@agent-utils/core

Core utility functions for building reliable, predictable AI agents.
Framework-agnostic • LLM-optional • Production-friendly.

Think of this as lodash for AI agents.

✨ Why this package exists

Every AI agent ends up re-building the same foundational logic:

• input normalization
• intent detection
• entity extraction
• command parsing
• guardrails
• confidence scoring
• lightweight memory helpers

@agent-utils/core provides these boring but essential utilities, so you can focus on business logic—not plumbing.

📦 Installation

npm install @agent-utils/core

🧩 What’s included

| Category | Utilities | |-------|---------| Normalization | normalizeBudget, normalizePhone, normalizeText | Intent | detectIntent | Entities | extractEntities | Commands | parseCommand | Guardrails | withinWorkingHours, rateLimit | Scoring | confidenceScore | Memory | summarizeConversation |

🚀 Quick Example

import {
  normalizeBudget,
  detectIntent,
  extractEntities,
  parseCommand,
  confidenceScore
} from "@agent-utils/core";

normalizeBudget("around 1 cr");
detectIntent("I want to visit this weekend");

📄 License

MIT