Files
test-repo/docs/.vitepress/theme/components/appendix/context-engineering/RAGSimulationDemo.vue
T

552 lines
13 KiB
Vue
Raw Normal View History

<!--
* Component: RAGSimulationDemo.vue
* Description: Demonstrates the Retrieval-Augmented Generation (RAG) process.
* Features:
* - Interactive search simulation
* - Visual representation of Vector DB and Document retrieval
* - Step-by-step animation of the RAG pipeline
* - Visualization of context augmentation
-->
<script setup>
import { ref, computed } from 'vue'
const query = ref('如何重置密码?')
const lastQuery = ref('')
const isSearching = ref(false)
const currentStep = ref(0)
const searchTime = ref(0)
const documents = ref([
{
id: 1,
title: '密码重置指南',
content: '用户可以通过点击设置页面的"忘记密码"链接来重置密码。系统会发送验证邮件。',
vector: [0.12, 0.88, 0.05],
score: 0
},
{
id: 2,
title: '定价策略',
content: '基础版每月 $10,专业版每月 $29。企业版需要联系销售团队获取报价。',
vector: [0.85, 0.15, 0.10],
score: 0
},
{
id: 3,
title: 'API 文档',
content: '所有 API 请求都需要在 Header 中包含 Bearer Token 进行身份验证。',
vector: [0.30, 0.20, 0.95],
score: 0
},
{
id: 4,
title: '账户安全',
content: '为了账户安全,建议开启双重认证 (2FA)。定期修改密码也是好习惯。',
vector: [0.15, 0.85, 0.12],
score: 0
}
])
const steps = [
{ id: 1, label: 'Embedding', desc: '将问题转换为向量' },
{ id: 2, label: 'Similarity Search', desc: '计算向量相似度' },
{ id: 3, label: 'Retrieval', desc: '提取 Top-K 相关文档' },
{ id: 4, label: 'Augmentation', desc: '注入上下文窗口' }
]
const retrievedDocs = computed(() => {
return documents.value
.filter(doc => doc.score > 0.7)
.sort((a, b) => b.score - a.score)
})
const calculateSimilarity = (q, docVector) => {
// Mock similarity calculation based on keywords for demo purposes
// In reality, this would be a vector dot product
if (q.includes('密码') || q.includes('安全')) {
if (docVector[1] > 0.8) return 0.92 + (Math.random() * 0.05)
if (docVector[0] > 0.8) return 0.15
return 0.4 + (Math.random() * 0.1)
}
if (q.includes('价格') || q.includes('多少钱')) {
if (docVector[0] > 0.8) return 0.95
return 0.1
}
return Math.random() * 0.3
}
const search = async () => {
if (isSearching.value || !query.value) return
isSearching.value = true
lastQuery.value = query.value
currentStep.value = 1
searchTime.value = 0
// Reset scores
documents.value.forEach(d => d.score = 0)
// Step 1: Embedding (Simulated)
await new Promise(r => setTimeout(r, 800))
currentStep.value = 2
// Step 2: Search
const startTime = performance.now()
documents.value.forEach(doc => {
doc.score = calculateSimilarity(query.value, doc.vector)
})
await new Promise(r => setTimeout(r, 800))
searchTime.value = Math.round(performance.now() - startTime) + 45 // Add base latency
currentStep.value = 3
// Step 3: Retrieval
await new Promise(r => setTimeout(r, 800))
currentStep.value = 4
// Step 4: Complete
await new Promise(r => setTimeout(r, 800))
isSearching.value = false
}
</script>
<template>
<div class="rag-simulation-demo">
<!-- Control Panel -->
<div class="control-panel">
<div class="search-bar">
<input
v-model="query"
type="text"
placeholder="输入问题 (例如: 怎么重置密码?)"
