Indexing & Search

Codebase indexing, dependency graphs, and semantic code search

PuzldAI can index your entire codebase for intelligent context injection and semantic search.

Quick Start

# TUI
/index              # Open indexing panel
/index full         # Full index with embeddings
/index quick        # Quick index without embeddings
/index search "auth" # Search indexed code

TUI Commands

Command	Description
/index	Open interactive indexing panel
/index full	Full index with AST parsing and embeddings
/index quick	Quick index without embeddings
/index search <query>	Search indexed code semantically
/index context <task>	Get relevant code for a task
/index config	Show detected project configuration
/index graph	Show dependency graph

What Gets Indexed

Data	Description
AST Structure	Functions, classes, interfaces, exports
Dependencies	Import/export relationships between files
Project Config	AGENTS.md, package.json, tsconfig.json
Embeddings	Semantic vectors for code search

Supported Languages

Language	Extensions
TypeScript	.ts, .tsx
JavaScript	.js, .jsx
Python	.py
Go	.go
Rust	.rs
Java	.java
C/C++	.c, .cpp, .h

Dependency Graph

The dependency graph tracks import/export relationships:

src/index.ts
├── imports: ./utils, ./config
└── exports: main, App

src/utils.ts
├── imports: lodash
└── exports: formatDate, parseJSON

Use this to understand how files connect and what might be affected by changes.

Semantic Search

Search your codebase using natural language:

/index search "authentication logic"
/index search "database connection"
/index search "error handling"

Results are ranked by semantic similarity, not just keyword matching.

Context Injection

When you ask questions or request changes, PuzldAI automatically:

1. Searches relevant code from the index
2. Injects matching files/functions as context
3. Gives agents better understanding of your codebase

Storage

Index embeddings are stored in the vector store (part of the memory system):

~/.puzldai/
└── puzldai.db       # SQLite database with vector embeddings

Indexing runs on-demand. Use /index full for complete re-indexing with embeddings, or /index quick for faster AST-only parsing.