5 fun, easy-to-start, runnable creative projects for JEV | 5个好玩、好上手、能直接跑的创意项目

·27 views

一句话先记住TEV能干啥 JEV*不会写作文、不会讲故事，它就像一个超快的选择题阅卷员*。你一次性丢给它好几个判断题/单选题/打分题，它一瞬间全部做完，每一题附带「我有几成把握」的概率。搭配一个会写东西的大模型（System2）一起玩，才出彩。大模型负责思考、写内容；JEV负责快速判断、做选择、把关校验。思路分支判断、选项选择、状态校验。玩家输入一句话，TEV并行一次性判断多件事： Noul：玩家这句话是不是攻击NPC？（是/否+概率） Choice：玩家动作属于【交谈 / 偷东西 / 逃跑 / 战斗】哪一类 Score：NPC愤怒等级（1平静～5暴怒）代码写规则：如果TEV判断「攻击」概率>0.8，直接触发战斗线；如果置信度低（0.4~0.7），交给大模型仔细理解；低于0.3，按普通对话走。亮点大量分支自动分流，游戏手感丝滑。你甚至可以做：TEV判断任务是否完成，满足条件自动解锁结局。 System1=TEV（快直觉）；System2=Qwen/Llama（慢思考）分工 TEV（高频调用，便宜快） Choice：我下一步该调用哪个工具？【搜索 / 读文件 / 代码执行 / 直接回答】 Noul：当前任务做完了吗？要不要继续思考？ Noul：模型输出内容有没有违规？是不是幻觉？大模型（低频调用，耗资源）只在TEV拿不准的时候（概率落在中间区间），才调用大

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

PHP PDF generation becomes slow when creating large documents

I am working on a PHP-based worksheet generation system and recently ran into a performance problem. The application generates printable PDF files based on user-selected options. The workflow is: User selects options ↓ Server generates data ↓ PHP creates PDF ↓ User downloads the file The problem is that the generation time increases significantly when the document contains many items. For example: 100+ exercises multiple pages answer pages custom layouts Current implementation The frontend sends a POST request to the PDF generation endpoint: $result = http_post( https://www.wativate.com/?api/p

0 comments Read more at DEV Community

ProgrammingDEV Community ·

I built a live map of natural events with Vue 3 and OpenLayers. Here are the mistakes I made.

Hi everyone 👋 I'm Andrea, a full-stack developer from Italy, living in Tenerife. Yes, that one, the island with the big Teide, a volcano, which is maybe why I ended up building this. I have been building Vue apps in production for about four years: a blockchain explorer, a marketplace, editors, dashboards, data visualisation pages. What I had never built is a map. No GIS, no projections, no OpenLayers.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

🌱 Last Frost: A Garden Planner That Only Works If You Go Outside

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass Last Frost is an offline garden planner. You type your ZIP code and what you want to grow; it looks up your local first fall frost and last spring frost dates, then a small open-weight model running on your own machine tells you what to plant this week, what to hold back, and why — based on the actual number of frost-free days you have left. It's for anyone whose gardening app experience is: open app → get generic advice → still not know if it's too late to plant tomatoes. Last Frost replaces "generic" with

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Jev-LCT: Free Calibrated Confidence from Looped Transformer Trajectories

Small decision models have two bad options today: generate text token-by-token (slow, fragile to parse), or "introspect" their confidence verbally (systematically miscalibrated). Jev-LCT is an open-source System-One decision engine that takes a third route — extracting calibrated probabilities directly from a transformer's internal recurrent dynamics. Parallel Looped Prefill: recurrently recomputes only the top k=2 layers of a causal transformer, preserving full representation fidelity. Endogenous Trajectory Confidence: reads calibrated probabilities from hidden-state convergence dynamics (cos

0 comments Read more at DEV Community