TE
科技回声
首页24小时热榜最新最佳问答展示工作
GitHubTwitter
首页

科技回声

基于 Next.js 构建的科技新闻平台,提供全球科技新闻和讨论内容。

GitHubTwitter

首页

首页最新最佳问答展示工作

资源链接

HackerNews API原版 HackerNewsNext.js

© 2025 科技回声. 版权所有。

Show HN: Zero dependency LLM standby, rollouts, no proxy

1 点作者 riddleronroof超过 1 年前
We believe you should not rely on any company for your critical LLM infra.<p>This does not proxy via another service that might go down or be beholden to whims of some board.<p>We wrote logos-shift for our own needs and are open sourcing it.<p>What it does:<p>- Save input&#x2F;output data, no proxy, no slowdowns, double buffering, separate thread<p>- Finetune your own model, or we can for a fee.<p>- Auto rollout your model when ready.<p>We recommend keeping your LLM at 80&#x2F;20 for most cases and rollout to 100 on failure.<p>How to use it:<p>pip install logos_shift_client<p>from logos_shift_client import LogosShift<p># Initialize with your API key (without if you just want the local copy)<p>logos_shift = LogosShift(api_key=&quot;YOUR_API_KEY&quot;)<p># Instrument your function<p>@logos_shift()<p>def add(x, y):<p><pre><code> return x + y </code></pre> result = add(1, 2)<p># Optionally, provide feedback logos_shift.provide_feedback(result[&#x27;bohita_logos_shift_id_123&#x27;, &quot;success&quot;)<p>We can manage all this for you of course, but its not hard to do it yourself.<p>Save data on our infra for free for upto 1GB.<p>$10 per month for after.<p>Download your data anytime. No lock in.<p>Feel free to ask questions and how can we make this easier for you.

暂无评论

暂无评论