@keyup.enter="search"
:disabled="isSearching"
/>
<button
class="search-btn"
@click="search"
:disabled="isSearching || !query"
>
{{ isSearching ? '检索中...' : '🔍 开始检索' }}
</button>
</div>
<div class="step-indicator">
<div
v-for="s in steps"
:key="s.id"
class="step-dot"
:class="{ active: currentStep >= s.id, current: currentStep === s.id }"
:title="s.label"
></div>
</div>
</div>
<!-- Main Visualization -->
<div class="viz-container">
<!-- Left: Vector Database -->
<div class="panel vector-db" :class="{ dimmed: currentStep === 4 }">
<div class="panel-header">
<span class="icon">🗄</span> 向量数据库 (Knowledge Base)
</div>
<div class="doc-list">
<div
v-for="doc in documents"
:key="doc.id"
class="doc-card"
:class="{
'scanning': currentStep === 2,
'matched': doc.score > 0.7 && currentStep >= 3,
'rejected': doc.score <= 0.7 && currentStep >= 3
}"
:style="{ '--score': doc.score }"
>
<div class="doc-icon">📄</div>
<div class="doc-info">
<div class="doc-title">{{ doc.title }}</div>
<div class="doc-preview">{{ doc.content.substring(0, 20) }}...</div>
</div>
<div class="doc-score" v-if="currentStep >= 2 && doc.score > 0">
{{ (doc.score * 100).toFixed(0) }}%
</div>
<div class="vector-visual">
<span v-for="(v,i) in doc.vector" :key="i" :style="{ height: v * 10 + 'px' }"></span>
</div>
</div>
</div>
</div>
<!-- Center: Pipeline Visuals -->
<div class="pipeline-arrow">
<div class="arrow-line" :class="{ active: isSearching }"></div>
<div class="pipeline-status" v-if="currentStep > 0">
{{ steps[currentStep - 1]?.label }}
</div>
</div>
<!-- Right: Augmented Context -->
<div class="panel context-window" :class="{ active: currentStep === 4 }">
<div class="panel-header">
<span class="icon">🤖</span> 增强后的上下文 (Final Prompt)
</div>
<div class="prompt-content">
<div class="prompt-section system">
<span class="tag">System</span>
<p>你是一个帮助用户的 AI 助手请基于以下上下文回答用户的问题</p>
</div>
<div class="prompt-section context" v-if="currentStep >= 3">
<span class="tag">Context (RAG)</span>
<div v-if="retrievedDocs.length > 0">
<div v-for="doc in retrievedDocs" :key="doc.id" class="retrieved-item">
<span class="bullet"></span> {{ doc.content }}
</div>
</div>
<div v-else class="empty-context">
(暂无相关文档)
</div>
</div>
<div class="prompt-section user" v-if="lastQuery">
<span class="tag">User</span>
<p>{{ lastQuery }}</p>
</div>
<div class="placeholder-text" v-else>
等待查询...
</div>
</div>
</div>
</div>
<!-- Metrics Footer -->
<div class="metrics-footer">
<div class="metric">
<span class="label">检索耗时:</span>
<span class="value">{{ searchTime }} ms</span>
</div>
<div class="metric">
<span class="label">命中数量:</span>
<span class="value">{{ retrievedDocs.length }} docs</span>
</div>
</div>
</div>
</template>
<style scoped>
.rag-simulation-demo {
border: 1px solid var(--vp-c-divider);
border-radius: 8px;
background-color: var(--vp-c-bg-soft);
overflow: hidden;
margin: 1rem 0;
display: flex;
flex-direction: column;
}
.control-panel {
padding: 1rem;
background-color: var(--vp-c-bg);
border-bottom: 1px solid var(--vp-c-divider);
display: flex;
justify-content: space-between;
align-items: center;
flex-wrap: wrap;
gap: 1rem;
}
.search-bar {
display: flex;
gap: 0.5rem;
flex: 1;
min-width: 280px;
}
input {
flex: 1;
padding: 0.5rem 0.8rem;
border: 1px solid var(--vp-c-divider);
border-radius: 6px;
background-color: var(--vp-c-bg-alt);
color: var(--vp-c-text-1);
}
input:focus {
border-color: var(--vp-c-brand);
outline: none;
}
.search-btn {
padding: 0.5rem 1rem;
background-color: var(--vp-c-brand);
color: white;
border-radius: 6px;
font-weight: 500;
transition: background-color 0.2s;
}
.search-btn:hover:not(:disabled) {
background-color: var(--vp-c-brand-dark);
}
.search-btn:disabled {
opacity: 0.6;
cursor: not-allowed;
}
.step-indicator {
display: flex;
gap: 0.4rem;
}
.step-dot {
width: 10px;
height: 10px;
border-radius: 50%;
background-color: var(--vp-c-divider);
transition: all 0.3s;
}
.step-dot.active {
background-color: var(--vp-c-brand);
}
.step-dot.current {
transform: scale(1.4);
box-shadow: 0 0 4px var(--vp-c-brand);
}
/* Viz Container */
.viz-container {
display: flex;
padding: 1.5rem;
gap: 1rem;
