Digital FTE - AI Employee (Bronze Tier)

Your life and business on autopilot. Local-first, agent-driven, human-in-the-loop.

A minimum viable Personal AI Employee built for the FTE-AI-Employee Hackathon. This Bronze Tier implementation provides the foundation for an autonomous AI agent that proactively manages your personal and business affairs using Claude Code and Obsidian.

🏆 Bronze Tier Deliverables

• ✅ Obsidian vault with Dashboard.md, Company_Handbook.md, and Business_Goals.md
• ✅ One working Watcher script (Filesystem monitoring)
• ✅ Claude Code integration - reads from and writes to the vault
• ✅ Basic folder structure: /Inbox, /Needs_Action, /Done, /Plans, /Pending_Approval, /Approved, /Rejected, /Logs
• ✅ Agent Skills ready - all AI functionality designed as extensible skills

📁 Project Structure

FTE-AI-Employee-Hackathon-0/
├── AI_Employee_Vault/          # Obsidian vault (local knowledge base)
│   ├── Dashboard.md            # Real-time status summary
│   ├── Company_Handbook.md     # Rules of engagement for AI
│   ├── Business_Goals.md       # Objectives and targets
│   ├── Inbox/                  # Raw incoming files
│   ├── Needs_Action/           # Items requiring AI attention
│   ├── Done/                   # Completed tasks
│   ├── Plans/                  # AI-generated action plans
│   ├── Pending_Approval/       # Awaiting human approval
│   ├── Approved/               # Human-approved actions
│   ├── Rejected/               # Rejected actions
│   ├── Logs/                   # Daily JSON activity logs
│   ├── Briefings/              # CEO briefings (future tiers)
│   └── Accounting/             # Financial records (future tiers)
│
├── watchers/                   # Watcher scripts (AI's "senses")
│   ├── base_watcher.py         # Abstract base class for all watchers
│   └── filesystem_watcher.py   # Monitors drop folder for files
│
├── drop_folder/                # Drop files here for AI processing
├── orchestrator.py             # Master process managing the system
├── .env.example                # Environment variable template
├── .gitignore                  # Git ignore rules (security-first)
└── requirements.txt            # Python dependencies

🚀 Quick Start

Prerequisites

Component	Required	Purpose
Python 3.13+	✅	Watcher scripts & orchestrator
Qwen Code	✅	AI reasoning engine
Obsidian (optional)	Recommended	View/edit vault files
Node.js 24+	For future tiers	MCP servers

Setup (5 minutes)

1. Clone the repository

   git clone <repository-url>
   cd FTE-AI-Employee-Hackathon-0

2. Copy environment template

   copy .env.example .env
   # Edit .env and add any API keys (not needed for Bronze tier)

3. Verify Qwen Code is installed

   qwen --version

4. Open the vault in Obsidian (optional but recommended)

   • Open Obsidian → Open folder as vault → Select AI_Employee_Vault/

📖 Usage

Method 1: File Drop Workflow (Easiest)

1. Drop a file into the drop_folder/ directory

   # Example: Copy a text file or document
   copy my_task.txt drop_folder\

2. Run the filesystem watcher (creates action file in vault)

   python watchers\filesystem_watcher.py --vault AI_Employee_Vault --drop drop_folder --once

3. Run the orchestrator (processes the action file, creates a plan)

   python orchestrator.py --vault AI_Employee_Vault --once

4. Review the plan in AI_Employee_Vault/Plans/

5. Use Qwen Code to process the plan:

   qwen --cwd AI_Employee_Vault
   # Prompt: "Check the Plans folder and process any pending plans"

Method 2: Direct Vault Editing

1. Create a task file directly in AI_Employee_Vault/Needs_Action/:

   ---
   type: task
   priority: high
   created: 2026-04-05T10:00:00Z
   status: pending
   ---
   
   # Task: Review weekly goals
   
   Please review the Business_Goals.md file and suggest improvements.

