How long does the integration take?

Change one base URL in an OpenAI-compatible client, or install the Python or TypeScript SDK; the quickstart routes a first request in a few minutes.

Will this lock us into another vendor?

No. The router is Apache-2.0 open source, your provider keys stay yours, and the API is OpenAI-compatible, so leaving is a change of base URL.

What happens when a provider fails?

The router falls back to the next model your policy allows and records the fallback in the trace, so an outage costs a retry, not an incident.

song-analysis - Skill - OpenSmartRoute

Name: song-analysis
Author: cholateio

song-analysis

花譜（KAF）cover 曲的 YouTube ingest + 音訊分析 pipeline：用 yt-dlp 抓音訊 → Meyda 做每幀 spectrum 分析 → 打包成 binary blob 寫入 Supabase （kaf_songs 表 + song-blobs storage bucket），供 3D 前端視覺化消費。前端消費者為 kaf-observatory（本 repo…

by cholateio(0) 0 installs

Imported from cholateio/song-analysis (AGENTS.md). Install upstream with npx skills add cholateio/song-analysis. Copyright stays with the author.

song-analysis

Project goal

花譜（KAF）cover 曲的 YouTube ingest + 音訊分析 pipeline：用 yt-dlp 抓音訊 → Meyda 做每幀 spectrum 分析 → 打包成 binary blob 寫入 Supabase （kaf_songs 表 + song-blobs storage bucket），供 3D 前端視覺化消費。前端消費者為 kaf-observatory（本 repo 的 docs/reference/clock-analysis-handover.md 對應其同名檔；歷史名 virtual-music-clock）。

Stack

Language: Node.js ≥20（ESM，"type": "module"），pnpm
核心依賴: meyda（spectrum 特徵）、@supabase/supabase-js、dotenv、minimist
外部 CLI 依賴（不在 package.json）: yt-dlp（src/youtube.mjs 經 child_process 呼叫）、ffmpeg（LRA 由 ebur128 解析；preflight 會檢查兩者）
Datastore: Supabase Postgres kaf_songs 表 + Storage bucket song-blobs（DDL 見 supabase-schema.sql）
Run: node analyzer.mjs --url "<youtube-url>" [--genre <tag>] [--force] [--dry-run]
Batch: node batch.mjs urls.txt [--sleep <sec>]（無人值守整夜批次，含限流退避）
Test: pnpm test（node --test，自動抓所有 *.test.mjs：src/title.test.mjs、generate-urls.test.mjs、analyzer.test.mjs）

File layout

analyzer.mjs — CLI 進入點（單曲 ingest）
batch.mjs — 依序批次跑 urls.txt
generate-urls.mjs — 用 yt-dlp 掃 YouTube 頻道（@virtual_kaf）產出 urls；過濾例外寫在 KEEP_IDS（強制保留）／DROP_IDS（強制排除），產出的 txt 是 gitignored 純產物，不要手改
lint-titles.mjs — 標題 dry-run 檢查
src/youtube.mjs — yt-dlp wrapper；src/audio.mjs — Meyda spectrum 抽取
src/clock_analysis.mjs — 前端 live FFT 的離線移植版
src/binary_pack.mjs — binary blob 打包；src/db.mjs — Supabase client / upsert
src/title.mjs（+ .test.mjs）— 花譜曲名正規化（前端 A-Z 索引用）
supabase-schema.sql — DB DDL
docs/reference/ — 給前端的交接文件（data-contracts.md＝資料格式權威來源、 clock-analysis-handover.md＝clock 移植的 why + gotchas）

Constraints

docs/reference/data-contracts.md 是資料格式的權威文件：binary layout 或 row shape 變更時，必須在同一個 commit bump BIN_VERSION（src/binary_pack.mjs）或 metadata.schemaVersion（analyzer.mjs）並更新該文件。
src/clock_analysis.mjs 是前端 live FFT 的離線移植版，必須與它所鏡像的前端一起改，否則視覺會漂移。
urls-virtual_kaf.txt 是 gitignored 產物，不手改；過濾例外只改 generate-urls.mjs 的 KEEP_IDS / DROP_IDS。
SUPABASE_SERVICE_KEY 只放在 gitignored 的 .env（名稱見 .env.example），絕不進 git。
Path-type constraints are mirrored in kit.toml [protected]; keep the two in sync (one glob per line).

