Gemini CLI 2.0: 7 Game-Changing Updates for AI-Powered Development

Key Concepts

• Gemini CLI: A command-line interface for interacting with Google's Gemini AI models.
• Add-ons: Integrations with editors like Cursor and Visual Studio Code.
• Context Problem: The challenge of providing AI models with sufficient information about a project within token limits.
• Gemini Markdown File: A file generated by Gemini CLI containing a high-level overview of a project's structure and functionality.
• Memories: Manually managed instructions that are applied to all agents in the future.
• Stats: Detailed information about model and tool usage within Gemini CLI.
• Directories/Workspaces: The ability to work with multiple project directories in a single Gemini CLI session.
• Conversation History: Saving and managing previous conversations within Gemini CLI.
• GitHub Actions: Integration with GitHub for automated PR reviews and issue triaging.
• Rate Limits: Restrictions on the number of requests that can be made to the Gemini API within a given timeframe.
• Agentic Workflows: Automated processes driven by AI agents.

1. Editor Integration with Gemini CLI Companion

• Main Point: Gemini CLI now integrates with editors like Cursor and Visual Studio Code via an add-on, allowing it to understand and interact with open files.
• Specific Details:
  • The add-on is called "Gemini CLI companion."
  • It can be installed from the editor's marketplace.
  • Commands: IDE install, IDE status
  • Gemini CLI can automatically understand the context of the current file.
  • Users can highlight code sections and ask questions about them.
• Example: Highlighting a code section and asking "what is going on in this section?" allows Gemini CLI to provide a description of the highlighted code.
• Process:
  1. Install the Gemini CLI companion add-on from the editor's marketplace.
  2. Type IDE install in the Gemini CLI.
  3. Restart the editor.
  4. Type IDE status to confirm the connection.

2. Gemini Markdown File Generation

• Main Point: The omit command analyzes a project and generates a Gemini markdown file, which provides context for future agent calls.
• Specific Details:
  • Addresses the context problem by providing a summary of the codebase.
  • Includes core layouts of the project.
  • The generated file can be further refined and expanded upon.
• Example: Running the omit command on a Next.js project generates a markdown file describing the project's structure and technologies used.
• Argument: The Gemini markdown file is a good starting point, but it should be iteratively improved to provide more detailed context.
• Data: Gemini CLI offers a generous free tier with up to 1,000 AI requests per day (Gemini 2.5 Pro).

3. Manual Memory Management

• Main Point: Users can now manually manage memories (rules) that are applied to all future agents.
• Specific Details:
  • Commands: memory show, memory add
  • memory show lists existing memories.
  • memory add allows adding new rules as text.
  • Memories are stored in a Gemini folder in the user's root directory.
  • Rules apply to all future projects.
• Example: Adding a rule to ensure all text in the UI is capitalized.
• Process:
  1. Type memory show to view existing memories.
  2. Type memory add followed by the rule text to add a new memory.
  3. Use phrases like "important rule" or "you must do this" for better results.
• Caveat: Rules are applied globally to all projects.

4. Detailed Stats Viewing

• Main Point: Gemini CLI now provides more detailed stats on model and tool usage.
• Specific Details:
  • Users can view model and tool calls, token usage, and caching information.
  • The tools command lists available tools (e.g., read files, write file, search text).
• Wishlist: The speaker wishes for the ability to view stats for specific sessions and overall usage to track quota consumption.

5. Multi-Directory/Workspace Support

• Main Point: Gemini CLI can now work with multiple directories and workspaces in a single session.
• Specific Details:
  • Commands: directory, directory add
  • Allows bringing code from one project into another.
• Example: Bringing a stripe.ts file from one project to another, along with its dependencies.
• Argument: This feature leverages the power of AI to understand and merge code from different projects.

6. Conversation History Management

• Main Point: Users can now save and manage their conversation history.
• Specific Details:
  • Commands: chat list, chat save, chat resume
  • chat list shows saved conversations.
  • chat save saves the current conversation.
  • chat resume resumes a saved conversation.
• Wishlist: The speaker wishes for automatic saving of all conversations.

7. GitHub Integration

• Main Point: Gemini CLI integrates with GitHub for automated PR reviews and issue triaging via GitHub Actions.
• Specific Details:
  • Command: setup GitHub
  • Requires setting the Gemini API key as a repository secret in GitHub.
  • Adds workflows to the project for PR reviews and issue handling.
  • Allows commenting on code changes with Gemini CLI to trigger a review.
• Issue: The PR review functionality is currently not working reliably (as of the video's recording).
• Process:
  1. Type setup GitHub in Gemini CLI.
  2. Add the Gemini API key as a repository secret in GitHub.
  3. The command updates the project's GitHub workflows.
  4. Comment on code changes with "Gemini CLI" to trigger a review.

8. Overall Assessment and Wishlist

• Target Audience: Gemini CLI is a good beginner tool for exploring agentic workflows with a generous free tier.
• Current Limitations:
  • Lack of real-time quota and rate limit information.
  • Inconsistent performance and long processing times.
  • Manual tasks that could be automated.
• Wishlist Items:
  • Real-time quota and rate limit display.
  • Improved performance and faster processing.
  • More agentic features for automation.
  • More detailed stat messages.

9. Shipkit.ai Promotion

• Main Point: A project and course called Shipkit.ai is promoted, which aims to help users build real-world AI applications quickly.
• Specific Details:
  • Includes pre-built templates for common AI applications (chat, RAG, multi-agent).
  • Provides a course on building AI applications from idea to production.
  • Includes AI prompts and real-world AI projects.

10. Conclusion

Gemini CLI is a promising tool for interacting with AI models and automating development tasks. The recent updates, including editor integration, markdown file generation, and GitHub integration, significantly enhance its capabilities. While some limitations and bugs exist, the tool's generous free tier and active development make it a worthwhile option for developers interested in exploring agentic workflows. The speaker expresses excitement about the tool's direction and provides valuable feedback for future improvements.