UETGPT

io.github.Ibrahim-Salman19v1.0.0更新于 Oct 2, 2026

Unofficial UET Taxila guide: merit calculator, programs, fees and cited admissions answers.

已验证Streamable HTTP可网页运行Knowledge & Memory

概览

AI 生成的概览

回答关于 UET Taxila 的招生、学费、专业、师资与校园生活问题，并给出引用官方来源的答案。

功能
UETGPT 是塔克西拉工程与技术大学（UET Taxila）的非官方指南。它用自然语言回答有关招生、学费结构、学术项目、院系、师资、考试、校园生活、交通、宿舍和奖学金的问题。答案来自对学校网页和 PDF 文档的检索增强生成流程，并附有指向官方来源的引用链接。
适用场景
当你需要为潜在学生、在读学生、家长或教师快速获取有引用的 UET Taxila 信息，而不想翻找零散的学校网页和 PDF 时使用。它不是通用助手，只覆盖这一所大学。
运行要求
远程使用只需 streamable HTTP 端点。自托管需要 pnpm、Node.js、Python 3.x、Convex 账户、Clerk 应用，以及 Groq、Gemini、Cerebras 的 API 密钥，并在 .env.local 和 Convex 控制台中设置。
安装前请注意
README 说明自托管需要 Convex 账户、Clerk 应用以及 Groq、Gemini、Cerebras 的 API 密钥，但未列出具体环境变量名，部署前请查阅示例环境变量文件。该托管端点是非官方指南，答案应与学校官方来源核对。远程端点未声明任何身份验证。

安装

在 SourceWeft 中

  1. 打开 控制台中的 UETGPT，将其添加到工作区。
  2. 为需要使用其工具的对话启用该服务。

Web executable，通过 Streamable HTTP。 远程服务在工作区中配置后即可从网页运行时运行。

其他 MCP 客户端

把它添加到你客户端的 mcpServers 配置中。

{
  "mcpServers": {
    "uetgpt": {
      "type": "http",
      "url": "https://uet-gpt.vercel.app/mcp"
    }
  }
}

README

UET GPT - Your AI Guide to UET Taxila

An intelligent AI chatbot that answers any question about the University of Engineering and Technology (UET) Taxila - admissions, fee structure, academic programs, departments, faculty, campus life, transport, hostels, scholarships, and more. Powered by RAG over official UET Taxila data.

One-line description (SEO): UET GPT is an AI chatbot that gives instant, cited answers about UET Taxila - admissions, fees, programs, faculty, and campus life - using RAG over the university's official data. Licensed under AGPL-3.0.

UET GPT is an autonomous RAG (Retrieval-Augmented Generation) chatbot for UET Taxila. It answers questions about admissions, departments, fees, exams, faculty, and campus life using a hybrid search + LLM generation pipeline over a corpus of crawled university web pages and ingested PDFs.

Live app: https://uet-gpt.vercel.app · Press kit

What is UET GPT?

UET GPT is a specialized AI assistant for the University of Engineering and Technology (UET), Taxila - one of Pakistan's premier engineering institutions, founded as a UET Lahore campus in 1975 and granted its independent charter in 1993, now serving 5,000+ students across 30+ programs. Instead of digging through scattered university web pages and PDFs, students, applicants, parents, and faculty ask UET GPT in plain language and get fast, accurate answers backed by citations from official UET Taxila sources.

Key things to know:

  • Built for UET Taxila - every answer is grounded in UET Taxila's official website (web.uettaxila.edu.pk), admissions portal (admissions.uettaxila.edu.pk), and ingested documents.
  • Retrieval-Augmented Generation (RAG) - answers are retrieved from a continuously crawled, indexed knowledge base, not hallucinated from generic model memory.
  • Cited and verifiable - responses link back to the official university material they came from.
  • Open source - the full stack is released under AGPL-3.0 and is free to use.

UET GPT is currently free to use at https://uet-gpt.vercel.app.

Stack

  • Frontend: Next.js 16 (App Router, Turbopack), React 19, Tailwind CSS 4, Radix UI
  • Backend: Convex (real-time DB + serverless functions, native vector index)
  • Auth: Clerk (RBAC: user / admin / superadmin)
  • LLM: Vercel AI SDK with a Groq → Cerebras → Gemini fallback chain
  • Crawler: Python async BFS crawler (curl_cffi, trafilatura)

See architecture.md - the single source of truth - for the full system design.

Prerequisites

  • pnpm (this repo is pnpm-locked; do not mix npm/yarn - see architecture.md §17)
  • Node.js (version per package.json / .nvmrc)
  • Python 3.x (for the async BFS crawler and PDF ingest scripts in scripts/)
  • A Convex account, a Clerk application, and API keys for Groq / Gemini / Cerebras

Environment Variables

Set these in .env.local (Next.js) and in the Convex dashboard as appropriate. See .env.local.example for the full list with comments, and architecture.md §8.2 for the authoritative descriptions.

Getting Started

bash
pnpm install
# Run the Convex backend (in one terminal)npx convex dev
# Run the Next.js dev server (in another terminal)pnpm dev

Open http://localhost:3000 to use the app.

Testing

bash
pnpm test        # Vitest unit + integration testspnpm lint        # Biome lintpnpm test:e2e    # Playwright end-to-end tests

Documentation

  • architecture.md - single source of truth for system design
  • AGENTS.md - agent / contributor instructions
  • testing.md - testing strategy (trimmed, agent-optimized)
  • THREAT_MODEL.md - security threat model
  • CRONJOB.md - autonomous maintenance protocol
  • DESIGN.md - design system (Neo Kinpaku)
  • PRODUCT.md - product definition
  • CHANGELOG.md - engineering work log
  • docs/ - focused reference docs:
    • docs/reference/ - API signatures, env vars, auth matrix, component inventory
    • docs/security.md - security controls summary
    • docs/embedding-strategy.md - embedding model + vector store
    • docs/agent-coordination.md - frontend/backend agent boundary rules
    • docs/frontend_backend_boundaries.md - ownership rules between src/ and convex/
    • docs/deployment.md - deployment architecture
    • docs/crawl-robots-audit.md - robots.txt compliance decisions

来源：README.md，提交 aaf1448

工具

0
工具元数据尚未被收录。

版本历史

1
  1. v1.0.0最新Oct 2, 2026