background-color: var(--vp-c-bg-alt);
min-height: 350px;
align-items: stretch;
}
.panel {
flex: 1;
background-color: var(--vp-c-bg);
border: 1px solid var(--vp-c-divider);
border-radius: 8px;
display: flex;
flex-direction: column;
transition: all 0.5s ease;
}
.panel.dimmed {
opacity: 0.6;
filter: grayscale(0.5);
}
.panel.active {
border-color: var(--vp-c-brand);
box-shadow: 0 0 15px rgba(var(--vp-c-brand-rgb), 0.1);
}
.panel-header {
padding: 0.8rem;
font-weight: 600;
font-size: 0.9rem;
border-bottom: 1px solid var(--vp-c-divider);
display: flex;
align-items: center;
gap: 0.5rem;
background-color: var(--vp-c-bg-soft);
}
.doc-list {
padding: 0.8rem;
display: flex;
flex-direction: column;
gap: 0.6rem;
overflow-y: auto;
max-height: 300px;
}
.doc-card {
padding: 0.6rem;
border: 1px solid var(--vp-c-divider);
border-radius: 6px;
display: flex;
align-items: center;
gap: 0.6rem;
font-size: 0.85rem;
position: relative;
transition: all 0.3s;
background-color: var(--vp-c-bg);
}
.doc-card.scanning {
animation: pulse 1s infinite;
border-color: var(--vp-c-brand-dimm);
}
.doc-card.matched {
border-color: var(--vp-c-green);
background-color: var(--vp-c-green-dimm);
transform: translateX(5px);
}
.doc-card.rejected {
opacity: 0.5;
}
.doc-icon {
font-size: 1.2rem;
}
.doc-info {
flex: 1;
overflow: hidden;
}
.doc-title {
font-weight: 600;
color: var(--vp-c-text-1);
}
.doc-preview {
color: var(--vp-c-text-2);
font-size: 0.75rem;
white-space: nowrap;
overflow: hidden;
text-overflow: ellipsis;
}
.doc-score {
font-family: var(--vp-font-mono);
font-weight: bold;
color: var(--vp-c-brand);
}
.vector-visual {
display: flex;
gap: 2px;
align-items: flex-end;
height: 15px;
width: 20px;
}
.vector-visual span {
width: 4px;
background-color: var(--vp-c-text-3);
border-radius: 1px;
}
/* Pipeline Arrow */
.pipeline-arrow {
width: 40px;
display: flex;
flex-direction: column;
justify-content: center;
align-items: center;
position: relative;
}
.arrow-line {
width: 100%;
height: 4px;
background-color: var(--vp-c-divider);
border-radius: 2px;
transition: all 0.3s;
}
.arrow-line.active {
background: linear-gradient(90deg, var(--vp-c-brand), var(--vp-c-brand-light));
background-size: 200% 100%;
animation: flow 1s linear infinite;
}
.pipeline-status {
position: absolute;
top: 40%;
left: 50%;
transform: translate(-50%, -50%);
background-color: var(--vp-c-brand);
color: white;
padding: 0.2rem 0.6rem;
border-radius: 12px;
font-size: 0.7rem;
white-space: nowrap;
z-index: 10;
}
/* Context Window */
.prompt-content {
padding: 1rem;
display: flex;
flex-direction: column;
gap: 1rem;
font-family: var(--vp-font-mono);
font-size: 0.85rem;
overflow-y: auto;
}
.prompt-section {
background-color: var(--vp-c-bg-soft);
padding: 0.8rem;
border-radius: 6px;
border-left: 3px solid transparent;
}
.prompt-section.system {
border-left-color: var(--vp-c-yellow);
}
.prompt-section.context {
border-left-color: var(--vp-c-green);
background-color: rgba(var(--vp-c-green-rgb), 0.1);
}
.prompt-section.user {
border-left-color: var(--vp-c-brand);
}
.tag {
display: inline-block;
font-size: 0.7rem;
font-weight: bold;
text-transform: uppercase;
margin-bottom: 0.4rem;
color: var(--vp-c-text-2);
}
.retrieved-item {
margin-top: 0.4rem;
color: var(--vp-c-text-1);
}
.empty-context {
color: var(--vp-c-text-3);
font-style: italic;
text-align: center;
}
.placeholder-text {
text-align: center;
color: var(--vp-c-text-3);
margin-top: 2rem;
}
/* Metrics Footer */
.metrics-footer {
display: flex;
justify-content: space-around;
padding: 0.8rem;
background-color: var(--vp-c-bg);
border-top: 1px solid var(--vp-c-divider);
font-size: 0.85rem;
}
.metric .label {
color: var(--vp-c-text-2);
margin-right: 0.5rem;
}
.metric .value {
font-weight: bold;
color: var(--vp-c-text-1);
}
@keyframes flow {
0% { background-position: 100% 0; }
100% { background-position: -100% 0; }
}
@keyframes pulse {
0% { opacity: 1; }
50% { opacity: 0.6; }
100% { opacity: 1; }
}
@media (max-width: 768px) {
.viz-container {
flex-direction: column;
}
.pipeline-arrow {
width: 100%;
height: 40px;
flex-direction: row;
}
.arrow-line {
width: 4px;
height: 100%;
}
}
</style>