Pace：我打造了開源 AI，處理我一直拖延的日程規劃

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 3: Slow and Steady

What I Built

I built Pace, an AI day coordinator for turning intentions into time you can protect.

A calendar holds fixed commitments. A task list holds intentions. Pace connects the two.

Write what needs to happen in plain language:

“I have an exam tomorrow morning. Find time to study.”

Pace reads the commitments you connected, understands the request, finds a realistic opening, and prepares a focus-block proposal.

You review the proposal before Pace writes anything to your calendar.

Pace supports the planning work that often gets pushed aside: checking a crowded calendar, finding time for meaningful work, and turning a vague intention into one clear next step.

Make a plan you can actually follow.

Demo

Watch the Pace demo:

Watch on YouTube →

Try Pace or deploy your own private workspace:

pace-hub.onrender.com

The demo shows this flow:

1. Connect Google Calendar.
2. Tell Pace what needs to happen.
3. Let Pace read the shape of the day.
4. Review a proposed focus block.
5. Approve the plan.

It also shows Pace using Gemma locally on an Arduino Uno Q, with LED signals for ready, planning, and attention states.

Code

Pace is open source:

Pace

Pace turns an intention into a calendar block you can follow through on.

Write what needs to happen in plain language. Pace reads the commitments you connected, asks your chosen model to interpret the request, finds a valid opening, and shows you a proposal. Nothing is written to your calendar until you approve it.

Pace has two pages:

• / is the product page.
• /app is your private workspace.

Use Pace from scratch

1. Open /app and go to Connections.
2. Connect Google Calendar, then select Sync today.
3. Go to Settings and choose an AI model: Arduino Uno Q, a local OpenAI-compatible model server, or a hosted compatible model.
4. Return to Today and select Tell Pace.
5. Write a request such as: I have an exam tomorrow morning. Find two hours to study.
6. Review Pace's proposed time. Select Use this plan only when it looks right.
7. Open Devices to…

Repository: github.com/abdushakurob/pace-ai

The repository includes the Pace Hub, Google Calendar connection, local-model support, Arduino Uno Q app, LED status signals, Docker configuration, and a Render Blueprint.

How I Built It

Pace uses a Python Hub, a lightweight web interface, and SQLite storage.

Google Calendar provides the existing commitments for the day. Pace uses that context to find an available time for a new focus block.

The AI layer converts a natural-language request into planning details: the task, likely duration, preferred time, and a reason for the suggested slot.

For local AI, I used Gemma 3 1B through llama.cpp on an Arduino Uno Q.

The Uno Q receives the request and calendar context, runs the local model, and returns planning information to Pace. Its LED matrix gives the system a physical state:

• Ready when Pace is available
• Moving while Pace is planning
• Alert when Pace needs attention

Pace also supports OpenAI-compatible local model servers such as Ollama, llama.cpp, and LM Studio, plus compatible hosted model endpoints.

For deployment, I added a Dockerfile and Render Blueprint. Each person can deploy a private Pace workspace with persistent storage for settings and calendar connection state.

Why Does Open Innovation Matter?

Pace works with personal context: a calendar, deadlines, study plans, and the work competing for time in a day.

Open-weight models let the planning layer run close to the person using it. Pace can run Gemma locally on an Arduino Uno Q, on a computer, or through another compatible model server.

People can inspect the system, choose their model, host their own workspace, and control their calendar connection.

Open innovation gives Pace room to adapt:

• Run a model locally
• Swap models as new options emerge
• Deploy a private workspace
• Keep calendar access under the user’s control
• Use small, accessible hardware such as Arduino Uno Q

Google Calendar provides connected calendar data. Gemma provides an open local planning path. Together, they make Pace useful while keeping model choice in the user’s hands.

My Agent Session

Pace developed through an iterative build focused on calendar-aware planning, local Gemma inference on Uno Q, Google OAuth, proposal approval, device pairing, LED feedback, a public landing page, and Render deployment.

The complete implementation is available here:

github.com/abdushakurob/pace-ai

The Lazy Payoff

Pace removes the repeated planning loop behind a simple thought:

“I should do this today.”

The person shares the intention once. Pace checks the day, finds a possible opening, and prepares a plan for review.

That leaves more energy for the work itself: studying, building, exercising, resting, and following through.

Pace handles the scheduling effort. The person keeps the decision.

Prize Categories

I am entering Pace for:

• Hacktoberfest Open-Source AI Challenge: Slow and Steady
• Best Use of Arduino — Gemma planning on Arduino Uno Q with physical LED status signals
• Best Use of Gemma — Gemma 3 1B powers Pace’s local planning path
• Best Use of Render — Render hosts the Pace product page and enables private Pace deployments