Changelog

Beanis - Redis ODM for Python

0.1.0 - 2025-01-15

Major Release - Complete Redis Refactor

Complete refactor from MongoDB (Beanie fork) to Redis-native implementation.

Core Features

  • ✅ Redis Hash Storage - Documents stored as Redis Hashes
  • ✅ Automatic Indexing - Sorted Sets for numeric fields, Sets for categorical fields
  • ✅ Type Safety - Full Pydantic v2 validation
  • ✅ Async/Await - Built on redis.asyncio
  • ✅ TTL Support - