// today's trending OSS

Signal over the noise.

Daily-ranked open-source repos from GitHub & Hacker News, distilled by AI into bilingual summaries.

● 93 repos listed (UTC)
last 14 days · new repos2026-04-19 → 2026-05-02
REPOS / WK
175
+113
AI SUMMARIES
381
+175
TAGS TRACKED
483
UPTIME
27d
● LIVE · 24h movers
AIFlashMLA: Optimized Attention Kernels for DeepSeek Models★ 12,617DatabaseInstant: A Real-time Client-side Database Backend★ 10,217AILiteLLM: Unified AI Gateway for 100+ LLMs★ 45,389AITabby: Self-Hosted AI Coding Assistant★ 33,480DatabaseTiDB: A Distributed SQL Database★ 40,048AIDeepEP: High-Performance Expert Parallelism Library★ 9,594DevOpsHyperDX Observability Platform for ClickHouse★ 9,473DatabasePostgres Language Server★ 5,225FrontendAnt Design: Enterprise-Class React UI Library★ 97,903AIKhoj: Your Personal AI Second Brain★ 34,330

// today's trending

view all 93 repos →
#1FlashMLA: Optimized Attention Kernels for DeepSeek Models[93]AI#2Instant: A Real-time Client-side Database Backend[92]Database#3LiteLLM: Unified AI Gateway for 100+ LLMs[89]AI#4Tabby: Self-Hosted AI Coding Assistant[88]AI#5TiDB: A Distributed SQL Database[87]Database

// weekly report

all reports →
2026-W182026-04-27 — 2026-05-03
93 repos93 new1 days
#1deepseek-ai-FlashMLA1d
#2instantdb-instant1d
#3BerriAI-litellm1d

// monthly rankings

all rankings →
2026-0593 repos
AI60Frontend8Other7Database6Backend4DevOps3Security3Mobile2

// newcomers

see all →
Released in last 6 months · 79
#1Khazix Skills: Personal Practical AI Skills and Prompt Collection🆕 26d#2ZhangXuefeng.skill: A Thinking Framework Based on Cognitive Models🆕 27d#3Hermes Agent: The Complete Guide (Orange Book)🆕 24d#4OpenSandbox: General-Purpose AI Sandbox Platform🆕 5mo#5Toonflow-app: One-stop AI Short Drama Creation Workbench🆕 3mo

// comparisons

all comparisons →
FlashMLA: Optimized Multi-head Latent Attention KernelsvsFlashMLA: Optimized Multi-head Latent Attention KernelsLiteLLM: Unified AI Gateway for 100+ LLMsvsFlashMLA: Optimized Multi-head Latent Attention KernelsPaddleOCR: Global Leading OCR Toolkit and Document AI EnginevsFlashMLA: Optimized Multi-head Latent Attention KernelsVoicebox: Local-First Open-Source Voice Synthesis StudiovsWillow Inference ServerVoicebox: Local-First Open-Source Voice Synthesis StudiovsNeuTTS: Open-Source On-Device Voice AI and Instant CloningNeuTTS: Open-Source On-Device Voice AI and Instant CloningvsWillow Inference Server

// browse by topic

all →
LLM117Automation90AI Agents55TypeScript46Python40React39CLI37Deep Learning33Docker29AI29Agent28Rust25
// what it does · capabilities
6 modules · all free tier
MODULE.01
Real-time trend detection
A rolling window over GitHub & HN flags star velocity anomalies as they break out, not days later.
MODULE.02
EN · 120w
A unified gateway that routes LLM calls across 100+ providers…
→
中文 · 120字
统一的 LLM 网关，支持 100+ 服务商之间的路由与调度…
AI summaries in two languages
Vertex AI reads the README, issues, and source structure, then writes a gist in both EN and 中文.
MODULE.03
Category-aware ranking
Separate leaderboards per category mean "infra" never drowns under an AI wave — specialists see their own signal.
MODULE.04
repo A
★ 42,978
◆ 312 contrib
vs
repo B
★ 12,097
◆ 48 contrib
Head-to-head comparisons
Pick two repos; get a side-by-side diff of stars, tags, category, and AI-extracted highlights.
MODULE.05
crawl
queue
llm
render
cdn
Serverless by design
Crawl, summarise, render, distribute — everything runs on Cloud Run + Firestore + Cloudflare for pennies per day.
MODULE.06
Static, fast, forkable
Output is plain HTML served from CDN. Public REST API + MCP server for anyone to build on top.
// integrate · mcp + api

Every ranking is also a JSON endpoint.

Plug HubLens into Claude, Cursor, Windsurf, or any agent via the official MCP server — or hit the public REST API straight from curl. No key for the first 1,000 requests per day.

→ MCP docs→ API reference→ npm package
MCP SERVER
~/.claude/mcp.json● stdio
1# Add HubLens to any MCP-compatible client
2npx -y @hublens/mcp-server
3
4# claude_desktop_config.json
5{
6 "mcpServers"
7 : {
8 "hublens"
9 : {
10 "command"
11 : "npx",
12 "args"
13 : ["-y", "@hublens/mcp-server"]
14 }
15 }
16}
REST API
~/hublens/query.sh● 200 OK · 184ms
1# Ask HubLens for this week's trending AI repos
2curl -s https://hublens.dev/api/v1/trending \
3 -G --data-urlencode "category=AI" \
4 --data "limit=5"
5
6→ {
7 "generated_at"
8 : "2026-04-21T06:42:00Z",
9 "repos"
10 : [
11 {
12 "slug"
13 : "berriai-litellm",
14 "stars"
15 : 42978,
16 "summary_en"
17 : "Unified AI gateway…"
18 }, …
19 ]
20}