使用 HazelJS 的提示管理、RAG 和代理

📄 中文摘要

有效管理大型语言模型(LLM)的提示非常具有挑战性,需要实现版本控制、无需重新部署的覆盖以及一个统一的读取源,供 RAG、代理和普通 AI 任务使用。HazelJS 提示启动器展示了如何实现这些功能。该工具基于 HazelJS 框架,结合了 @hazeljs/prompts 和 @hazeljs/rag 等库,提供了类型化的提示模板、实时注册、文件存储持久性等功能,旨在简化 AI 任务的管理和执行。

📄 English Summary

Prompt management, RAG, and agents with HazelJS

Managing prompts for large language models (LLMs) is challenging, requiring versioning, overrides without redeploys, and a single source for RAG, agents, and standard AI tasks. The HazelJS Prompt Starter demonstrates how to achieve this. Built on the HazelJS framework, it integrates libraries such as @hazeljs/prompts and @hazeljs/rag, offering typed prompt templates, a live registry, and FileStore persistence. This tool aims to streamline the management and execution of AI tasks.

Powered by Cloudflare Workers + Payload CMS + Claude 3.5

数据源: OpenAI, Google AI, DeepMind, AWS ML Blog, HuggingFace 等