Transforms any text into classified, structured semantic units, acting as a cognitive primitive for AI agents.
Sponsored
Decompose provides AI agents with a crucial preprocessing step, dissecting raw text into meaningful, classified semantic units like authority, risk, attention, and entities. This tool operates deterministically without relying on Large Language Models (LLMs), ensuring speed, consistency, and offline capability. By structuring textual information before it reaches an LLM, Decompose significantly reduces token consumption, enhances relevance in retrieval-augmented generation (RAG) systems, and enables intelligent routing of content based on its importance and type, making agent workflows more efficient and precise.
Key Features
01Deterministic and LLM-free operation using regex and heuristics for speed and consistency
02Generates structured JSON output for easy integration with AI agent workflows
034 GitHub stars
04Available as an MCP server, CLI tool, or Python library
05Classifies text into semantic units (authority, risk, type, attention, entities, actionable, irreducible)
06Supports decomposition of raw text and content fetched from URLs
Use Cases
01Measure: Quantify token cost reduction by identifying and skipping boilerplate text before LLM analysis.
02Route: Direct risk-critical or financial content to specialized processing chains for targeted analysis.
03Filter: Implement attention-weighted RAG by only embedding high-priority content into vector stores.
