Don't worry about getting it perfect the first time. You can refine:

Actually, make that function async instead

Add error handling to the code you just wrote

Common Workflows¶

Code Review Workflow¶

User: I just updated the payment processor. Can you review it?
Assistant: [reads file, analyzes code, provides feedback]
User: Make those changes
Assistant: [edits file with improvements]
User: Now add tests for the edge cases
Assistant: [creates test file]

Debugging Workflow¶

User: My tests are failing
Assistant: [runs tests, analyzes output]
User: What's causing the failure?
Assistant: [explains issue with specific line numbers]
User: How do I fix it?
Assistant: [provides solution and can implement it]

Learning Workflow¶

User: Explain how async/await works in Python
Assistant: [provides explanation with examples]
User: Show me a practical example
Assistant: [creates working code example]
User: What are common pitfalls?
Assistant: [explains gotchas and best practices]

Ending a Session¶

To exit Amplifier, use:

Ctrl+D (or type 'exit')

Your conversation history is saved, and any changes made to files are preserved in your workspace.

Pro Tips¶

Parallel Operations¶

Request multiple things at once:

Read both config.py and utils.py, then tell me how they interact

Asking for Alternatives¶

Show me three different ways to implement this feature

Context Awareness¶

Amplifier remembers your conversation history:

User: Create a User class
Assistant: [creates User class]
User: Now add a method to validate email addresses
Assistant: [knows which class you mean]

Skill Loading¶

For specialized domains:

Load the REST API design skill and help me design an endpoint

What Not to Expect¶

• No internet access by default: Amplifier works with local files (unless web-research agent is used)
• No memory between sessions: Each session starts fresh
• No real-time interactions: Can't run interactive programs directly
• No destructive operations without review: Safety guardrails prevent accidental damage

Next Steps¶

Now that you understand the basics, explore:

1. Key Commands - Essential commands and workflows
2. Your First Bundle - Create custom behavior packages
3. Concepts - Deeper understanding of how Amplifier works
4. Tools - Complete documentation of available tools

Try experimenting with different types of requests to discover what Amplifier can do for you!

Quick Reference Card¶

Ready to dive deeper? Continue to Key Commands to learn about Amplifier's capabilities in detail.