返回资讯中心
外部精选
软件工程
#开发者实时流

Show HN: Risphai – picks the AI model, tools and prompt for each request

I noticed that a lot of non-technical people don’t use AI to its full extent. Like they use the wrong model for the task, the wrong reasoning effort, or vague prompts. That gap between "what the AI can do" and "what peo…

Hacker News 最新risphere2026/10/18 · 2026/10/184 分钟阅读

以下正文同步自 Hacker News 最新,版权归原站所有,已转换为易读排版。

I noticed that a lot of non-technical people don’t use AI to its full extent. Like they use the wrong model for the task, the wrong reasoning effort, or vague prompts. That gap between "what the AI can do" and "what people actually get" is what I wanted to fix.So I spent the last two years in my spare time building a consumer AI product that tries to fix this issue.This is how it works: When it receives the request, an LLM decides the task type (chat, reasoning, coding, writing) and the tool calls that are needed (web search, image generation, video generation, audio generation). Then it routes it to a suitable model and applies a prompt based on the task type. If it's a media generation task, it rewrites the prompt into a detailed one using prompt formulas (the prompt formulas for specific styles are mentioned in the system prompt). The LLM that is used for answering the request can also use a map, generate a mermaid diagram, generate a quiz or flashcards, generate files, etc.A limitation right now is that routing adds latency, but I'm currently experimenting with decision-models like Jev, PPLX Decider, etc. and might use decision-models to route to the right LLM and give the tool calls like web search and media generation to the answering LLM.It launched this week, so it's still early and I will keep improving it and adding new things.Also sign-up/in is required to prevent abuse.I’d really like honest feedback: what’s confusing, what’s missing, etc. Comments URL: https://news.ycombinator.com/item?id=50139344 Points: 1

Comments: 1

正文由 探How 从来源站点 RSS 同步,内容未经改写;遇到排版缺失或需要图片、视频时请以原文为准。

打开原文
去社区讨论 →