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

What I Built

Every developer knows the dreaded "Morning Friction" penalty.
You finish a late-night hacking session, close your laptop without writing documentation, and go to sleep. The next morning, you spend 25 minutes staring at your terminal trying to reconstruct your train of thought:

• What branch was I on?
• Why did I leave this git diff uncommitted?
• What was that StackOverflow tab I had open at 1:00 AM?
• What broken test was I about to fix? Writing daily handoff notes is the responsible thing to do, but at the end of the day, you're exhausted and want to be a sloth. ReMind is a local-by-default, privacy-preserving developer resumption daemon and dashboard. It allows you to be totally lazy: close your laptop without writing a single line of notes. In the background, ReMind automatically aggregates your day's work across:
• Multi-Project Git Activity: Commit logs, branch states, and unstaged diffs across all repositories you touched.
• Terminal History: Recent PowerShell and Bash/Zsh command history.
• Active Desktop Windows: Open IDE files, devtools, and editor windows you were looking at right before signing off.
• Browser Research Tabs: Local Chrome/Edge/Brave history to identify technical docs and discussions you referenced. It scrubs all secrets and passes the sanitized context to Google's Gemma 2 running locally on your machine via Ollama. In seconds, it produces a structured, actionable Tomorrow-Me Resumption Brief complete with:
• A plain-English task summary
• Evidence-based roadblock hypotheses
• Interactive, clickable next-step checkboxes
• Click-to-copy terminal commands to immediately resume your dev servers

Demo

Live App:

Youtube Demo Video:

Here is the ReMind workflow in action:

1. Multi-Project Daily Activity Scanner

ReMind detects activity across all active repositories, uncommitted file changes, 25 terminal commands, and 5 active windows:

2. The Gemma 2 Tomorrow-Me Resumption Brief

In one click, Gemma 2 synthesizes the chaotic context into a calm, structured game plan with interactive next steps and modified file tags:

3. Context & Sanitization Inspector

Transparency is key. Before inference, you can inspect the exact terminal commands, git diffs, and window titles captured:

4. Past Handoff History

Searchable chronological archive of previous days' sessions to track project velocity over time:

Code

The complete source code is open source under the MIT License:

How I Built It

ReMind is built from the ground up around Google's Gemma 2 (gemma2:2b) running locally through Ollama.

The Pipeline Architecture:

[ Git Diffs & Commits ] [ Terminal History ] ──► [ Secret Scrubber ] ──► [ Gemma 2 (Local Ollama) ] ──► [ Structured Brief ] [ Browser History ] (Masks Keys/Tokens) (Typed Pydantic Schema) (Checklist + Commands) [ Open Desktop Windows]

1. Context Harvesting Daemon (context_service.py):
   • Discovers all active git repositories on your drive by scanning commit timestamps.
   • Extracts shell commands from PowerShell (ConsoleHost_history.txt) and Bash.
   • Inspects desktop window titles via Windows API to record open files and tools without recording screen pixels.
   • Reads local Chromium SQLite history for recent technical domains (StackOverflow, GitHub, official docs).
2. Regex Secret Scrubber & Token Budgeter (sanitizer.py):
   • Protects sensitive credentials by auto-redacting OpenAI API keys, GitHub PATs, AWS secrets, bearer tokens, and private SSH keys across all inputs.
   • Prioritizes high-signal application code over lockfiles (package-lock.json, poetry.lock) to stay within Gemma's context window.
3. Gemma 2 Prompt Engineering & Structured Inference (ai.py):
   • Uses strict system prompts and Pydantic JSON schemas to constrain Gemma 2 to produce clean, anti-hallucinatory resumption structures: current_task, progress_summary, roadblocks_and_hypotheses, and recommended_next_steps.
4. Web Dashboard & Background Scheduler (frontend/ & scheduler.py):
   • An end-of-day timer triggers automatically when you log off, sending a native desktop toast notification when tomorrow's briefing is compiled.
   • The React frontend provides an interactive checklist so you can cross off morning tasks as you complete them.

Why Does Open Innovation Matter?

This project simply could not exist with closed, cloud-based LLM APIs. Here is why:

1. Proprietary Source Code Never Leaves Your Device: Developers and enterprise organizations cannot pipe uncommitted git diffs, internal repository names, and private terminal commands to closed cloud APIs like OpenAI or Anthropic without violating privacy and IP policies. Because Gemma 2 is an open-weight model, all inference runs entirely in-memory on your local CPU/GPU.
2. Zero Operating Cost: Running a daily automated daemon using commercial API tokens would incur monthly recurring bills. Open-weight models like Gemma cost $0 to run indefinitely.
3. 100% Offline Capability: Whether you are working on a flight, in a café with spotty Wi-Fi, or in an air-gapped environment, ReMind works seamlessly without internet access.
4. Model Swapping & Freedom: Because ReMind integrates with standard open-inference protocols, users can switch between gemma2:2b for lightweight ultra-fast laptop inference, or scale up to gemma2:9b or fine-tuned coding weights with zero code changes.

The Lazy Payoff

International Sloth Day reminds us that efficiency isn't about working harder—it's about eliminating unnecessary friction.
Before ReMind, resuming a multi-repo project took an average of 15 to 25 minutes of cognitive overhead every morning:

• Retracing terminal history
• Reading git status across multiple folders
• Re-orienting mental models With ReMind:
• Time saved: ~20 minutes every single morning (~1.5 hours per week).
• Effort saved: You never have to write end-of-day status summaries, slack updates, or standup notes manually again. Open your dashboard, review Gemma's checklist, copy your standup notes in 1 click, and start coding.

Prize Categories

Featured Categories:

• Best Use of Gemma ($200): ReMind is built entirely around Google's open-weight gemma2:2b model, utilizing local Ollama inference, structured prompt schemas, and on-device privacy guarantees.
• Best Use of Render ($200): The ReMind web dashboard and landing page are hosted directly on Render using an Infrastructure-as-Code Blueprint (render.yaml). Render automates the continuous deployment of our React/Vite frontend as a global CDN static site, providing an interactive online showcase for developers to explore the resumption interface. (Live Deployment: remind-dashboard on Render)

Thank you to DEV, GitHub, and Google for organizing Hacktoberfest 2026 Week 3!