$ whoami

Hi, I'm Christian John
but everybody calls me potaterrr 🥔

Python automation developer growing pipelines that connect data to Make.com and n8n — from job scrapers to trading-signal engines to AI agents that write follow-up emails for me.

If it's repetitive, I'll automate it. If it involves potatoes, even better.

# Featured Projects

Dead Lead Follow-up

An AI-powered, human-in-the-loop automation that resurrects dormant leads: an LLM writes personalized check-in emails, files them as Gmail drafts (never auto-sent!), drops a booking button into every one, and tracks each lead through the review lifecycle on ClickUp.

  • Human-in-the-loop by design — AI drafts, human approves, nothing sends itself
  • Anti-hallucination guardrails: the only link in any email is mine
  • Full setup guide, config reference and production-tested lessons in the README
  • ▶ Watch it run below · repo ↗

AI Voice Receptionist

An AI receptionist that answers missed calls and books appointments — built three times (Make / n8n / Zapier) on a shared Vapi + Twilio voice layer, so platforms can be compared apples-to-apples before choosing one.

  • Answers instantly, capturing name, number, intent and preferred slot as structured data
  • Conflict-checks Google Calendar, books real slots, alerts the owner with a summary
  • Simulated end-to-end today — live phone calls are a plug-in away
  • ▶ Watch a call flow run · repo ↗

# Other Projects

job-scrapper ↗

A Python scraper that searches OnlineJobs.ph for remote jobs matching automation keywords (n8n, make.com, zapier) and pushes each listing as JSON to a Make.com webhook for downstream automation.

  • Extracts title, salary, employment type, and full job description per listing
  • Keyword dedupe + polite, human-like request delays
  • Fallback notifications so downstream automations never stall

signal-checker ↗

An automated trading-signal pipeline that evaluates an SMA crossover strategy on daily stock prices and dispatches BUY / SELL / HOLD signals as JSON to a Make.com webhook.

  • ~400 days of historical closes via yfinance (50/200-day defaults)
  • Fully configurable through environment variables (symbol, SMA windows, webhook URL)
  • Runs automatically on GitHub Actions every weekday at 21:00 UTC

ai-folklore ↗ (theory)

The research stage of a folklore-story video pipeline: research → AI script → (Veo video) → publish. A zero-dependency Python scraper pulls folklore stories (Philippine mythology front-loaded) from Wikipedia and feeds them to Make.com or n8n — Gemini writes the script, a Gmail draft waits for review.

  • Zero pip dependencies — stdlib urllib + json, Python 3.9+ is all it needs
  • Built-in dedupe remembers delivered titles, so each run pushes only new stories
  • Ships with importable Make.com and n8n blueprints plus a full testing report

rss-news ↗ (personal automation)

An automated news curation pipeline for the AI Tech & Automation Hub Facebook page: a Python script fetches tech & AI RSS feeds (Hacker News, Phoronix, The Verge…), filters by keyword, and POSTs a digest to a Make.com webhook that drafts the post with Gemini, generates a graphic with Imagen, and publishes to Facebook Pages.

  • GitHub Actions cron runs it every Sunday at 6:00 AM PHT
  • Keyword filtering + seen_articles.json dedupe so nothing posts twice
  • A random banner image from the repo rides along with every digest

make-commits ↗

A Make.com automation that posts every GitHub push to a Facebook Page as a digest — instantly, via webhooks. Built and running entirely on the Make.com free tier.

  • Instant pipeline: git push → Facebook post lands seconds later (webhooks, no polling)
  • One post per push listing every commit message as a bullet (~2 ops per push)
  • Ships with a sanitized scenario blueprint + step-by-step setup guide

# Setup

🥔 Want to know what potato I'm running?

potaterrr@linux: ~

Works on Debian/Ubuntu, Fedora, Arch, openSUSE, Alpine and Void — see what it installs →

# Skills

Python automation

Scrapers, schedulers and pipelines that run unattended and fail loudly instead of silently.

n8n workflows

Self-hosted orchestration with clean logic, error branches and modular design.

Make.com scenarios

Connecting apps and data with minimized API calls and zero redundant steps.

Webhooks & APIs

Event-driven glue between systems — Postman and dev consoles are my stethoscope.

LLM pipelines

OpenRouter-powered agents with guardrails, structured output and human checkpoints.

