The Reddit node fetches top posts from a +-separated list of subreddits you define
A code node normalises each item into {url, title, publishedAt} while preserving Reddit-specific fields like ups, upvote_ratio, and num_comments
A deduplication step drops repeats by normalised URL and then by title
A scoring step ranks candidates by ups x upvote_ratio and passes the winner plus the top 10 candidates to GPT-4
A LangChain agent writes three platform-specific drafts in a single call: a short Telegram blurb with a link
Gmail sends you one email containing a draft with APPROVE / DECLINE buttons (sendAndWait, double-confirm)
On APPROVE, a Gate node fans out to three parallel branches that post to Telegram, and each result is appended to its own tab in Google Sheets
On DECLINE, reviewer feedback is injected into the agent and it picks another topic to retry
Set up steps
Setup takes about 15 to 20 minutes
Connect a Reddit OAuth2 credential with read access
Connect an OpenAI API credential (defaults to gpt-4.1-mini; swap to any model you prefer)
Connect a Gmail OAuth2 credential for the approval email and set sendTo to your own address
Connect a Telegram Bot API credential and replace @your_telegram_channel in the Post to Telegram node with your real channel handle (your bot must be an admin of the channel)
Connect a Google Sheets OAuth2 credential and pick your destination spreadsheet + correct tab in each of the four Google Sheets nodes
Edit the default subreddit list inside the Reddit node to match your niche
Optional: adjust the GPT prompt inside the Social Posts Generator to match your tone per platform
Detailed per-step notes live inside the workflow as sticky notes