2. Run the orchestrator:

   python orchestrator.py --vault AI_Employee_Vault --once

3. Check the Plans folder for the generated plan.

Method 3: Continuous Monitoring

Run the watchers and orchestrator continuously in the background:

# Terminal 1: Filesystem watcher
python watchers\filesystem_watcher.py --vault AI_Employee_Vault --drop drop_folder

# Terminal 2: Orchestrator
python orchestrator.py --vault AI_Employee_Vault

Now just drop files into drop_folder/ and the system will automatically process them.

🤖 Qwen Code Integration

Qwen Code is the "brain" of the AI Employee. Point it at the vault:

cd AI_Employee_Vault
qwen

Useful Qwen Code Prompts

• "Check the Needs_Action folder and process all pending tasks"
• "Review the Plans folder and execute the first pending plan"
• "Read Company_Handbook.md and summarize the key rules"
• "Update Dashboard.md with today's activity"
• "Generate a weekly summary of all files in the Done folder"

Ralph Wiggum Loop (Autonomous Task Completion)

Keep Qwen working until a task is complete:

/ralph-loop "Process all files in Needs_Action, move to Done when complete" \
  --completion-promise "TASK_COMPLETE" \
  --max-iterations 10

📋 Vault Templates

Dashboard.md

Real-time summary of system status, pending tasks, and recent activity.

Company_Handbook.md

Defines how the AI should behave - communication rules, financial rules, safety rules, and ethics.

Business_Goals.md

Tracks revenue targets, active projects, and key metrics. Update weekly.

🔒 Security

• No credentials stored in vault - all secrets go in .env
• .env is gitignored - never committed to version control
• Human-in-the-loop - sensitive actions require approval file movement
• Audit logging - all actions logged to Logs/YYYY-MM-DD.json
• Dry-run mode - set DRY_RUN=true in .env for testing

🧪 Testing

Test the filesystem watcher:

# Create a test file
echo "Test task for AI Employee" > drop_folder\test_task.txt

# Run watcher once
python watchers\filesystem_watcher.py --vault AI_Employee_Vault --drop drop_folder --once

# Check Needs_Action folder for created action file
dir AI_Employee_Vault\Needs_Action

# Run orchestrator
python orchestrator.py --vault AI_Employee_Vault --once

# Check Plans folder
dir AI_Employee_Vault\Plans

🛣️ Roadmap to Higher Tiers

Silver Tier (Next Steps)

• Add Gmail Watcher (requires Google API credentials)
• Add WhatsApp Watcher (requires Playwright)
• MCP server for sending emails
• Human-in-the-loop approval workflow
• Scheduled tasks via cron/Task Scheduler

Gold Tier

• Full cross-domain integration
• Odoo accounting integration
• Social media posting
• Weekly CEO Briefing generation
• Ralph Wiggum loop for autonomous completion

Platinum Tier

• Always-on cloud deployment
• Cloud/Local work-zone specialization
• Vault sync with claim-by-move rules

🐛 Troubleshooting

"Command not found: qwen"

Install Qwen Code first, then verify installation:

qwen --version

No action files created

• Check that drop_folder/ exists and has files
• Verify the watcher is pointed at the correct paths
• Check console output for errors

Orchestrator not processing files

• Ensure AI_Employee_Vault/Needs_Action/ has .md files
• Check file permissions allow read/write access
• Review logs in AI_Employee_Vault/Logs/

📚 Resources

• Qwen Code Documentation
• Obsidian Download
• Agent Skills Documentation
• Hackathon Blueprint

👥 Community

• Research Meetings: Every Wednesday at 10:00 pm on Zoom
• YouTube: https://www.youtube.com/@panaversity

📄 License

Built for the FTE-AI-Employee Hackathon 2026.

---

AI Employee v0.1.0 - Bronze Tier Local-first, agent-driven, human-in-the-loop