csf123321/InsightRadar
1
0
Fork 0
mirror of https://github.com/stardrophere/InsightRadar.git synced 2026-06-05 23:07:51 +08:00
Code Issues Packages Projects Releases Wiki Activity

optimize

Browse Source
This commit is contained in:
stardrophere
2026-03-12 01:50:08 +08:00
parent 966bcfbba4
commit e28b893a12
7 changed files with 123 additions and 14 deletions
Download Patch File Download Diff File Expand all files Collapse all files
backend/app/services/fetcher_service.py
+1 -1
View File
@@ -35,7 +35,7 @@ def generate_md5(text: str) -> str:
def generate_embedding_json(text: str) -> str: def generate_embedding_json(text: str) -> str:
"""辅助函数：调用大模型生成向量，并序列化为 JSON 字符串""" """辅助函数：调用大模型生成向量，并序列化为 JSON 字符串"""
raw_vec = embedder_model.encode([text], normalize_embeddings=True)[0] raw_vec = embedder_model.encode([text], normalize_embeddings=True, show_progress_bar=False)[0]
truncated_vec = [round(float(x), 5) for x in raw_vec] truncated_vec = [round(float(x), 5) for x in raw_vec]
return json.dumps(truncated_vec, separators=(',', ':')) return json.dumps(truncated_vec, separators=(',', ':'))
backend/app/services/matching_service.py
+38 -8
View File
@@ -37,18 +37,48 @@ def _normalize_text(text: str) -> str:
return text.strip().casefold() return text.strip().casefold()
_EMBEDDING_CACHE: dict[str, np.ndarray] = {}
MAX_CACHE_SIZE = 10000
def _build_keyword_embedding_map(keywords: list[str]) -> dict[str, np.ndarray]: def _build_keyword_embedding_map(keywords: list[str]) -> dict[str, np.ndarray]:
""" """
批量生成关键词向量，并返回原词到向量的映射。 批量生成或从缓存获取关键词向量，并返回原词到向量的映射。
这里要求向量已归一化，后续可直接用点积表示余弦相似度。 结合了批量推理（Batching）的极速优势和内存缓存的 O(1) 读取优势。
""" """
if not keywords:
return {}
vectors = embedder_model.encode(keywords, normalize_embeddings=True)
result: dict[str, np.ndarray] = {} result: dict[str, np.ndarray] = {}
for keyword, vec in zip(keywords, vectors): if not keywords:
result[keyword] = np.asarray(vec, dtype=np.float32) return result
uncached_keywords = []
# 1. 尝试从缓存获取
for keyword in keywords:
if not keyword:
continue
if keyword in _EMBEDDING_CACHE:
result[keyword] = _EMBEDDING_CACHE[keyword]
else:
uncached_keywords.append(keyword)
# 2. 对未命中的词进行统一的批量推理
if uncached_keywords:
# 去重，避免同一个未缓存的词被计算多次
unique_uncached = list(dict.fromkeys(uncached_keywords))
vectors = embedder_model.encode(unique_uncached, normalize_embeddings=True, show_progress_bar=False)
# 防止缓存无限增长：超过阈值时清空最早存入的一半（简单粗暴的内存控制）
if len(_EMBEDDING_CACHE) > MAX_CACHE_SIZE:
keys_to_delete = list(_EMBEDDING_CACHE.keys())[: MAX_CACHE_SIZE // 2]
for k in keys_to_delete:
del _EMBEDDING_CACHE[k]
# 3. 将新计算的向量存入缓存并回填结果
for keyword, vec in zip(unique_uncached, vectors):
vec_array = np.asarray(vec, dtype=np.float32)
_EMBEDDING_CACHE[keyword] = vec_array
result[keyword] = vec_array
return result return result
backend/app/services/summary_service.py
+1 -1
View File
@@ -108,7 +108,7 @@ def normalize_topic_keywords(topic_candidates: list[dict[str, Any]]) -> list[dic
return [] return []
keywords = [item["keyword"] for item in topic_candidates] keywords = [item["keyword"] for item in topic_candidates]
vectors = embedder_model.encode(keywords, normalize_embeddings=True) vectors = embedder_model.encode(keywords, normalize_embeddings=True, show_progress_bar=False)
clusters: list[dict[str, Any]] = [] clusters: list[dict[str, Any]] = []
for item, vector in zip(topic_candidates, vectors): for item, vector in zip(topic_candidates, vectors):
frontend/index.html
+1 -1
View File
@@ -2,7 +2,7 @@
<html lang="zh-CN"> <html lang="zh-CN">
<head> <head>
<meta charset="UTF-8"> <meta charset="UTF-8">
<link rel="icon" href="/favicon.ico"> <link rel="icon" href="/favicon.svg">
<meta name="viewport" content="width=device-width, initial-scale=1.0"> <meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>InsightRadar - 全网热点监控中枢</title> <title>InsightRadar - 全网热点监控中枢</title>
<!-- Font Awesome 图标库 --> <!-- Font Awesome 图标库 -->
frontend/public/favicon.svg
+69
View File
@@ -0,0 +1,69 @@
<svg viewBox="0 0 32 32" fill="none" xmlns="http://www.w3.org/2000/svg" width="32" height="32">
<style>
/* 核心呼吸灯动画 */
.ai-core-glow {
transform-origin: center;
animation: core-pulse 2s cubic-bezier(0.4, 0, 0.6, 1) infinite;
}
/* 雷达圈旋转动画 */
.radar-ring {
transform-origin: center;
}
.radar-ring.outer {
animation: spin-reverse 20s linear infinite;
}
.radar-ring.inner {
animation: spin 12s linear infinite;
}
/* 数据连线光流效果 */
.data-link {
stroke-dasharray: 4;
animation: flow 3s linear infinite;
}
@keyframes core-pulse {
0%,
100% {
transform: scale(1);
opacity: 0.4;
}
50% {
transform: scale(2.2);
opacity: 0;
}
}
@keyframes spin {
from { transform: rotate(0deg); }
to { transform: rotate(360deg); }
}
@keyframes spin-reverse {
from { transform: rotate(360deg); }
to { transform: rotate(0deg); }
}
@keyframes flow {
from { stroke-dashoffset: 8; }
to { stroke-dashoffset: 0; }
}
</style>
<circle class="radar-ring outer" cx="16" cy="16" r="14" stroke="#3b82f6" stroke-width="1" stroke-dasharray="4 8" opacity="0.4" />
<circle class="radar-ring inner" cx="16" cy="16" r="9" stroke="#3b82f6" stroke-width="1.5" stroke-dasharray="12 4" opacity="0.6" />
<path class="data-link" d="M16 16 L25 7 M16 16 L7 22 L5 20 M16 16 L23 25" stroke="#3b82f6" stroke-width="1" opacity="0.3" />
<circle class="data-node" cx="25" cy="7" r="1.5" fill="#3b82f6" opacity="0.7" />
<circle class="data-node" cx="7" cy="22" r="1.5" fill="#3b82f6" opacity="0.7" />
<circle class="data-node" cx="23" cy="25" r="1" fill="#3b82f6" opacity="0.5" />
<circle class="ai-core" cx="16" cy="16" r="3.5" fill="#3b82f6" />
<circle class="ai-core-glow" cx="16" cy="16" r="3.5" fill="#3b82f6" opacity="0.4" />
</svg>
Side by Side

