用你的 AI 助理回 Instagram 留言 — 程式只抓資料和送出,回覆內容由 AI 在對話裡生成,零付費 AI API。 Reply to Instagram comments with your AI assistant. The program only fetches & sends; the AI writes the replies in chat. No paid AI API.
繁中 ｜ English
一次送出多語言回覆 — 留言 ID 與粉絲暱稱已塗銷 ｜ multilingual replies in one run (comment IDs & handles redacted)
| 💸 零 AI 費用 | 不在程式裡呼叫任何付費 AI 介面。回覆由你的 AI 助理(如 Claude)在對話裡生成,量再大也不燒 API 錢。 |
| 🌍 跟著留言者的語言回 | 中／英／西／葡／義／日／土／德／印尼／馬來／波斯… 留言者用什麼語言,就用什麼語言回。 |
| 🙂 純表情留言也照顧 | 只有 emoji 的留言會被獨立標出,可用 username 風格推測語言,判不準就走英語安全牌。 |
| 🤝 真愛粉熟客語氣 | 同一人多次留言時,每則回覆內容/emoji/CTA 都不同,像主理人真心一條條看完。 |
| 🛡️ 防機器人節流 | 每則之間隨機等 30~90 秒,降低被平台判定為自動化的風險。 |
| ✅ 自動去重 | replied.json 記錄已回過的留言,絕不重複回同一則。 |
| 🔑 一鍵拿長期 Token | oauth.js 走 Facebook Login,自動換 60 天長期 Page Token 並寫回 .env。 |
在 IG 上經營,常會遇到各國語言的粉絲。雖然我們看得懂翻譯,但要一則一則把留言複製去翻譯、再切回來回覆,實在太麻煩 —— 尤其外語粉絲一多,光是判斷「這則該用哪種語言回」就很頭痛。
這套工具就是為了解決這件事:
- 🌐 自動判斷該用哪種語言回每一則留言。
- 🙂 就算對方只丟表情符號、沒有文字,也能從帳號名稱推測語言;真的判斷不出來,就統一用英文這張安全牌,絕不亂猜回錯語言。
- 🤝 還會記住常來留言的老粉,給他們更熱情、更熟客的回應 —— 讓路人更容易變成鐵粉,對貼文的觸及與互動也大有幫助。
而在「怎麼生成回覆」這件事上,大多數「自動回留言」工具把付費 AI 寫死在程式裡 — 量大燒錢、語氣難客製。 本工具把流程拆成「程式做苦工、AI 動腦」三段:
① node fetch.js <貼文ID> → pending.json(該篇待回留言清單)
② 你的 AI 助理讀 pending.json,在對話裡逐則生成回覆 → replies.json
③ node reply.js → 逐則送到 IG(每則間隔 30~90 秒,自動記錄已回)
語氣、語言、CTA 全由你和 AI 在對話裡掌控;程式端不碰 AI、零成本。
需要 Node.js 18+。
npm install
cp .env.example .env # Windows: copy .env.example .env
# 在 .env 填入 FB_APP_ID 與 FB_APP_SECRET(到 https://developers.facebook.com/apps 建立 App 取得)
node oauth.js # 跟著瀏覽器流程走完,自動把 Page Token 寫回 .env
node test.js # 列出最近貼文,複製想處理的 media_id
node fetch.js <media_id> # 抓該篇新留言 → pending.json
# → 讓你的 AI 助理讀 pending.json,生成 replies.json
node reply.js # 逐則送出IG 帳號需為商業帳號並連結到一個 Facebook 粉專。完整逐步說明見
SKILL.md。
| 檔案 | 角色 |
|---|---|
oauth.js |
一次性:走 FB Login 拿 60 天 Page Token,寫回 .env |
test.js |
測試憑證、列出最近貼文與其 media_id |
fetch.js |
抓某篇貼文的新留言,過濾已回過的 → pending.json |
reply.js |
讀 replies.json 逐則送出,間隔 30~90 秒,記錄到 replied.json |
.env.example |
憑證範本(複製成 .env 再填) |
- 絕對不要提交
.env— 內含可操控你帳號的權杖(已被.gitignore擋住)。 pending.json/replies.json/replied.json含留言者帳號與留言內容(第三方個資),同樣已被.gitignore擋住,請勿公開。- 請遵守 Instagram 平台政策,自動化留言頻率請節制。
這個工具對你有幫助的話,歡迎來我的粉專逛逛 🐱 👉 Facebook：https://www.facebook.com/VavaAiArtist
A helper for replying to Instagram comments. The code only fetches comments and sends replies — deciding what to say is handed to your AI assistant (e.g. Claude) in a chat. No paid AI API calls.
- 💸 Zero AI cost — no paid AI endpoint baked in; your assistant drafts replies in chat.
- 🌍 Replies in the commenter's own language (EN/ES/PT/IT/JA/TR/DE/ID/MS/FA…).
- 🙂 Emoji-only comments handled — language guessed from username, English as the safe fallback.
- 🤝 Loyal-fan voice — repeat commenters get distinct content/emoji/CTA each time.
- 🛡️ Anti-bot throttle — random 30–90s gap between replies.
- ✅ De-dupe —
replied.jsonensures no comment is answered twice. - 🔑 One-shot OAuth —
oauth.jsgrabs a 60-day long-lived Page token.
Running an IG account means constantly meeting fans who speak different languages. Even if you can read a translation, copying each comment into a translator and back just to reply is a real pain — and the bigger your international audience, the harder it gets to even decide which language to answer in.
That's exactly what this tool solves:
- 🌐 It automatically decides which language to reply in for every comment.
- 🙂 Even when someone leaves only emojis, it guesses their language from their username — and when it truly can't tell, it safely falls back to English instead of guessing wrong.
- 🤝 It also remembers your regular commenters and gives them warmer, familiar replies — turning passers-by into loyal fans and boosting reach & engagement along the way.
As for writing the replies, most auto-reply tools bake a paid AI endpoint into the program (expensive at scale, hard to fine-tune). This one splits the work instead:
① node fetch.js <postID> → pending.json (comments to answer)
② your AI assistant reads pending.json and drafts replies → replies.json
③ node reply.js → posts each reply (30–90s apart, auto-recorded)
Requires Node.js 18+.
npm install
cp .env.example .env
# put FB_APP_ID & FB_APP_SECRET in .env (create an app at https://developers.facebook.com/apps)
node oauth.js # complete the browser flow; writes the Page token to .env
node test.js # list recent posts, copy a media_id
node fetch.js <media_id> # fetch new comments → pending.json
# → let your AI assistant read pending.json and write replies.json
node reply.js # send them outYour IG account must be a Business account linked to a Facebook Page. Full walkthrough in
SKILL.md.
- Never commit
.env(account tokens — already git-ignored). pending.json/replies.json/replied.jsonhold commenters' handles and text (third-party PII) — also git-ignored; don't publish them.- Follow Instagram's platform policies; keep automation rates reasonable.
If this helped you, come say hi on my Facebook page 🐱 → https://www.facebook.com/VavaAiArtist