uregistry 0.2.0

Directory-based configuration registry — file paths become env var names

uregistry: Directory-based configuration registry

Your config lives in plain files. File paths become env var names, file content becomes the value. No YAML, no JSON, no .env parsers — just a directory tree. 🌳

env/
  AI/
    DEFAULT_MODEL    → AI_DEFAULT_MODEL = "claude-sonnet-4-20250514"
    API_KEY          → AI_API_KEY = "sk-..."
  DEBUG              → DEBUG = "1"

Features

• 📁 Files are config: Each file is one variable. The path is the name, the content is the value.
• 🔀 Smart path resolution: Underscores become directories, or use / for explicit control.
• 🎚️ Two-layer registry: System (shared) + local (project) — local values always win.
• 🐚 Shell integration: Generate export statements for bash/zsh/fish — drop it into your rc file.
• 🧹 Zero dependencies: Pure Python, nothing to install beyond the package itself.
• 🔄 Read, write, delete: Full CRUD for your config with automatic directory cleanup.

Installation

pip install uregistry

Quick Start

from uregistry import load_env, get_env, set_env, delete_env, list_env, dump_shell

# Load all vars from env/ into os.environ
load_env()

# Read a single var (does not modify os.environ)
model = get_env("AI_DEFAULT_MODEL")
model = get_env("AI_DEFAULT_MODEL", default="gpt-4o")

# Write a var (creates dirs as needed, updates os.environ)
set_env("AI/API_KEY", "sk-new-key")
set_env("DEBUG", "1")

# Delete a var (removes file, cleans empty dirs)
delete_env("DEBUG")

# List all vars without modifying os.environ
all_vars = list_env()

# Generate shell export statements
print(dump_shell())                    # bash/zsh format
print(dump_shell(shell="fish"))        # fish format

System + Local Registry

uregistry supports two layers: system (shared/global defaults) and local (project-specific overrides). Local values always take priority. Writes always go to local.

# Load system defaults, then apply local overrides
load_env(system_dir="/etc/myapp/env")

# Same for reading — local wins over system
model = get_env("AI_DEFAULT_MODEL", system_dir="/etc/myapp/env")

# Writes always go to local registry
set_env("AI_DEFAULT_MODEL", "gpt-4o", system_dir="/etc/myapp/env")

# Delete from local — system value becomes visible again
delete_env("AI_DEFAULT_MODEL")

# List and dump merge both layers
all_vars = list_env(system_dir="/etc/myapp/env")
print(dump_shell(system_dir="/etc/myapp/env"))

/etc/myapp/env/              (system)        ./env/                  (local)
  AI/                                          AI/
    DEFAULT_MODEL = "gpt-4o"                     DEFAULT_MODEL = "claude-sonnet-4-20250514"  ← wins
    API_KEY = "sk-shared-..."                  DEBUG = "1"

If the system registry doesn't exist or system_dir is not specified — everything works with local only, same as before.

Path Resolution

Use / in names to set explicit directory boundaries:

set_env("AI/CRON_MODEL", "haiku")   # → env/AI/CRON_MODEL

Without /, underscores become directory separators automatically:

set_env("AI_MODEL", "sonnet")       # → env/AI/MODEL

If a file already exists that matches the full var name, it wins — no ambiguity.

Shell Integration

Drop this into your .bashrc / .zshrc and your env dir becomes the source of truth for your shell:

eval "$(python -m uregistry.registry)"

Or for fish, in config.fish:

python -m uregistry.registry --shell fish | source

Auto-detects your shell when run directly, so python -m uregistry.registry just works. 🐟

Why?

• dotenv files get messy — merge conflicts, quoting bugs, one giant blob.
• Directory trees are git-friendly, diffable, and you can manage individual vars with standard Unix tools (cat, echo >, rm).
• Secret management fits naturally — just gitignore sensitive subtrees.

Contributing

Pull requests and issues are welcome!

1. Fork it
2. Create your feature branch (git checkout -b my-new-feature)
3. Commit your changes (git commit -am 'Add some feature')
4. Push to the branch (git push origin my-new-feature)
5. Create a new Pull Request

License

MIT