RLM

Implements the Recursive Language Model (RLM) pattern for handling massive contexts that exceed token limits by treating content as external variables rather than loading directly into prompts. Supports multiple LLM providers (Claude SDK, Ollama), automatic content type detection with optimized chunking strategies, parallel batch processing with concurrency controls, and context persistence to disk. Designed for analyzing large codebases, documents, or datasets by decomposing them into manageable pieces while maintaining hierarchical processing capabilities through controlled recursion depth.