Skip to content

docs

Directory actions

More options

Directory actions

More options

Latest commit

 

History

History

docs

README.md

RLM-REPL Documentation

Welcome to the RLM-REPL documentation! This directory contains comprehensive guides for using the library.

Documentation Index

Getting Started

  • Getting Started Guide - Installation, setup, and your first query
    • Installation instructions
    • Model setup (Ollama, OpenAI, etc.)
    • First examples
    • Common issues

API Reference

  • API Reference - Complete API documentation
    • Core classes (RLMREPL, RLMConfig, DatabaseConfig)
    • All methods and parameters
    • Data classes and events
    • Complete examples

Configuration

  • Configuration Guide - All configuration options
    • RLMConfig parameters
    • DatabaseConfig options
    • Environment variables
    • Best practices
    • Configuration examples

Examples

  • Examples - Comprehensive usage examples
    • Basic usage
    • Streaming events
    • Persistent database
    • Building applications
    • Advanced queries
    • Error handling

Architecture

  • Architecture - How the system works
    • System overview
    • Core components
    • Reading process
    • Event system
    • Performance considerations

Troubleshooting

  • Troubleshooting Guide - Common issues and solutions
    • Installation issues
    • Configuration problems
    • API connection issues
    • Query problems
    • Performance issues

Quick Links

Documentation Structure

docs/
├── README.md              # This file
├── getting-started.md     # Installation and first steps
├── api-reference.md       # Complete API documentation
├── configuration.md       # Configuration options
├── examples.md            # Usage examples
├── architecture.md        # System architecture
└── troubleshooting.md     # Common issues and solutions

Getting Help

  1. Start here: Getting Started Guide
  2. Need examples?: Examples
  3. API questions?: API Reference
  4. Having issues?: Troubleshooting Guide

About

Author: Remy Gakwaya

Background: RLM-REPL was created after reading the MIT paper on Recursive Language Models. The initial approach used a REPL where LLMs would generate Python functions, but this proved challenging with smaller models.

After hundreds of iterations, Remy developed the RLM-REPL v8 concept - a human-like reading strategy optimized for local, smaller language models. The philosophy: if it works with poor and small models on limited computation, it will excel with leading LLMs.

The library evolved to use SQL-based retrieval with DuckDB, implementing the proven v8 reading strategy (overview → search → deep read → synthesize) in a more reliable way that works with models of all sizes.

Contributing

Found an error or want to improve the documentation? Contributions are welcome!

  1. Check existing issues
  2. Open a new issue or pull request
  3. Follow the existing documentation style