Skip to main content

​
Overview

Adist uses advanced block-based indexing to parse and organize your project files into semantic blocks. This enables precise document searching, contextual AI responses, and efficient codebase analysis.

​
Block-Based Indexing

The default indexing method splits documents into semantic blocks such as functions, classes, methods, headings, and paragraphs. Each block is indexed individually with metadata, allowing for:
  • More precise searching and better context understanding
  • Improved AI interactions with relevant code snippets
  • Hierarchical structure preservation
  • Parent-child block relationships
Block-based indexing replaced the legacy full-document indexing in recent versions for better performance and accuracy.

​
Indexing Your Project

​
Initial Indexing

When you initialize a new project, Adist automatically indexes all supported files: 
adist init my-project
During initialization, you’ll be prompted to enable LLM summarization: 
Would you like to enable LLM summarization? (y/N):
If you choose yes, Adist will:
  1. Generate summaries for each indexed file
  2. Create an overall project summary
  3. Store summaries for faster future queries

​
Reindexing

To update your project’s index after code changes: 
adist reindex
Add the --summarize flag to regenerate summaries: 
adist reindex --summarize
Use verbose output to see detailed indexing progress:
adist reindex --summarize --verbose

​
Reindexing All Projects

You can reindex all configured projects at once: 
adist reindex --all
Or with summarization: 
adist reindex --all --summarize

​
Supported File Types

Adist indexes a wide variety of file types by default: 
.js, .jsx, .ts, .tsx
.py, .go, .java, .cpp
.rb, .php, .rs, .swift

​
Include and Exclude Patterns

The indexer uses fast-glob patterns to find files: Default Include Patterns: 
**/*.{js,jsx,ts,tsx,md,markdown,json,yaml,yml,toml}
Default Exclude Patterns: 
**/node_modules/**
**/dist/**
**/build/**
**/.git/**
**/coverage/**
**/*.min.*
Custom include/exclude patterns can be configured when using the indexing API programmatically.

​
Index Structure

Each indexed document contains:
  • Path: Relative path from project root
  • Title: Document filename or first heading
  • Last Modified: Timestamp of last file modification
  • Size: File size in bytes
  • Blocks: Array of semantic blocks with:
    • id: Unique block identifier
    • type: Block type (function, class, heading, etc.)
    • content: Block content
    • startLine and endLine: Source location
    • parent: Parent block ID (if applicable)
    • children: Child block IDs
    • metadata: Type-specific metadata (name, signature, etc.)
    • summary: AI-generated summary (if enabled)

​
Performance Considerations

​
Indexing Speed

Adist uses a progress bar to show indexing status: 
█████████████████████ 100% | ETA: 0s | 143/143 files | src/utils/parser.ts

​
Summarization Costs

Enabling LLM summarization incurs API costs when using cloud providers (Anthropic, OpenAI). Use Ollama for free local summarization.
The tool displays:
  • File count and total size before summarizing
  • Estimated processing time
  • Actual costs after completion

​
Storage Location

Indexes are stored in the configuration directory: 
~/Library/Application Support/adist

​
Advanced Options

​
Verbose Mode

See detailed indexing information: 
adist reindex --verbose
This shows:
  • LLM service status
  • Individual file processing
  • Summary generation progress
  • Error details

​
Legacy Indexing

The previous full-document indexing method is still available: 
adist legacy-reindex
Legacy indexing is maintained for backward compatibility but block-based indexing is recommended for better results.

​
Next Steps

Search Your Code

Learn how to search indexed projects

AI Chat

Start chatting with your codebase