Hi everyone,
We run a small automation studio and keep exporting the same building blocks from project to project. We cleaned eleven of them up and published them under MIT — maybe they save someone a day of tinkering.
Yandex 360 without native nodes (n8n has none, so these speak the protocols through plain HTTP Request nodes):
- Yandex Calendar — create an event / list events (CalDAV: PUT an iCal file, REPORT a calendar-query and parse the VEVENTs);
- Yandex Contacts — read the address book (CardDAV: PROPFIND, multiget REPORT, vCard parsing);
- Yandex Disk — upload a file (WebDAV, base64 in).
Lead intake
ai-lead-intake— one webhook for site, Telegram, MAX, VK and e-mail: normalizes the payload, reads attachments with an LLM, classifies the request (category, priority, needs a human), builds a CRM card and pings the manager;site-lead-to-sheets-telegram— the classic “form → Google Sheets → Telegram”;website-audit-form— public “audit my site” form with validation and a stored request.
AI assistants and media
rag-support-assistant— answers from a knowledge base through any OpenAI-compatible endpoint;voice-assistant-knowledge-base— the “brain” behind a phone assistant: takes the recognized phrase, scores it against a topic table, answers from the KB and asks for a human when it keeps failing;ai-video-generation-proxyapi— create → poll → download → Yandex Disk → Telegram.
Reporting
daily-deals-report— scheduled CRM summary to Telegram, no BI tool involved.
Two things we learned the hard way and baked into every template:
- A webhook that spends money (LLM call, SMS, video generation) checks a shared-secret header in an
onlyRunIfcondition before doing any work — otherwise a public endpoint quietly burns your API budget. - Credentials are stripped from the exported JSON and live webhook paths are renamed, so importing a template can never call someone else’s production URLs. Every README has a
curlexample and the list of credentials to create.
Repo (one folder per template, workflow.json + README): GitHub - VavilkinAlex/n8n-bowhard-templates: Production-tested n8n workflow templates: AI lead intake, voice assistant, Yandex Disk/Calendar/Contacts (WebDAV/CalDAV/CardDAV) and reporting. By Bowhard (bowhard.ru). · GitHub
Some of these are voice/transcript adjacent, so we also built two small HTTP services our own workflows call: speech-to-text (mp3/wav/ogg → text, SRT or word-level JSON) and text-to-speech. The API is documented on the page: Расшифровка аудио и видео в текст — Bowhard: текст и субтитры SRT
Happy to answer questions about the Yandex part — the CalDAV/CardDAV quirks are the least documented bit of that ecosystem.