Resumer-Agent

• 

  setup

• 

  dashboard

Inspiration

Developers don't just lose their place when they're interrupted—they lose their train of thought. Existing tools can restore open files or terminal sessions, but they can't restore the reasoning behind a debugging session or a half-finished feature. I wanted to build an AI collaborator that remembers why I was working on something, so getting interrupted doesn't mean rebuilding my mental context from scratch.

What it does

Resume-Agent is an autonomous background agent that quietly watches a developer's active workflow through code diffs, terminal output, and Git commit history. When an interruption happens, it reconstructs the developer's intent—not just which file was open, but the problem being solved and the next logical step to take.

The agent becomes more useful over time. It learns which types of work, such as debugging, refactoring, or building new features, are interrupted most often and automatically adjusts its checkpointing frequency for those tasks. Instead of waiting for the developer to ask for a summary, Resume-Agent proactively preserves context before it's lost.

How I built it

I built Resume-Agent as an autonomous AI workflow powered by Google's ecosystem.

• Gemini 3.5 reconstructs developer intent from recent activity.
• Google Agent Development Kit (ADK) orchestrates the background agent's workflow.
• Python powers the installable command-line interface and local monitoring.
• Cloud Firestore stores checkpoints and learning data in real time.
• Google Stitch provides a live dashboard for visualizing interruption patterns.
• Firebase Hosting deploys the dashboard with seamless synchronization.

Challenges I ran into

The biggest challenge was inferring intent rather than simply recording activity. A list of file changes doesn't explain whether someone was chasing a bug, refactoring a module, or implementing a feature.

Another challenge was making the adaptive behavior feel natural. I wanted the agent to become smarter from repeated interruptions without requiring developers to constantly configure or train it.

Accomplishments that I'm proud of

• Built a fully autonomous background agent that works without interrupting the developer's workflow.
• Created AI-generated checkpoints that preserve reasoning instead of just editor state.
• Implemented adaptive checkpointing that learns from a developer's personal interruption patterns.
• Connected the CLI with a live Firestore-backed dashboard for real-time visibility into workflow trends.
• Delivered the project as a real, installable command-line tool.

What I learned

This project taught me that collaboration isn't just about answering questions—it's about anticipating when someone will need help. I learned how to combine multiple developer signals into meaningful context reconstruction while building autonomous workflows with Gemini, ADK, Firestore, and Firebase.

What's next for Resume-Agent

I want to make Resume-Agent an even stronger collaborative partner by expanding beyond the terminal.

• VS Code and JetBrains integrations.
• Smarter interruption detection using system events.
• Cross-device checkpoint synchronization.
• Richer productivity insights based on long-term interruption patterns.
• Privacy-first local intelligence with more personalized learning over time.

Built With

• agent-development-kit
• claude
• cli
• cloud-firestore
• firebase
• firebase-hosting
• gemini
• gemini-api
• git
• gitpython
• google-adk
• google-cloud
• google-stitch
• python