Codebase Indexing

Codebase Indexing enables semantic code search across your entire project via AI embeddings. Instead of searching for exact text matches, it understands the meaning of your query, helping VJSP locate relevant code even when you don't know specific function names or file locations.

Functionality

Once enabled, the indexing system:

1. Parses your code with Tree-sitter to identify semantic blocks (functions, classes, methods)

2. Generates embeddings for each code block using an AI model

3. Stores vectors in a Qdrant database for fast similarity search

4. Provides VJSP with the codebase_search tool for intelligent code discovery

This allows natural language queries (e.g., "user authentication logic" or "database connection handling") to find relevant code across your entire project.

Key Benefits

• Semantic Search: Locate code by meaning, not just keywords

• Enhanced AI Understanding: VJSP can better comprehend and utilize your codebase

• Cross-project Discovery: Search all files, not just open ones

• Pattern Recognition: Identify similar implementations and code patterns

Qdrant Vector Database Configuration

Qdrant is the core tool for embedding vector storage and retrieval, supporting two configuration methods as follows:

Method 1: Local Vector Database

• Default Address: http://localhost:6333

• Authentication: API key configuration is supported and optional for secure deployment

Method 2: Apply for a Dedicated Vector Database via VJSP Official Website

• Dedicated Address: Obtain the exclusive URL through the VJSP Official Website Console - API Key

• Authentication: A dedicated API key applied via the official website is required (mandatory field)

ⓘ API Key Application Instructions

The Qdrant exclusive URL and API key must be applied for through the VJSP Official Website Console. After successful application, copy the relevant information to the corresponding configuration items of Codebase Indexing. For detailed acquisition steps, refer to the corresponding help documentation.

Configuration

1. Open the Index Ready icon in the bottom right corner of the VJSP dialog

2. Toggle the switch to enable "Enable Codebase Indexing"

3. Configure your embedding provider

4. Set the Qdrant URL and optional API key

5. Configure Maximum Search Results (Default: 50, Range: 1-100)

6. Click Save to start the initial indexing

Enable/Disable Toggle

The Codebase Indexing feature includes a convenient toggle switch that allows you to:

• Enable: Start indexing your codebase and activate the search tool

• Disable: Stop indexing, pause file monitoring, and disable the search function

• Retain Settings: Your configurations are saved when turned off

This toggle is useful for temporarily disabling indexing during intensive development work or when handling sensitive codebases.

Understanding Index Status

The interface displays real-time status with color indicators:

• Standby (Gray): Not running, waiting for configuration

• Indexing (Yellow): Currently processing files

• Indexed (Green): Up-to-date and ready for search

• Error (Red): Failed status requiring attention

File Processing Methods

Intelligent Code Parsing

• Tree-sitter Integration: Identify semantic code blocks using AST parsing

• Language Support: All languages supported by Tree-sitter

• Markdown Support: Full support for markdown files and documentation

• Fallback: Line-based chunking for unsupported file types

• Chunk Size:

  • Minimum: 100 characters
  • Maximum: 1,000 characters
  • Intelligent splitting for large functions

Automatic File Filtering

The indexer automatically excludes:

• Binary files and images

• Large files (>1MB)

• Git repositories (.git folder)

• Dependencies (node_modules, vendor, etc.)

• Files matching .gitignore and .vjsp patterns

Incremental Updates

• File Monitoring: Monitor changes in the workspace

• Intelligent Updates: Only reprocess modified files

• Hash-based Caching: Avoid reprocessing unchanged content

• Branch Switching: Automatically handle Git branch changes

Current Limitations

• File Size: Maximum 1MB per file

• Single Workspace: One workspace at a time

• Dependencies: Requires external services (embedding provider + Qdrant)

• Language Coverage: Optimal parsing limited to Tree-sitter supported languages

Using the Search Function

After indexing, VJSP can use the codebase_search tool to find relevant code:

Example Queries:

• "How is user authentication handled?"

• "Database connection settings"

• "Error handling patterns"

• "API endpoint definitions"

The tool provides VJSP with:

• Relevant code snippets (up to the configured maximum result limit)

• File paths and line numbers

• Similarity scores

• Context information

Search Result Configuration

You can control the number of returned search results by adjusting the Maximum Search Results setting:

• Default: 50 results

• Range: 10-200 results

• Performance: Lower values improve response speed

• Comprehensiveness: Higher values provide more context but may slow down responses

Privacy & Security

• Code Remains Local: Only small code snippets are sent for embedding generation

• Embeddings Are Numeric: Not human-readable representations

• Secure Storage: API keys are encrypted in VS Code and IDEA storage

• Access Control: Respects existing file permissions