CI/CD & ops

GitHub Actions cron jobs, env-driven config, monitoring and fallback alerts.

# How It Works

Pick a featured build, hit run, and step through its pipeline stage by stage. Everything is simulated right here in your browser — no APIs, no keys, no emails harmed. 🥔

📋 Tonight's dead lead — edit me
Current StageSchedule Trigger
Live Packets15
Statushealthy
⏰ Schedule Trigger n8n · scheduleTrigger

The pipeline wakes up every morning at 10:00 — nobody clicks anything.

Pipeline velocity
1.0x
Failure drill

Ready — hit “Run the pipeline”.

Same nodes as production, zero side effects: the real workflow only ever drafts emails (never sends) and runs calls in simulation. Want one of your own? →

# Timeline

  1. Dotfiles go public with a one-line installer ↗

    The full potato setup — Bash, Neovim, Starship, Fastfetch, Hyprland, Waybar, Wofi and GTK — now installable on Debian, Fedora, Arch, openSUSE, Alpine and Void with a single command: curl -fsSL https://potaterrr.github.io/install | sh.

  2. ai-folklore released ↗

    The folklore pipeline lives again — as free software: a zero-dependency Wikipedia folklore scraper feeding Make.com or n8n, with importable blueprints, built-in dedupe and a full testing report.

  3. rss-news goes live ↗

    Weekly AI & tech digest automation: Python fetches Hacker News, Phoronix, The Verge and friends, filters by keyword, and ships the digest to Make.com for Gemini drafting, Imagen graphics and Facebook Pages publishing.

  4. Signal Checker Gains Forex Support

    The SMA crossover signal-checker now handles forex pairs like USD/PHP and EURUSD with automatic yfinance format conversion. Users can specify symbols intuitively (e.g., "USD/PHP") while the tool internally converts to yfinance format (USDPHP=X) for accurate data retrieval. Existing stock functionality remains unchanged.

  5. Folklore AI video pipeline: research live, publishing blocked

    Weekly folklore topic research is running on Make.com, but the generate & publish stage is stuck on Gemini API free-tier quota (429) for video generation. Fully working text pipeline; video awaits a plan decision.

  7. Launched potaterrr.github.io ↗

    This very site: a dark dev-theme portfolio shipped as a single static page and deployed to GitHub Pages automatically through GitHub Actions.

  8. Dead Lead Follow-up released ↗

    AI-powered lead resurrection goes public: an OpenRouter LLM drafts personalized check-ins, files them as Gmail drafts for human approval, and tracks every lead through ClickUp.

  9. job-scrapper hardened ↗

    v2 pass: real salary and employment-type extraction, full descriptions instead of teasers, bot-check retries, and one-job-per-keyword queue discipline for Make.com.

  10. signal-checker goes live ↗

    SMA crossover engine over ~400 days of yfinance closes pushing BUY / SELL / HOLD signals to a Make.com webhook every weekday at 21:00 UTC.

  12. potaterrr joins GitHub ↗

    Profile repo created and the automation journey begins. The rule is simple: if it's repetitive, it gets automated. 🥔

# Blog

5 min

My dotfiles, one line

Bash, Neovim, Hyprland and friends — installed and stowed on six distro families by a single curl | sh. How the bootstrap and GNU Stow make it safe.

7 min

One receptionist, three engines

I built the same AI phone receptionist on Make, n8n and Zapier. Here's what broke, what shone, and which engine to pick for whom.

8 min

Why my AI never sends emails

Human-in-the-loop AI outreach: the LLM drafts into Gmail, a human approves, and guardrails keep hallucinations out of your inbox.

7 min

Feeding Make.com from Python

Flat JSON payloads, polite scraping and queue discipline: how job-scrapper keeps its downstream automations cheap and quiet.

6 min

The free-plan ops math

Polling every 15 minutes burns 2,880 ops/month before anything happens. The arithmetic that should decide every trigger choice.

# Contact

You can reach me anywhere below.

✈️ Telegram @portaterrrr preferred 💬 WhatsApp @potaiterrr chat anytime

Open to discussing new technologies, automation strategies, or freelance opportunities. Recently enhanced with HIPAA-compliant voice receptionists and forex-capable signal checkers. If your leads are gathering dust, I know a workflow for that. 🥔

GOT QUESTIONS? LET'S GET FARMING 🥔