llmproxy

ONE API，LLM_Proxy，统一代理，支持请求响应自定义转换。

A * very * simple guide to using Deepseek or any other models on Janitor.ai

omnillm is the only LLM proxy you need for any api shapes and any LLM providers, one endpoint for all.

llm proxy as openai/ollama， upstream llm only support openai

A production-inspired AI platform framework, built module by module: a FastAPI Gateway in front of a provider-agnostic model layer, an agentic Runtime with tool-calling and persisted memory, span-based tracing, an offline evaluation harness, and a containerized, CI-tested deployment path.

该组件为基于 Go 构建的高性能 HTTP 反向代理，专用于将客户端请求透明转发至兼容 OpenAI 或 Anthropic 协议的上游服务。在转发的同时，代理以无侵入方式完整捕获每次会话的请求与响应，包括流式传输中产生的所有数据分片，并以 JSONL 格式实时持久化至本地存储。这一机制使每一轮交互均可被精确追溯，从而满足合规审计、行为分析及训练语料构建等需求。系统还提供可选的按日历日合并导出功能，可将当日全部会话记录重组织为统一的行式数据集，直接服务于监督微调、偏好对齐等训练数据的准备工作

Go proxy that logs every LLM request, tracks token usage and costs in SQLite, and serves a live dashboard with rate limiting and model remapping, I built this to be used with Hermes to see token usage

Clausy allows agent frameworks like OpenClaw to use browser-based LLM providers like ChatGPT, xAi, Gemini and Claude.