OpsFlow — AI Chief of Staff

A multi-agent AI system that manages your tasks, schedule, notes, and research from a single natural-language goal.

Built for the Google Gen AI APAC Hackathon using Vertex AI Gemini 2.5 Pro.

---

How it works

You type one sentence. OpsFlow coordinates four specialised AI agents in sequence:

User goal
   └── Orchestrator Agent (Gemini 2.5 Pro)
           ├── Research Agent   → live Google Search grounding
           ├── Task Agent       → creates prioritised tasks in DB
           ├── Calendar Agent   → schedules focus blocks in DB
           └── Notes Agent      → saves structured briefs in DB

Every agent action streams live to the dashboard over WebSocket.

---

Demo

Type: I have a product launch in 3 days. Help me prepare.

Within ~20 seconds you get:

• 6 prioritised tasks with due dates
• 4 calendar focus blocks scheduled
• A "Launch Day Playbook" note with checklist and contacts
• Research summary from live web search

---

Project Structure

opsflow/
├── main.py                  # FastAPI — REST + WebSocket
├── config.py                # Vertex AI client (ADC)
├── mcp_config.py            # MCP-style tool registry
├── requirements.txt
├── Dockerfile
├── deploy.sh                # One-command Cloud Run deploy
├── agents/
│   ├── orchestrator.py      # Plans + coordinates sub-agents
│   ├── task_agent.py        # Creates tasks → SQLite
│   ├── calendar_agent.py    # Schedules events → SQLite
│   ├── notes_agent.py       # Saves notes → SQLite
│   └── research_agent.py    # Web search via Vertex AI grounding
├── db/
│   └── database.py          # SQLite schema + helpers
└── frontend/
    └── index.html           # Live dashboard (no build step)

---

Prerequisites

• Python 3.10+
• Google Cloud SDK (gcloud)
• A GCP project with billing enabled
• Vertex AI API enabled on the project

---

Local Setup

1. Clone and install dependencies

git clone <repo-url>
cd opsflow
pip install -r requirements.txt

2. Authenticate with GCP

gcloud auth application-default login
gcloud auth application-default set-quota-project YOUR_PROJECT_ID

3. Enable Vertex AI API

gcloud services enable aiplatform.googleapis.com --project YOUR_PROJECT_ID

4. Configure environment

cp .env.example .env
# Edit .env and set GOOGLE_CLOUD_PROJECT=your-project-id

5. Run

uvicorn main:app --reload

Open http://localhost:8000

---

Deploy to Cloud Run

./deploy.sh

The script will:

1. Enable required GCP APIs
2. Create a service account with Vertex AI permissions
3. Build and push the Docker image via Cloud Build
4. Deploy to Cloud Run with auto-scaling

---

Tech Stack

Layer	Technology
AI model	Vertex AI Gemini 2.5 Pro
Web search	Vertex AI Google Search Grounding
Backend	FastAPI + Python 3.12
Real-time	WebSocket
Database	SQLite
Frontend	HTML/JS + Tailwind CSS
Deployment	Google Cloud Run
Auth	Application Default Credentials

---

API Endpoints

Method	Path	Description
POST	/api/execute	Run a goal (REST fallback)
WS	/ws/{user_id}	Real-time goal execution
GET	/api/tasks/{user_id}	List tasks
GET	/api/events/{user_id}	List calendar events
GET	/api/notes/{user_id}	List notes
GET	/api/history/{user_id}	Workflow history
DELETE	/api/reset/{user_id}	Clear all data