gdrive-organizer 0.1.0b10

AI-powered Google Drive organizer CLI — scan, classify, review, apply, rollback

gdrive-organizer

AI-powered Google Drive organizer for technical users.

Scan your entire Drive with a single flat-crawl API call, classify files with AI (xAI Grok), review a generated execution plan, dry-run it, apply it, and roll back if anything looks wrong. All from the CLI, with an optional local web viewer for visual inspection.

Status: Beta for technical users

---

Why this exists

Large Google Drive accounts get messy fast. Manual cleanup is slow, repetitive, and risky. One bad bulk operation can make things worse.

gdrive-organizer automates the grunt work while keeping you in control:

scan --> classify --> review plan --> dry-run --> apply --> rollback

AI suggests changes. You review them before anything touches your Drive.

---

Key features

• Flat-crawl scanner -- one files.list call fetches everything, tree is reconstructed in memory (~100x fewer API calls than recursive folder queries)
• AI classification -- xAI Grok analyzes file metadata and suggests renames, moves, and folder structures via configurable presets
• Plan/Apply separation -- AI output becomes a reviewable JSON plan, never directly mutates your Drive
• Dry-run -- preview exactly what will change before committing
• Rollback -- undo applied plans with proper ordering and untrash semantics
• Readonly OAuth mode -- scan and inspect without granting write access
• Local web viewer -- FastAPI dashboard for visual plan review, tree exploration, duplicate detection, and permission audit
• Incremental sync -- Changes API integration for fast re-scans
• Safety by default -- atomic snapshot writes, advisory locking, session-gated viewer mutations, LLM response validation

---

Quick Start

Prerequisites

• Python 3.12+
• A Google Cloud Platform project with Drive API enabled (free)
• An xAI API key for AI classification (console.x.ai)

1. Install

# macOS / Linux — Homebrew tap
brew tap papa-channy/tap
brew install gdrive-organizer

# Any platform — isolated install via pipx (recommended)
pipx install gdrive-organizer

# Any platform — pip
pip install gdrive-organizer

For development or contributing, install from source instead:

git clone https://github.com/papa-channy/google-drive-manager.git
cd google-drive-manager
uv sync --all-extras

2. Set up Google OAuth

You need a GCP OAuth client credential. This is a one-time setup:

1. Create a GCP project at console.cloud.google.com
2. Enable the Google Drive API
3. Configure OAuth consent screen (External, testing mode)
4. Create an OAuth client ID (Desktop app) and download credentials.json
5. Place it in the config directory:

mkdir -p ~/.config/gdrive-organizer
cp ~/Downloads/credentials.json ~/.config/gdrive-organizer/

See docs/04-USER-GUIDE.md for detailed step-by-step GCP setup with screenshots.

3. Authenticate

gdrive-organizer auth login
# Browser opens -> select Google account -> authorize -> done

4. Scan your Drive

gdrive-organizer scan
# Scanning Google Drive...
# Page 1: 1000 files (total: 1000)
# ...
# Scan complete: 1,282 files, 314 folders

5. Explore what you have

gdrive-organizer tree --depth 3     # folder tree
gdrive-organizer stats              # file type breakdown
gdrive-organizer permissions        # sharing audit

6. Classify with AI and review

# Register your xAI API key (stored in system keychain)
gdrive-organizer config api-key set

# Generate a classification plan
gdrive-organizer classify

# Review the plan
gdrive-organizer plan show

7. Dry-run first, then apply

# Preview changes without touching Drive
gdrive-organizer plan apply --dry-run

# Apply the approved selection from the viewer (default)
gdrive-organizer plan apply

# Bypass the approved subset and force the latest full plan
gdrive-organizer plan apply --latest-plan

# Roll back if needed
gdrive-organizer plan rollback

plan apply and plan show now prefer approved-plan.json when it exists. A fresh classify run automatically invalidates any older approval, and a real apply archives the consumed approval as history.

8. Visual inspection (optional)

gdrive-organizer view
# Opens http://127.0.0.1:8765 in your browser

---

Screenshots

The local viewer renders your Drive state, AI-generated plan, and applied-plan history.

