Trigger StarShip CodeReviewer

Default Trigger

To manually initiate the StarShip CodeReviewer, you can enhance your Merge Request (MR) by adding a specific note. Each command performs a unique function that aids in automating aspects of code review. Below is a detailed guide on how to utilize these commands:

• Triggering Code Reviewing Process:

  Use the command @codegpt /review in your note. This triggers an automated code review process, analyzing your code for quality standards, potential bugs, and adherence to coding best practices.

• Initiating Code Description Process:

  To have a detailed description or summary of your code, append the note with @codegpt /describe. This command prompts an automated process that provides insights into what your code does, potentially improving understandability for reviewers who may not be intimately familiar with the codebase.

• Starting Code Evaluation Process:

  By adding @codegpt /evaluate to your note, you can trigger a comprehensive evaluation of your code. This includes performance analysis, security vulnerability checks, and suggestions for optimization, ensuring your code is robust and efficient.

• Comprehensive Code Analysis:

  If you wish to carry out a full spectrum analysis involving reviewing, describing, and evaluating your code, simply use @codegpt /all in your note. This command consolidates the functionalities of the previous commands, providing a thorough inspection and assessment of your code in one go.

• Accessing Help Information:

  For assistance or to get more information about the available commands and how to use them, include @codegpt /help in your note. This will display a help guide, detailing each command and its function, ensuring you make the most out of the StarShip CodeReviewer.

Implementing these commands in your MRs can significantly streamline the code review process, enhance code quality, and foster a culture of continuous improvement within your development team.

Using Custom Configured LLMs

The StarShip CodeReview platform is designed to provide a versatile and comprehensive code review experience, supporting a wide range of development projects by offering integration with both SaaS (Software as a Service) Large Language Model (LLM) services, such as GPT-4, and custom on-premise LLM services, for example, deepseek-code-33B. This enhanced support ensures that teams can choose the most suitable LLM service for their specific code review needs, whether they prioritize cutting-edge AI capabilities, custom solutions tailored to their specific coding standards, or data privacy considerations.

If you've set up a custom LLM, you can activate this specific review process within your Merge Requests (MRs) by appending a note in the format @codegpt-custom ..., followed by the specific command you wish to execute (e.g.,/review, /describe, etc.).

For example, if you have configured a custom local LLM that is better tuned to your project's specific requirements or coding standards, simply trigger it by using:

@codegpt-custom /review

This command would initiate a code review process utilizing your custom configured LLM, providing insights and suggestions tailored to your unique setup.

Switching Between Default and Custom LLMs

If you have set a custom local LLM as your default model but occasionally need to leverage the advanced capabilities of another model, such as the GPT-4 assistant API, you can easily switch for specific tasks. To do this within an MR, append a note with @codegpt-gpt4 ..., followed by the desired command. This enables the StarShip CodeReviewer to temporarily utilize the GPT-4 model for the stated purpose in your note.

For instance, if you wish to conduct a code review using the GPT-4 model’s advanced analytical capabilities, you would input:

@codegpt-gpt4 /review

This explicitly directs the StarShip CodeReviewer to process your request using the GPT-4 API, ensuring you benefit from the cutting-edge analysis and feedback GPT-4 offers, even if it's not your configured default.

Limitation: some messages (such as help information) may also show codegpt but that works with codegpt-custom.