Skip to main content
Joining a new project can be challenging. Where are the main files? What is the architecture? How does the authentication flow work? This recipe shows how to use ChatCLI to drastically accelerate your onboarding process.

​
The Problem

You cloned a repository with hundreds of files and need to become productive as quickly as possible. Reading everything is impractical, and the documentation may be outdated.

​
Ingredients

  • A code repository cloned to your local machine.

​
Step by Step

This workflow uses a “top-down” approach: first we get an architecture overview and then we dive into specific details.
1

The Panoramic View with @file --mode=summary

The first step is to get a map of the territory. The summary mode is perfect for this, as it lists the file and directory structure without sending the code content.
@file --mode=summary . > Describe the overall architecture of this project based on the file structure. Which components appear to be the most important?
What to expect from the AI:The AI will return a high-level analysis, such as:
This appears to be a Go backend project.
The /cmd directory suggests there are multiple binaries.

The /internal/api folder likely contains the HTTP route handlers,
while /internal/database manages the database access logic.

The docker-compose.yml file indicates the project uses Docker for
the development environment.
With this, you already have a mental map of where to look for things.
2

Specific Investigation with @file --mode=smart

Now that you have an overview, you can ask specific questions. The smart mode is ideal here, because you don’t need to know which files contain the answer. The AI will find them for you.Let’s investigate the authentication flow.
@file --mode=smart ./internal/ > Explain the end-to-end authentication flow, from the API route to token validation.
How it works:
  1. ChatCLI sends the list of all files inside ./internal/ to the AI.
  2. The AI, based on the question “authentication flow,” identifies that files like auth_handler.go, jwt_service.go, and user_repository.go are likely the most relevant.
  3. ChatCLI then sends the content of only those selected files.
  4. The AI answers your question based on the actual code from the most important files.
3

Deep Dive into a File with @file --mode=full

During your investigation, you may want to understand a specific file in detail. Now, use the full mode (or simply omit the mode, as it is the default).
@file ./internal/api/middleware/auth.go > Refactor this middleware to follow Go best practices, adding more robust error handling and logging.
The AI now has the complete file content and can perform complex tasks such as refactoring, documentation, or test generation.
Result: In less than 15 minutes, you went from zero knowledge of the project to:
  1. Understanding the high-level architecture.
  2. Mapping a critical business flow (authentication).
  3. Analyzing and even refactoring a specific code component.
This workflow transforms hours or days of manual exploration into a focused and efficient question-and-answer session.