Dashboard — storage overview, indexed counts, quick actions:

Plan Review — name/type filters, bulk actions, approved-subset banner, last-applied panel, and per-op before/after diff:

Explorer — in-memory file tree from the flat-crawl scanner:

Permissions — public / domain / specific-user / private breakdown with a list of exposed files:

Screenshots are generated from synthetic data by scripts/capture_screenshots.py. Re-run after viewer UI changes to refresh them.

---

CLI Reference

gdrive-organizer auth login|status|logout
gdrive-organizer scan [--folder ID] [--incremental] [--no-filter]
gdrive-organizer tree [--depth N]
gdrive-organizer stats
gdrive-organizer permissions
gdrive-organizer classify [--preset migration|second-brain|custom] [--prompt-file PATH]
gdrive-organizer plan show [--latest-plan] [--phase N] [--per-phase N]
gdrive-organizer plan apply [--dry-run] [--verbose] [--force] [--latest-plan] [-y]
gdrive-organizer plan rollback [-y]
gdrive-organizer view [--port N]
gdrive-organizer config show
gdrive-organizer config set KEY VALUE
gdrive-organizer config api-key set|show|delete
gdrive-organizer config exclude --show|--add-folder PATTERN|--remove-folder PATTERN|--add-ext .EXT|--add-name PATTERN|--reset
gdrive-organizer cat FILE_ID
gdrive-organizer export FILE_ID [--mime-type TYPE]
gdrive-organizer create-folder NAME [--parent ID]

---

CLI vs Viewer

gdrive-organizer is CLI-first. The local web viewer is a review surface, not a full application.

Use the CLI for	Use the viewer for
Authentication, scanning, classification	Visual tree exploration
Applying and rolling back plans	Plan review with checkboxes
Configuration and operational control	Duplicate detection
Scripting and automation	Permission audit dashboard

The viewer runs locally (127.0.0.1 only) and has session-gated mutation controls. It is intentionally secondary to the CLI. When you approve a subset in the viewer, the CLI treats that approved subset as the default execution source until it is applied or invalidated by a newer full plan.

---

Safety Model

This project is designed around automation with guardrails.

Feature	What it does
Plan/Apply separation	AI output becomes a reviewable plan, not direct mutations
Approved subset authority	Viewer approval becomes the default CLI apply/show source
LLM validation boundary	AI responses are validated before becoming executable operations
--dry-run	Preview all changes without touching Drive
Rollback	Undo applied plans with correct ordering and untrash semantics
Readonly OAuth	Scan and inspect without write credentials
Atomic snapshot writes	Local state files are written atomically to prevent corruption
Advisory snapshot locking	Reduces concurrent write races
Viewer session gates	Viewer mutations require confirmation and session tokens
Localhost-only viewer	Web UI binds to 127.0.0.1 by default

Important: Once you explicitly run plan apply, the tool performs real Google Drive mutations. Always use --dry-run first, and use --latest-plan only when you intentionally want to bypass the approved subset.

---

Trust, Privacy, and Data Handling

• OAuth credentials are stored locally in ~/.config/gdrive-organizer/
• API keys are stored in your system keychain (via keyring)
• Directories are created with 0700 permissions on POSIX systems
• Snapshots are stored locally in ~/.local/share/gdrive-organizer/
• The viewer binds to localhost only
• During classification, file metadata (names, paths, types, sizes) is sent to the configured AI provider (xAI). File contents are not sent by default
• No telemetry or analytics are collected

If you want to explore your Drive without any data leaving your machine, skip the classify step -- scan, tree, stats, and permissions are fully local operations.

