How does the budget control work in RLM?

Using the --max-budget flag, you can set a USD limit for a task; the skill will automatically stop execution and return results if the cost exceeds your specified limit.

Does RLM support searching through local files?

Yes, RLM includes index and search commands using high-performance tools like ripgrep and Tantivy for fast, concept-based codebase exploration.

Can RLM access the internet for research?

Yes, by using the --exa flag and an Exa API key, RLM can perform semantic web searches to bring external documentation and articles into its analysis.

What is a Recursive Language Model (RLM)?

RLM allows an LLM to call itself over decomposed parts of a larger input, effectively bypassing context window limits by processing data in recursive layers.

What happens if a recursive task takes too long?

You can use the --max-timeout or --max-iterations flags to limit execution time. If stopped early, RLM will still attempt to provide a partial answer based on completed steps.

RLM - Recursive Language Model CLI

Name: RLM - Recursive Language Model CLI
Author: rawwerks

byrawwerks

•

Herramientas para Desarrolladores

Enables Claude to process massive contexts through recursive input decomposition and self-calling capabilities.

The RLM (Recursive Language Models) skill empowers Claude with the ability to handle near-infinite context by recursively breaking down complex inputs into manageable parts. By leveraging specialized commands for asking, searching, and indexing, this tool allows for deep codebase analysis, comprehensive git diff reviews, and structured navigation of large documents. It features built-in budget controls, local search integration via ripgrep, and optional web search capabilities, making it an essential utility for developers and researchers working with information-dense environments that exceed standard model context limits.

Características Principales

0152 GitHub stars

02Recursive Processing: Decomposes large inputs into sub-tasks with configurable depth for exhaustive analysis.

03Smart Codebase Indexing: Provides powerful local search tools like ripgrep and Tantivy to locate relevant code segments quickly.

04Budget & Resource Management: Sets strict spending limits and token caps to prevent unexpected API costs during recursive runs.

05Web & PDF Integration: Optional support for Exa web search and hierarchical PageIndex PDF navigation for external research.

06Multi-Source Input Support: Processes directories, files, URLs, and stdin seamlessly for versatile data ingestion.

Casos de Uso

01Deep Git Diff Reviews: Summarizing and bug-checking extensive changesets that would normally overflow a standard context window.

02Recursive Knowledge Extraction: Iteratively searching through local files and web documentation to solve multi-step technical problems.

03Comprehensive Codebase Analysis: Explaining complex architectures or locating specific implementation patterns across large repositories.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add rawwerks/rlm-cli rlm

For use in Claude.ai and ChatGPT

Download Skill