---
title: "n8n-x-autopost: DeepSeek & Telegram Social Auto-Publisher"
description: Open-source n8n workflow combining topic feeds, DeepSeek A/B draft generation, Telegram inline-button review, and Buffer GraphQL multi-platform publishing.
image: "https://zengliangyi.com/_astro/n8n-x-autopost.DW1Pc8d5_ZUu39G.webp"
---

## The problem: unattended auto-posters sound robotic, while manual curation burns hours

Staying active across social platforms as a developer or indie builder (X / Twitter, Threads, LinkedIn, Bluesky) rarely stalls on clicking “Post”—it stalls on **scanning feeds, filtering out noise, drafting sharp angles, and preparing preview visuals**. Yet fully unattended auto-posters quickly degrade an account’s voice with generic AI filler, hallucinated context, or ban-inducing browser-cookie automation.

**n8n-x-autopost** takes a **human-in-the-loop (80% automated preparation + 20% human editorial judgment)** approach. An n8n workflow rotates across three high-signal topic pools, deduplicates against a 300-item sliding window, generates two distinct A/B drafts with DeepSeek alongside a 1200×630 preview card, and delivers an interactive review message to Telegram. A single tap on an inline button publishes immediately—or schedules to a queue—across multiple social networks through Buffer’s official GraphQL API.

> Why Human-in-the-Loop Instead of Unattended Auto-Posting?

> Automating feed curation, deduplication, A/B drafting, and image rendering removes over 80% of daily friction, while keeping a 5-second Telegram approval gate ensures every published post reflects a deliberate editorial choice.

## Core workflow capabilities

### 1. Three rotating topic pools & deduplication

Triggered on a cron schedule (`08:15`, `11:15`, `20:15` daily) or on demand via Telegram commands (`/biz`, `/tech`, `/trends`), rotating across Indie Business (`Show HN` via Algolia), Tech (HN front page), and US Daily Trends (Google Trends RSS) with a 300-item static-data deduplication window.

### 2. DeepSeek A/B drafting & automatic visuals

Generates two contrasting narrative angles (Draft A and Draft B) per story with mobile-friendly short-line spacing and anti-cliché prompt constraints, paired with a 1200×630 preview image (GitHub OpenGraph social card or 11ty page screenshot).

### 3. Single-user Telegram gate & inline controls

Enforces a strict `YOUR_TELEGRAM_CHAT_ID` allowlist across all three entry points. Each review card includes inline buttons to publish A or B with an image, publish text-only, add to the Buffer queue, switch topic pools, or skip.

### 4. Multi-platform fanout via Buffer GraphQL

Uses Buffer’s official GraphQL API (`createPost` mutation by `channelId`) instead of brittle headless browsers, supporting X, Threads, LinkedIn, Bluesky, Mastodon, and parallel multi-channel publishing.

## End-to-end pipeline architecture

The entire pipeline ships as a single self-contained n8n workflow JSON organized into four stages:

| Pipeline stage | Automated responsibility | Key engineering constraint |

| --- | --- | --- |

| 1. Rotating feeds & dedup | Polls `Show HN`, HN front page, and Google Trends RSS | Sliding 300-item static-data history prevents duplicate pitches |

| 2. DeepSeek A/B rewrite & image | Produces Draft A (insight hook) + Draft B (builder breakdown) + 1200×630 image | Prompt rules enforce short lines and strip generic AI filler |

| 3. Telegram inline review gate | Delivers preview card with publish, text-only, queue, pool-switch, and skip buttons | Enforces `YOUR_TELEGRAM_CHAT_ID` allowlist across all 3 trigger nodes |

| 4. Buffer GraphQL fanout | Calls Buffer `createPost` and edits the Telegram message in place | Turns Telegram chat history into an auditable publishing log |

## Four-step editorial loop

- Scheduled push or on-demand pull (`/biz`, `/tech`, `/trends`) : In addition to three daily scheduled runs, sending `/biz` (Show HN indie products), `/tech` (top Hacker News discussions), or `/trends` (Google Trends US daily stories) in Telegram immediately fetches the next unseen candidate from that pool.

- Context extraction, A/B rewriting & preview card : n8n extracts the source title, URL, and discussion context, prompts DeepSeek to write two contrasting drafts (Draft A and Draft B), attaches a 1200×630 social card or webpage screenshot, and sends the package to Telegram.

- One-tap decision in Telegram : Review both drafts on mobile and tap an inline button: `🚀 Publish A/B` (with image), `📝 Text-Only A/B`, `📅 Queue A/B` (push to Buffer’s scheduled queue), `🔄 Biz/Tech/Trends` (switch topic pool immediately), or `❌ Skip`.

- In-place message audit trail & multi-channel delivery : Once a button is tapped, n8n executes the Buffer `createPost` mutation and edits the original Telegram message in place with the final status (`✅ Published`, `📅 Queued`, or `❌ Skipped`).

![Screenshot of the Telegram review card showing the source headline, DeepSeek A/B draft options, and the inline action button matrix](/_astro/n8n-telegram-review-card.CuwWDJKc_dTKT2.webp)

Live Telegram review card displaying the source link, Draft A and Draft B copy, and the inline

button matrix for instant publishing, queuing, or switching topic pools.

