# piragi playground & codelabs

> Interactive zero-config RAG (Retrieval-Augmented Generation) playground and codelabs for Python.

## Quick Links
- **Interactive Playground**: https://hemanth.github.io/piragi/
- **PyPI Package**: https://pypi.org/project/piragi/
- **GitHub Repository**: https://github.com/hemanth/piragi
- **Architecture Overview**: https://hemanth.github.io/piragi/docs/architecture/architecture.html

## Key Features
- **Zero Configuration**: Built-in embeddings and serverless vector store (LanceDB) by default.
- **Inference Engines**: On-device Gemini Nano (via Chrome AI `window.ai` / Prompt API), Ollama, OpenRouter, and OpenAI-compatible endpoints.
- **Composable Pipelines**: Modular `IngestionPipeline` (streaming, chunking, embedding, storage) and `RetrievalPipeline` (query transformation, parallel search, reranking, answering).
- **Format Support**: Markdown, PDF, Word, Excel, source code, text files, and remote blob storage (`s3://`, `gs://`, `az://`, web crawling).
- **Concurrency & Reliability**: Thread-safe `FilteredRagi` immutable querying, model cache locking, and exponential backoff retry.

## Installation
```bash
pip install piragi
```

## CLI Usage
Launch local playground with live Python kernel:
```bash
piragi playground
```