After

Width:  |  Height:  |  Size: 1.9 KiB

frontend/src/views/DashboardView.vue
+3
View File
@@ -135,6 +135,9 @@ function getHotLevel(score: number): { label: string; color: string; bg: string
} }
function formatRelativeTime(dateStr: string): string { function formatRelativeTime(dateStr: string): string {
if (!dateStr.endsWith('Z') && !dateStr.includes('+')) {
dateStr += 'Z' // 补偿 SQLite 丢失的 UTC 时区标识
}
const now = Date.now() const now = Date.now()
const target = new Date(dateStr).getTime() const target = new Date(dateStr).getTime()
const diff = now - target const diff = now - target
frontend/src/views/RevisionsView.vue
+10 -3
View File
@@ -47,8 +47,15 @@ function getPlatformIcon(name: string): string {
} }
/** 格式化时间 */ /** 格式化时间 */
function safeParseTime(dateStr: string): number {
if (!dateStr.endsWith('Z') && !dateStr.includes('+')) {
dateStr += 'Z'
}
return new Date(dateStr).getTime()
}
function formatTime(dateStr: string): string { function formatTime(dateStr: string): string {
const d = new Date(dateStr) const d = new Date(safeParseTime(dateStr))
const now = Date.now() const now = Date.now()
const diff = now - d.getTime() const diff = now - d.getTime()
const minutes = Math.floor(diff / 60000) const minutes = Math.floor(diff / 60000)
@@ -75,7 +82,7 @@ const revisionChains = computed<RevisionChain[]>(() => {
const chains: RevisionChain[] = [] const chains: RevisionChain[] = []
for (const [event_id, items] of groups) { for (const [event_id, items] of groups) {
// 组内按时间升序 // 组内按时间升序
items.sort((a, b) => new Date(a.created_at).getTime() - new Date(b.created_at).getTime()) items.sort((a, b) => safeParseTime(a.created_at) - safeParseTime(b.created_at))
// 拼接标题链，避免重复（相邻记录的 revised 与下一条 previous 通常相同） // 拼接标题链，避免重复（相邻记录的 revised 与下一条 previous 通常相同）
const titles: string[] = [items[0].previous_headline] const titles: string[] = [items[0].previous_headline]
@@ -102,7 +109,7 @@ const revisionChains = computed<RevisionChain[]>(() => {
} }
// 最终按最新修改时间降序 // 最终按最新修改时间降序
chains.sort((a, b) => new Date(b.last_at).getTime() - new Date(a.last_at).getTime()) chains.sort((a, b) => safeParseTime(b.last_at) - safeParseTime(a.last_at))
return chains return chains
}) })