LlamaIndex Playground

A playground repository for experimenting with LlamaIndex, a powerful framework for building context-augmented LLM applications.

Overview

This repository contains various examples and experiments with LlamaIndex features:

• Semantic Search (semantic_search/) - Vector store indexing for efficient similarity search
• Summarization (summarization/) - Document summarization using SummaryIndex
• Structured Data Queries (query_structured_data/) - Natural language queries to SQL databases
• Context-Augmented Agents (context_augmentation/) - Query engines with context retrieval over financial documents (Uber 10Q filings)
• RAG Agents (rag_agent/) - Retrieval-augmented generation with custom function tools
• Notebooks (notebook/) - Jupyter notebooks with LATS and DAG agent implementations

Features

• Vector store indexing and retrieval
• Document summarization with tree-summarize mode
• SQL database querying via natural language
• Multi-document context augmentation
• Custom tool-augmented agents
• Async agent streaming support

Setup

1. Clone the repository
2. Install dependencies with Poetry:

   poetry install

3. Copy .env.example to .env and add your OpenAI API key:

   cp .env.example .env
   # Edit .env and add: OPENAI_API_KEY=your_key_here

Usage

Each module can be run independently. For example:

# Semantic search
python -m semantic_search

# Summarization
python -m summarization

# Query structured data
python -m query_structured_data

# Context-augmented agent
python -m context_augmentation

Key Findings

• Semantic search is fast and efficient - pre-indexing and storing vectors enables faster retrieval and cheaper operations
• Summarization requires more processing time due to the tree-summarize approach

Dependencies

• Python 3.12+
• LlamaIndex 0.10.39+
• llama-index-agent-openai
• llama-index-llms-openai
• python-dotenv

License

MIT