Yusei Miyakoshi
Workflows by Yusei Miyakoshi
Generate AI summaries from Hacker News to Slack with OpenRouter
## Who's it for This template is for teams that want to stay updated on industry trends, tech news, or competitor mentions without manually browsing news sites. It's ideal for marketing, development, and research teams who use Slack as their central hub for automated, timely information. ## What it does / How it works This workflow runs on a daily schedule (default 9 AM), fetches the top articles from Hacker News for a specific keyword you define (e.g., 'AI'), and uses an AI agent with OpenRouter to generate a concise, 3-bullet point summary in Japanese for each article. The final formatted summary, including the article title, is then posted to a designated Slack channel. The entire process is guided by descriptive sticky notes on the canvas, explaining each configuration step. ## How to set up 1. In the **Configure Your Settings** node, change the default keyword `AI` to your topic of interest and update the `slack_channel` to your target channel name. 2. Click the **OpenRouter Chat Model** node and select your OpenRouter API key from the Credentials dropdown. If you haven't connected it yet, you will need to create a new credential. 3. Click the **Send Summary to Slack** node and connect your Slack account using OAuth2 credentials. 4. (Optional) Adjust the schedule in the **Trigger Daily at 9 AM** node to change how often the workflow runs. 5. Activate the workflow. ## Requirements * An n8n instance (Cloud or self-hosted). * A Slack account and workspace. * An OpenRouter API key stored in your n8n credentials. * If self-hosting, ensure the LangChain nodes are enabled. ## How to customize the workflow * **Change the News Source:** Replace the **Hacker News** node with an **RSS Feed Read** node or another news integration to pull articles from different sources. * **Modify the AI Prompt:** In the **Summarize Article with AI** node, you can edit the system message to change the summary language, length, or tone. * **Use a Different AI Model:** Swap the **OpenRouter** node for an **OpenAI**, **Anthropic**, or any other supported chat model. * **Track Multiple Keywords:** Modify the workflow to loop through a list of keywords in the **Configure Your Settings** node to monitor several topics at once.
Automated meeting minutes from audio using OpenAI transcription and Google Docs
## Who’s it for Teams and operators who record meetings/interviews and want fast, standardized, action-oriented minutes—without writing code. New n8n users welcome. ## What it does / How it works On form submission, the workflow ingests an audio file and basic metadata, transcribes the audio with OpenAI, summarizes it into concise minutes (key points, next actions with owner & due date, and counterpart concerns/requests), then creates a Google Doc (if missing) and appends the minutes. Sticky notes on the canvas explain each step and required inputs. ## How to set up 1. Connect **OpenAI** via Credentials (never hardcode API keys). 2. Connect **Google (OAuth2)** with Docs/Drive scopes. 3. Replace any hardcoded Drive folder ID with an env var (e.g., `MINUTES_FOLDER_ID`). 4. In the Form Trigger, keep the audio upload field and optional metadata (Manager, Partner, Situation). 5. Test with a short audio sample (30–120 sec), then iterate on the summary prompt. Paste this description into a yellow sticky note at the top of the canvas. ## Requirements * n8n (Cloud or self-hosted) * OpenAI credential with transcription + LLM access * Google account with Docs/Drive access and available storage ## How to customize the workflow * Tweak the prompt tone, length, or sections (e.g., add “Risks/Blockers”). * Change the document title pattern or parent folder. * Add Slack or email notifications after the Doc is updated. * Log executions to Google Sheets for weekly reporting and audits. ## Security & publishing tips remove personal IDs (folder IDs, real emails) before sharing; keep node names descriptive; retain sticky notes for setup guidance.
Automated email digest from Gmail to Slack with GPT-4o summary
## Who’s it for Teams that start their day in Slack and want a concise, automated summary of yesterday’s emails—ops leads, PMs, founders, and anyone handling busy inboxes without writing code. ## What it does / How it works Runs every morning at **08:00** (cron `0 0 8 * * *`), fetches all emails received **yesterday**, and routes the result: if none were found, it posts a polite “no emails” notice; if emails exist, it aggregates them and asks an AI agent to produce a structured digest, then formats and posts to your chosen Slack channel. The flow uses **Gmail → If → Aggregate (Item Lists) → AI Agent (OpenRouter model with structured output) → Code (Slack formatter) → Slack**. A set of sticky notes on the canvas explains each step and required inputs. ## How to set up 1. Connect **Gmail (OAuth2)** and keep the default date window (yesterday → today at 00:00). 2. Connect **Slack (OAuth2)** and select your target channel. 3. Add **OpenRouter** credentials and pick a compact model (e.g., `gpt-4o-mini`). 4. Keep the provided structured-output schema and formatter code. 5. Adjust the schedule/timezone if needed (the fallback message includes an Asia/Tokyo example). 6. Paste this description into the **yellow sticky note** at the top of the canvas. ## Requirements * Gmail & Slack accounts with appropriate scopes * OpenRouter API key stored in **Credentials** (no hard-coded keys) * n8n Cloud or self-host with LangChain agent nodes enabled ## How to customize the workflow * Narrow Gmail results with label/search filters (e.g., `from:`, `subject:`). * Change the digest sections or tone in the **AI Agent** system prompt. * Swap the model for cost/quality needs and tweak temperature/max tokens. * Localize dates/timezones in the formatter code and Slack messages. * Branch the output to email, Google Docs, or Sheets for archival. ## Security & publishing tips Rename all nodes clearly, **do not** hardcode API keys, remove real channel IDs/emails before sharing, and group end-user variables in a **Set (Fields)** node. Keep the sticky notes—they’re mandatory for reviewers.