Bitwarden Organizer with AI + Rules

Intelligently organizes your Bitwarden vault using Local AI classification using Ollama with manual review for ambiguous cases and a rules engine for future runs.

Features

• AI Classification: Uses local LLM (Ollama) to categorize 600+ logins in minutes
• Rules Engine: Build reusable rules so items are auto-classified on future runs
• Ambiguity Detection: Flags uncertain items for your review with AI suggestions
• Folder Creation: Automatically creates folders and assigns items
• Validation: Count checks and interactive confirmations at every step

Workflow

First Run

1. AI classifies most items automatically
2. Flags 20-50 ambiguous items (generic names, missing URLs, etc.)
3. You review each with AI suggestions + option to create rules
4. Rules saved to bitwarden_rules.json

Subsequent Runs

1. Rules applied first (instant classification)
2. AI only classifies new/unknown items
3. Fewer ambiguous cases (rules handle most edge cases)

Usage

# Install Ollama and pull model
ollama pull qwen3:8b

# Run organizer
uv run bitwarden_organizer.py input.json output.json

Local Preview

Spin up a local bitwarden to preview your newly organized vault (using vaultwarden)

• Install mkcert
• Install docker

mkdir vw-certs
cd vw-certs

mkcert install
mkcert -key-file vaultwarden-key.pem -cert-file vaultwarden-cert.pem 

docker run -d --name vaultwarden \
  -v /tmp/vw-data:/data \
  -v "$(pwd):/certs:ro" \
  -e DOMAIN=https://localhost:8443 \
  -e ROCKET_TLS='{certs="/certs/vaultwarden-cert.pem",key="/certs/vaultwarden-key.pem"}' \\
  -p 8443:80 \\
  vaultwarden/server:latest

Then open https://localhost:8443.

• Create an account, and import bitwarden_organized.json.
• After Preview, you can remove this vaultwarden

# Stop vaultwarden
docker stop vaultwarden && docker rm vaultwarden
export CERTS_DIR=$(pwd)
cd $CERTS_DIR/..
# Delete the certs and vaultwarden data
rm -rf /tmp/vw-data && $CERTS_DIR

Interactive Review

When ambiguous items are found:

Item #47: "Google"
  URL:      mail.google.com
  Username: eznix@company.com
  Notes:    (none)

  AI suggests:
    1. ★ Work & Productivity
    2.   Email
    3.   Other
  Reasoning: Company email address indicates work use

  Choice [1-3, or type category name]: 1
  → Classified as: Work & Productivity

  Create a rule for future items like this?
  Options:
    1. URL contains 'mail.google.com'
    2. Name contains 'Google'
    3. Username ends with '@company.com'
    4. Exact match: 'Google' + this URL
    0. No rule
  Choice [0-4]: 3
  ✓ Rule added: Username ends with '@company.com' → Work & Productivity

Rules File Structure

bitwarden_rules.json stores your decisions:

{
  "url_patterns": {
    "chase.com": "Banking & Finance",
    "*amazon.com": "Shopping"
  },
  "name_keywords": {
    "VPN": "Work & Productivity",
    "PIN": "Banking & Finance"
  },
  "username_patterns": {
    "*@company.com": "Work & Productivity"
  },
  "notes_keywords": {
    "authenticator": "Security Tools"
  },
  "exact_matches": {
    "My Bank|https://bank.com": "Banking & Finance"
  }
}

Categories

• Social Media
• Banking & Finance
• Shopping
• Work & Productivity
• Email
• Streaming & Entertainment
• Gaming
• Developer & Tech
• Health
• Travel
• Government & Legal
• Education
• Utilities & Home
• Other

Tips

• First run takes longer — review ambiguous items carefully and create good rules
• Subsequent runs are fast — rules handle most items automatically
• URL rules are powerful — *amazon.com matches any Amazon subdomain
• Username patterns are great — catch all work accounts with *@company.com
• Start broad, refine later — you can always edit bitwarden_rules.json manually

Requirements

• Python 3.13+
• Ollama with qwen3:8b (or modify MODEL variable for other models)
• Unencrypted Bitwarden JSON export