---

Classification Presets

Preset	Description
migration	Analyzes existing folder structure, organizes into time-series archive with folder shortcuts
second-brain	Extracts multi-tag classifications from text content, organizes by topic and timeline
custom	Uses your own prompt file via --prompt-file

gdrive-organizer classify --preset second-brain
gdrive-organizer classify --preset custom --prompt-file my-rules.txt

---

Configuration

gdrive-organizer config show            # View current settings
gdrive-organizer config set KEY VALUE   # Update a setting
gdrive-organizer config api-key set     # Register xAI API key
gdrive-organizer config api-key delete  # Remove API key

Key settings:

Setting	Default	Description
llm_model	grok-4-1-fast-non-reasoning	AI model for classification
viewer_port	8765	Local viewer port
scan_page_size	1000	Files per API page
batch_size	100	Operations per batch (max 100)
write_rate_limit	3.0	Max writes/sec to Drive API
oauth_readonly	false	Use readonly OAuth scope

---

Development

git clone https://github.com/papa-channy/google-drive-manager.git
cd google-drive-manager
uv sync --all-extras

# Using justfile (requires just: https://github.com/casey/just)
just test          # Run tests (210 tests)
just lint          # Lint with ruff
just fmt           # Format with ruff
just check         # All of the above
just start         # Start viewer in background
just stop          # Stop viewer
just               # List all available commands

See CONTRIBUTING.md for development workflow details.

---

Current Limitations

This is a working beta. Known limitations:

• OAuth testing mode: Refresh tokens expire after 7 days (Google restriction for unverified apps). Re-run auth login to refresh.
• Memory: Full scan loads all file metadata into memory. Drives with 1M+ files may need more RAM.
• AI classification: Requires an xAI API key. Without it, scan/tree/stats/permissions still work.
• Content-based classification: Currently classifies by metadata only (name, path, type, size). Content export for deeper analysis is on the roadmap.
• Google Workspace files: No MD5 checksums available, so duplicate detection falls back to name+size heuristics.

These are maturity limitations, not safety concerns. The core scan-classify-review-apply workflow is stable and tested.

---

Troubleshooting

"Token has been expired or revoked"

Google OAuth tokens in testing mode expire after 7 days. This is a Google restriction for unverified apps.

gdrive-organizer auth logout
gdrive-organizer auth login

"Rate limit exceeded" during plan apply

Google Drive enforces ~3 writes/sec. The tool handles this automatically with exponential backoff, but if it persists:

gdrive-organizer config set write_rate_limit 2.0   # lower the rate
gdrive-organizer config set batch_size 50           # smaller batches

"Snapshot is stale" warning

If your scan is older than the configured freshness window (default: 60 minutes), the tool warns before applying. Options:

gdrive-organizer scan --incremental    # fast re-scan via Changes API
gdrive-organizer plan apply --force    # override freshness check (use carefully)

"File not found (404)" during apply

A file was deleted or moved in Drive after your last scan. Re-scan and re-classify:

gdrive-organizer scan
gdrive-organizer classify

Safe first run recommendation

If this is your first time using the tool, try the readonly exploration path first:

gdrive-organizer config set oauth_readonly true
gdrive-organizer auth login           # grants read-only access
gdrive-organizer scan
gdrive-organizer tree --depth 3
gdrive-organizer stats

Once comfortable, switch to full mode for classification and plan execution:

gdrive-organizer config set oauth_readonly false
gdrive-organizer auth logout
gdrive-organizer auth login           # re-authenticate with write access

credentials.json not found

Download it again from GCP Console -> APIs & Services -> Credentials -> your OAuth client -> Download JSON. See docs/04-USER-GUIDE.md for the full GCP setup walkthrough.

---

Roadmap

Near-term:

• Content-based classification (export + analyze file text)
• Performance benchmarks at larger Drive sizes
• Homebrew formula
• Screenshots and demo assets for documentation

Contributions welcome -- see CONTRIBUTING.md.

---

Contributing

See CONTRIBUTING.md for guidelines.

---

Project Structure

src/gdrive_organizer/
  cli.py              # Typer CLI entry point
  config.py           # Pydantic Settings
  auth/oauth.py       # Google OAuth flow + readonly mode
  scanner/            # Flat crawl, tree reconstruction, permissions
  store/              # JSON snapshot, Changes API sync
  ai/                 # xAI classifier, naming engine, plan generator
  executor/           # Batch API, rate limiter, rollback
  viewer/             # FastAPI local dashboard + templates
  utils/              # Retry, logging

---

License

Apache License 2.0