How long a meeting can it handle?

The author says recordings of about an hour were handled comfortably, and longer recordings sometimes cut off or returned empty text when too many tokens went to thinking.

What happens if I close the page while it is processing?

Once processing has started, the document still gets the notes even if the page is closed, because the work is queued on Cloud Tasks.

What tools does the app use?

It runs on Node.js 20 with Express 4 on Google Cloud Run, uses Cloud Storage and Firestore, and relies on the Gemini API and Google Apps Script, all packaged in Docker.

Large Language ModelsAI Coding AssistantsZenn AI/ML2d agoOct 13, 2026, 22:00 JST

Developer builds minutes app on Gemini API and Google Apps Script

Send on LINE

3 Key Points

What happened
The app records audio in the browser, stores it in Cloud Storage, and has Gemini API's gemini-2.5-flash transcribe it and draft minutes, which Google Apps Script appends to a Google Doc table.
Why it matters
After shifting tokens away from thinking and toward output, recordings of up to about an hour were handled without much quality loss, suggesting light summarization tasks may not need heavy reasoning budgets.

Not sure about something? Ask the AI

Questions and answers are published on this page.

Summaries like this, in your inbox every morning.

Get news like this every morning, free

Context & Analysis

The author's motivation was ordinary: running a weekly operations meeting for a mahjong-themed social media account with two classmates, taking notes in Google Docs, and finding the note-taking tedious. Building the app doubled as a way to study APIs, cloud services, and application work.

The stack is notable for how much of the glue comes from Google: Cloud Run for the container, Cloud Storage for audio, Firestore for job state, Cloud Tasks to queue the work with OIDC authentication, and Apps Script to append a row to a table inside a Google Doc. The author was candid that JavaScript was beyond their existing knowledge, so they leaned on AI to explain and then implement.

The most practical finding concerns token budgets on the free tier. Early versions sent too many tokens to thinking, so recordings over 30 minutes were cut off or returned empty text; shifting tokens toward output preserved quality while handling longer recordings. The author concludes that for fixed-format tasks like summarization, thinking can be cut back almost entirely.

FAQ

How long a meeting can it handle?: The author says recordings of about an hour were handled comfortably, and longer recordings sometimes cut off or returned empty text when too many tokens went to thinking.
What happens if I close the page while it is processing?: Once processing has started, the document still gets the notes even if the page is closed, because the work is queued on Cloud Tasks.
What tools does the app use?: It runs on Node.js 20 with Express 4 on Google Cloud Run, uses Cloud Storage and Firestore, and relies on the Gemini API and Google Apps Script, all packaged in Docker.

Zenn AI/MLRead Original Article

Send on LINE

AI news that matters for your work, delivered every morning.

Pick your industry and the AI tools you use, and get news related to your work every day.

Get it free

Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.