Conventions

Conventional commits（type(scope): 摘要）；commit 前跑 pnpm test。
以 video_id 對帳，不用標題（頻道掃描拿英文標題、單曲 metadata 拿日文標題）。
GitHub Actions: top-level permissions: contents: read, raised per job only when needed.
GitHub Actions: pin every uses: to a 40-character commit SHA with a trailing # vX.Y.Z comment.

Where to look

docs/DEVELOPMENT.md — 指令、batch 限流、環境變數、輸出、gotchas
docs/reference/data-contracts.md — row shape 與 binary layout（權威）
docs/reference/clock-analysis-handover.md — clock 移植的 why + gotchas
docs/architecture/overview.svg — 系統架構總覽

Install song-analysis and it becomes one more candidate the router can pick - when it fits.

A listing is a routing target with a manifest: what it does, which domains it covers, what it costs and who publishes it. Once installed it sits beside your own models and tools, is scored like any other candidate for each request, and shows up in the trace when it wins. Ratings come from workspaces that installed it, one per account.

{ "ocm": "1", "id": "cholateio-song-analysis-song-analysis", "kind": "skill", "name": "song-analysis", "description": "花譜（KAF）cover 曲的 YouTube ingest + 音訊分析 pipeline：用 yt-dlp 抓音訊 → Meyda 做每幀 spectrum 分析 → 打包成 binary blob 寫入 Supabase （`kaf_songs` 表 + `song-blobs` storage bucket），供 3D 前端視覺化消費。前端消費者為 **kaf-observatory**（本 repo 的 `docs/reference/clock-analysis-handover.md` 對應其同名檔；歷史名 virtual-music-clock）。", "publisher": "cholateio", "version": "1.0.0", "capabilities": { "domains": [ "creative", "coding" ], "tags": [ "skill-md", "github-agents" ], "languages": [ "en" ] }, "quality_prior": 0.6, "examples": [ "花譜（KAF）cover 曲的 YouTube ingest + 音訊分析 pipeline：用 yt-dlp 抓音訊 → Meyda 做每幀 spectrum 分析 → 打包成 binary blob 寫入 Supabase （`kaf_songs` 表 + `song-blobs` storage bucket），供 3D 前端視覺化消費。前端消費者為 **kaf-observatory**（本 repo 的 `docs/reference/clock-analysis-handover.md` 對應其同名檔；歷史名 virtual-music-clock）。" ], "primary": false, "metadata": { "source": { "provider": "github-agents", "repository": "https://github.com/cholateio/song-analysis", "path": "AGENTS.md", "ref": "bbf5656f99f78131b664051943f46a75059ef677", "url": "https://github.com/cholateio/song-analysis/blob/bbf5656f99f78131b664051943f46a75059ef677/AGENTS.md", "key": "cholateio/song-analysis/AGENTS.md" } }, "instructions": "# song-analysis\n\n## Project goal\n\n花譜（KAF）cover 曲的 YouTube ingest + 音訊分析 pipeline：用 yt-dlp 抓音訊 →\nMeyda 做每幀 spectrum 分析 → 打包成 binary blob 寫入 Supabase\n（`kaf_songs` 表 + `song-blobs` storage bucket），供 3D 前端視覺化消費。前端消費者\n為 **kaf-observatory**（本 repo 的 `docs/reference/clock-analysis-handover.md`\n對應其同名檔；歷史名 virtual-music-clock）。\n\n## Stack\n\n- Language: Node.js ≥20（ESM，`\"type\": \"module\"`），pnpm\n- 核心依賴: `meyda`（spectrum 特徵）、`@supabase/supabase-js`、`dotenv`、`minimist`\n- 外部 CLI 依賴（不在 package.json）: **yt-dlp**（`src/youtube.mjs` 經 child_process 呼叫）、**ffmpeg**（LRA 由 `ebur128` 解析；preflight 會檢查兩者）\n- Datastore: Sup", "cost": { "context_tokens": 692 } }

song-analysis

About

song-analysis

Project goal

Stack

File layout

Constraints

Conventions

Where to look

Install song-analysis and it becomes one more candidate the router can pick - when it fits.

What it costs to adopt - in engineering time.

One base URL

No vendor lock-in

Failover built in

Use it

Manifest

Reviews