Show HN: Jevper – an LLM API client constrained to the Jev wire format

AI 导读

Jevper 是一个基于 System One 协议的 LLM 客户端，兼容 OpenAI 接口，支持本地或托管模型，提供结构化问答与概率输出。

Jevper is an LLM client implementing System One protocol, compatible with OpenAI APIs, supporting both hosted and self-hosted models with structured question-answer outputs.

重点速览

• 独立实现 System One 协议，不依赖 TypeSafe AI 或 openai 库 Independent implementation of System One protocol without TypeSafe AI dependency
• 支持 OpenAI 兼容模型及本地 llama.cpp 服务器 Supports OpenAI-compatible models and local llama.cpp servers
• 通过 OpenResponses 规范统一响应格式，兼容 LM Studio 等工具 Unified response format via OpenResponses specification
• 提供意图识别示例，输出包含概率与置信度数据 Provides intent recognition example with probability/confidence outputs

一句话：Jevper 为多模型兼容的结构化问答提供了灵活的协议适配方案。 / Jevper offers flexible protocol adaptation for structured question-answering across multiple models.

推荐理由 值得读取其多模型兼容设计与协议适配方案，适合需要灵活部署的 AI 应用开发。