From c64b58b560f4dc8586697466ca4a0464ac6bff49 Mon Sep 17 00:00:00 2001 From: Sunil Tripathy <143227935+CyberSunil@users.noreply.github.com> Date: Mon, 10 Aug 2026 14:33:05 +0100 Subject: [PATCH] Update README for LLMVault v2.0 release --- README.md | 53 ++++++++++++++++++++++++----------------------------- 1 file changed, 24 insertions(+), 29 deletions(-) diff --git a/README.md b/README.md index f72b998..e1b594a 100644 --- a/README.md +++ b/README.md @@ -2,11 +2,11 @@ LLMVault Banner

-# LLMVault +# LLMVault v2.0

-The Ultimate Hands-On OWASP LLM Top 10 Training Platform +### 🚀 The Ultimate Hands-On OWASP LLM Top 10 Training Platform

@@ -56,33 +56,28 @@ Every lab pairs the **attack** with a **defense**: solve it, then read what woul > ⚠ī¸ **Everything here is intentionally insecure. Authorised, self-hosted security education only.** Don't expose it to the internet or reuse its code in production. -## 🆕 What's new in v1.2.0 - -**Live Mode.** The headline feature. Instead of a scripted bot, you're now talking to a real -local model (Ollama, ~2 GB, CPU is fine) and the target secret is minted per session. The model -is genuinely unpredictable, which is the point: a payload that worked yesterday may not work -today, and you have to read what the model actually does with your text. Three scenarios ship -with it, covering prompt injection, indirect injection through an uploaded image, and output -handling where the leak only shows up downstream. Hints are free and unlimited here — there's -no score to protect. Play Mode is untouched, and if you skip the Ollama setup the Live Zone just -shows you the install steps and everything else works as before. - -**A real dashboard.** `/` is now Mission Command rather than a splash page: the two modes -side by side, per-tier progress meters, your rank and score, a live activity feed across both -modes, badges, and a status readout that tells you whether the local model is actually reachable. -Everything on it is computed from your real progress — if there's no data yet, the panel says so -instead of inventing a number. - -**Milestone pop-ups that don't skip.** Finish Core, Advanced or Expert and the card pops up on -the spot, each one carrying the badges you now hold. Earning Advanced lights Core alongside it -rather than quietly replacing it. - -**Notification centre.** One bell, identical on every page, with a Play / Live toggle so you see -one mode at a time, a minimise control, and a dot on the Live tab when the local model needs -attention. - -**Two dashboard themes.** Neon (the orange-and-lime default) and Phosphor, a green-CRT reading of -the same screen. Colour only — no layout shifts either way. +## 🚀 What's new in v2.0.0 + +### 🔴 Live Mode + +A real local model instead of a scripted bot — genuinely unpredictable, so yesterday's payload may die today. + +- 🧠 Ollama, ~2 GB, CPU is fine; secret minted fresh each session +- đŸŽ¯ Three scenarios: prompt injection, indirect injection via image, downstream output handling + +### 📊 A real dashboard + +- ⚔ī¸ Both modes side by side, per-tier progress, rank, score, live activity feed +- 🎨 Two themes: Neon (default) and Phosphor green-CRT + +### 🏅 Milestones and alerts + +- 🎖ī¸ Core / Advanced / Expert cards pop on the spot, carrying your badges +- 🔔 One Notification everywhere, Play / Live toggle, dot when the model needs attention + +### 🕹ī¸ Play Mode is untouched + +Everything you had still works exactly as before. ### Previously, in v1.1.0