Skip to main content
Cheetah AI provides powerful search capabilities that help you find code, understand your codebase, and research solutions. Unlike simple text search, the AI understands the meaning of code and can find relevant results even when they don’t contain your exact search terms. Search is fundamental to how the AI operates. When you ask questions or request changes, the AI automatically searches your codebase to gather relevant context. You can also explicitly search to explore your project or find specific code. Semantic search finds code by meaning and intent, not just keywords. This is the most powerful way to explore your codebase because it understands what code does, not just what it’s called.
This search finds authentication logic even if it’s in files called session.ts, login.ts, or middleware.ts. The AI understands that these are all related to authentication.
How semantic search works: The AI analyzes your codebase to understand what each piece of code does. When you search, it compares your query against this understanding to find relevant matches. This means you can describe what you’re looking for in natural language without knowing the exact function names or file locations. When to use semantic search:
  • Exploring an unfamiliar codebase
  • Finding code when you don’t know the exact terminology
  • Understanding how a feature is implemented across multiple files
  • Locating code related to a concept or functionality

​
Content Search

Content search finds exact text matches in your codebase. Use this when you know the specific text you’re looking for.
Content search is literal - it finds exactly what you type. This is useful for finding specific strings, variable names, or patterns that you know exist in your code. For complex patterns, you can use regular expressions:
Regex search gives you precise control over pattern matching. It’s useful for finding variations of names, specific code patterns, or text that follows a particular format. Find files by their name or location:
Path search is useful when you know roughly where something is or what it’s called, but not the exact location. The AI can search the web to find documentation, solutions, and best practices. This is useful for troubleshooting errors, researching libraries, and learning new techniques.
What web search is good for:
  • Troubleshooting error messages
  • Finding documentation for libraries and frameworks
  • Comparing different approaches or libraries
  • Learning best practices and patterns
  • Researching new technologies
How the AI uses web search: When you ask a question that requires current information or external knowledge, the AI searches the web, reads relevant pages, and synthesizes the information into a response. It cites sources so you can verify the information or read more.

​
Documentation Lookup

The AI can access official documentation for libraries and frameworks:
Documentation lookup fetches current, accurate information from official sources. This is more reliable than the AI’s training data, which may be outdated for rapidly evolving libraries. Supported sources:
  • Official library documentation
  • Framework guides and tutorials
  • API references
  • Package documentation from npm, PyPI, and other registries

​
Search Within Files

Find specific content within a single file:
This is useful when you’re working in a large file and need to locate specific code quickly.

​
Directory Tree

Visualize and understand your project structure:
Understanding your project’s structure helps you navigate effectively and understand how code is organized. Different search types are suited for different needs:

​
Search Strategies

​
Start Broad, Then Narrow

Begin with semantic search to find the general area of code, then use content search for specifics:
This approach helps you locate relevant code even when you don’t know exactly what you’re looking for.

​
Use Natural Language

Describe what you’re looking for in plain English. The AI understands context and intent:
This is more effective than trying to guess specific function names or file locations.

​
Combine Search Types

Chain different searches for comprehensive results:
The AI uses multiple search strategies to gather complete information.

​
Search Before Editing

In unfamiliar codebases, always search first to understand existing patterns:
This helps you follow established conventions when making changes.

​
Use Web Search for Errors

When encountering unfamiliar errors, web search often finds solutions faster than trying to debug from first principles:
The AI finds relevant Stack Overflow answers, GitHub issues, and documentation that explain the error and how to fix it.

​
Search in Different Modes

Search capabilities vary by mode: Normal Mode - No search capabilities. The AI only sees open files and conversation context. Gather Mode - Full search capabilities including semantic search, content search, path search, and web search. Read-only access to files. Plan Mode - Same search capabilities as Gather mode, plus the ability to use search results to inform planning. Agent Mode - Full search capabilities plus the ability to act on search results by editing files, running commands, etc. For exploration and research, Gather mode provides all the search capabilities you need without the risk of unintended changes. For tasks that require both searching and acting on results, Agent mode combines search with full editing capabilities.