1
Show HN: Llama CPU Benchmarks
TurboQuant号称8倍速，实测CPU端到端慢2.2倍，Qwen准确率还降17个百分点，别被合成数据骗了。
Article URL: https://deemwar-products.github.io/llama-cpu-benchmarks/ Comments URL: https://news.ycombinator.com/item?id=48212222 Points: 1 # Comments…
2
Experimenting with graph-based semantic memory for AI agents
开源AI代理图语义记忆项目，整合SQLite、llama.cpp、BGE-M3等，30秒可安装体验。
Article URL: https://github.com/AEndrix03/Graft Comments URL: https://news.ycombinator.com/item?id=48216282 Points: 4 # Comments: 0
3
Show HN: Autodidact – Self-evolving local-first AI agent
本地优先AI代理自动处理Ollama依赖，安装模型启动一气呵成，适合不想折腾环境的开发者。
pip install autodidact && autodidact init Comments URL: https://news.ycombinator.com/item?id=48194739 Points: 4 # Comments: 0
4
Show HN: Llama-dash – local LLM operators dashboard and proxy
本地LLM运维仪表板，集成路由、日志、监控，轻松管理多个模型上游。
Article URL: https://github.com/ndom91/llama-dash Comments URL: https://news.ycombinator.com/item?id=48196202 Points: 1 # Comments: 0
5
Streaming Ollama Responses in Next.js: The SSE Pattern That Actually Works
告别8秒等待！用SSE在Next.js中实现Ollama流式输出，生产级代码不到百行
Streaming Ollama Responses in Next.js: The SSE Pattern That Actually Works Most Next.js + Ollama tutorials show a single await fetch and call it a day…
6
Show HN: I containerized my AI agents and dev tools
容器化集成多种顶级AI编码代理和开发工具，一键部署本地开发环境。
I have been juggling a bunch of different tooling to keep agents locked down on my local system. This weekend I formalized a container build + python …
7
I tested 8 LLM models on Linux without using the GPU
8款LLM模型无需GPU在Linux上实测，CPU推理新选择关注真实可用性
Article URL: https://itsfoss.com/testing-local-llms-without-gpu/ Comments URL: https://news.ycombinator.com/item?id=48147334 Points: 2 # Comments: 0
8
ClickBook – Offline Android eReader with local LLM inference via llama.rn
想在手机上离线看书还能用AI帮你理解？ClickBook把本地大模型塞进安卓阅读器，不联网也能玩转摘要、问答，真正把AI装进口袋。
Article URL: https://play.google.com/store/apps/details?id=com.clickbook.reader&hl=en_US Comments URL: https://news.ycombinator.com/item?id=481632…