simple_keystore

A Python library for encrypted key storage in SQLite with optional Redis-based rate limiting. Designed for development and on-prem use. For production, consider secrets managers or other built-in secure provider methods.

Features

• Encrypted storage -- Keys are encrypted with Fernet symmetric encryption and stored in SQLite with metadata (name, source, login, batch, expiration dates).
• Rate limiting -- Optional sliding-window rate limiting per key using Redis and atomic Lua scripts.
• Key orchestration -- High-level API that combines storage and throttling with exponential backoff to find and claim an available key.
• CLI tool -- Interactive menu-driven interface for managing keys.

Installation

pip install simple_keystore

Redis is required only if you use the rate limiting features (SKSRateThrottler, get_available_key_for_use).

Configuration

Set the encryption key via environment variable (preferred) or .netrc:

export SIMPLE_KEYSTORE_KEY="your-fernet-key"

Or add an entry to ~/.netrc:

machine SIMPLE_KEYSTORE_KEY
login unused
password your-fernet-key

Generate a valid Fernet key with:

from cryptography.fernet import Fernet
print(Fernet.generate_key().decode())

Usage

Key Storage

from simple_keystore import SimpleKeyStore

ks = SimpleKeyStore("my_keys.db")

# Add keys
ks.add_key("openai", "sk-abc123", active=True)
ks.add_key("openai", "sk-def456", active=True, source="vendor-a", batch="2024-q1")

# Retrieve a key by name (raises if more than one match)
key = ks.get_key_by_name("openai")

# Query matching records
records = ks.get_matching_key_records(name="openai", active=True)
for r in records:
    print(r["id"], r["key"], r["usable"])

# Deactivate / reactivate
ks.mark_key_inactive("sk-abc123")
ks.mark_key_active("sk-abc123")

# Smart key selection (soonest-expiring, smallest set)
next_key = ks.get_next_usable_key(name="openai")

Rate Limiting

from datetime import timedelta
from simple_keystore import SKSRateThrottler

# Allow 10 uses per 60-second window
throttler = SKSRateThrottler(
    api_key_id=1,
    number_of_uses_allowed=10,
    amount_of_time=timedelta(seconds=60),
    redis_host="localhost",
    redis_port=6379,
)

# Check remaining uses
remaining, claimed = throttler.remaining_uses(claim_slot=False)

# Claim a slot
remaining, claimed = throttler.remaining_uses(claim_slot=True)

# Target a different key without mutating instance state
remaining, claimed = throttler.remaining_uses(claim_slot=True, api_key_id=42)

# Block until a slot opens (with timeout)
remaining = throttler.wait_until_available(timeout=300, verbose=True)

You can also inject a redis_client directly (e.g., for testing with fakeredis):

throttler = SKSRateThrottler(
    api_key_id=1,
    number_of_uses_allowed=10,
    amount_of_time=timedelta(seconds=60),
    redis_client=my_redis_instance,
)

When an external redis_client is provided, the throttler will not close it on destruction.

Combined: Get an Available Key

from simple_keystore import SimpleKeyStore, get_available_key_for_use

ks = SimpleKeyStore("my_keys.db")

key_record, remaining = get_available_key_for_use(
    key_name="openai",
    keystore=ks,
    key_number_of_uses_allowed=10,
    key_use_window_in_seconds=60,
    redis_host="localhost",
    redis_port=6379,
    how_long_to_try_in_seconds=3600,  # retry for up to 1 hour
    max_wait_cap_in_seconds=180.0,    # cap backoff at 3 minutes
)

print(key_record["key"], remaining)

This iterates through matching keys, attempts to claim a rate-limit slot for each, and retries with exponential backoff (1.5x multiplier) until one is available or the timeout is reached.

CLI

manage_simple_keys my_keys.db

Provides an interactive menu for adding, removing, activating/deactivating, and listing keys.

API Reference

SimpleKeyStore(name="simple_keystore.db")

SKSRateThrottler(api_key_id, number_of_uses_allowed, amount_of_time, ...)

get_available_key_for_use(key_name, keystore, ...)

Finds and claims an available key matching key_name with rate limiting. Returns (key_record, remaining_uses). Retries with exponential backoff.

get_key_with_most_uses_remaining(key_name, keystore, throttler=None, verbose=False)

Returns the id of the key with the most remaining rate-limit uses, or None. When throttler is provided, it is reused with per-call api_key_id; when None, a new throttler is created per key with hardcoded defaults.

License

MIT