# Stephen v1.0.1 - Table of Contents

> ColBERT-style neural retrieval for Elixir

## Pages

- [Stephen](readme.md)

- Livebooks
  - [Getting Started with Stephen](getting_started.md)

- Guides
  - [Architecture](architecture.md)
  - [Index Types](index_types.md)
  - [Compression](compression.md)
  - [Chunking](chunking.md)
  - [Configuration](configuration.md)

## Modules

- Core
  - [Stephen](Stephen.md): ColBERT-style neural retrieval for Elixir.
  - [Stephen.Encoder](Stephen.Encoder.md): Encodes text into per-token embeddings using BERT.
  - [Stephen.Retriever](Stephen.Retriever.md): High-level retrieval operations combining encoding, indexing, and scoring.
  - [Stephen.Scorer](Stephen.Scorer.md): Implements ColBERT's late interaction scoring mechanism (MaxSim).

- Indexes
  - [Stephen.Index](Stephen.Index.md): Manages the ColBERT document index.
  - [Stephen.Index.Compressed](Stephen.Index.Compressed.md): Compressed document index using residual compression.
  - [Stephen.Plaid](Stephen.Plaid.md): PLAID-style indexing for efficient ColBERT retrieval.

- Utilities
  - [Stephen.Chunker](Stephen.Chunker.md): Passage chunking for long documents.
  - [Stephen.Compression](Stephen.Compression.md): ColBERTv2-style residual compression for token embeddings.
  - [Stephen.KMeans](Stephen.KMeans.md): K-means clustering wrapper around Scholar.Cluster.KMeans.