## Sanitized, ready-to-import blueprint

To make the workflow easy to adopt on any self-hosted or cloud n8n instance:

- Zero hardcoded secrets : All Telegram chat IDs, Buffer channel IDs, and API credentials are parameterized with clean placeholders alongside bilingual step-by-step setup documentation.

- Composable multi-channel fanout : Because Buffer’s GraphQL endpoint routes posts by `channelId`, duplicating the publish node lets a single Telegram approval fan out to X, Threads, LinkedIn, and Bluesky simultaneously.

For the importable `workflow.json`, Buffer channel discovery queries, and setup instructions, visit the [n8n-x-autopost GitHub repository](https://github.com/ZengLiangYi/n8n-x-autopost).

##  More projects

[2026

### ChatCrystal

Local-first AI coding knowledge base syncing Claude Code, Cursor, Codex, and OpenCode sessions into structured notes, semantic search, and MCP memory.

#TypeScript

#AI

#MCP](/en/projects/chatcrystal/)[2026

### KeelDesk

Windows 11 desktop MCP runtime built with Tauri v2 and Rust: embedded Streamable HTTP server, tray daemon, and human approval for Claude Code and Cursor.

#Tauri

#Rust

#TypeScript

#MCP](/en/projects/keeldesk/)

##  Have a question or an idea?

Send me a message to discuss an implementation or explore a project idea.

[Contact me](/en/contact/) [All projects](/en/projects/)

```json
{
  "@context": "https://schema.org",
  "@id": "https://zengliangyi.com/#website",
  "@type": "WebSite",
  "name": "ZengLiangYi Blog",
  "alternateName": [
    "Zeng Liangyi",
    "ZengLiangYi",
    "zengliangyi.com"
  ],
  "url": "https://zengliangyi.com",
  "description": "Frontend engineering, AI coding and open-source practice",
  "inLanguage": "en",
  "author": {
    "@id": "https://zengliangyi.com/#person"
  }
}
{
  "@context": "https://schema.org",
  "@type": "SoftwareApplication",
  "name": "n8n-x-autopost",
  "description": "Open-source n8n workflow combining topic feeds, DeepSeek A/B draft generation, Telegram inline-button review, and Buffer GraphQL multi-platform publishing.",
  "url": "https://zengliangyi.com/en/projects/n8n-x-autopost/",
  "applicationCategory": "DeveloperApplication",
  "operatingSystem": "Any",
  "offers": {
    "@type": "Offer",
    "price": "0",
    "priceCurrency": "USD"
  },
  "codeRepository": "https://github.com/ZengLiangYi/n8n-x-autopost",
  "image": "https://zengliangyi.com/_astro/n8n-x-autopost.DW1Pc8d5_ZUu39G.webp",
  "datePublished": "2026",
  "dateModified": "2026-11-15T00:00:00.000Z",
  "inLanguage": "en",
  "keywords": [
    "n8n",
    "DeepSeek",
    "Telegram",
    "Workflow"
  ],
  "isPartOf": {
    "@id": "https://zengliangyi.com/#website"
  },
  "author": {
    "@id": "https://zengliangyi.com/#person",
    "@type": "Person",
    "name": "曾亮懿",
    "alternateName": [
      "Leon Zeng",
      "Zeng Liangyi",
      "ZengLiangYi",
      "夜雪闻竹"
    ],
    "jobTitle": "AI Native Developer",
    "knowsAbout": [
      "AI Native Development",
      "AI Agents",
      "Model Context Protocol",
      "TypeScript",
      "React",
      "Full-Stack Development",
      "Developer Tools"
    ],
    "url": "https://zengliangyi.com",
    "email": "contact@zengliangyi.com",
    "image": "https://zengliangyi.com/branding/zeng-liangyi-avatar.png",
    "address": {
      "@type": "PostalAddress",
      "addressLocality": "深圳",
      "addressRegion": "广东",
      "addressCountry": "CN"
    },
    "sameAs": [
      "https://github.com/ZengLiangYi",
      "https://x.com/yizel0501",
      "https://bsky.app/profile/yizel9.bsky.social",
      "https://juejin.cn/user/1669474445171547",
      "https://blog.csdn.net/2201_75708499",
      "https://www.linkedin.com/in/zeng-liangyi/",
      "https://gravatar.com/radangelb905e883e0"
    ]
  }
}
{
  "@context": "https://schema.org",
  "@type": "BreadcrumbList",
  "name": "Breadcrumb",
  "itemListElement": [
    {
      "@type": "ListItem",
      "position": 1,
      "name": "Home",
      "item": "https://zengliangyi.com/en/"
    },
    {
      "@type": "ListItem",
      "position": 2,
      "name": "Projects",
      "item": "https://zengliangyi.com/en/projects/"
    },
    {
      "@type": "ListItem",
      "position": 3,
      "name": "n8n-x-autopost",
      "item": "https://zengliangyi.com/en/projects/n8n-x-autopost/"
    }
  ]
}
```
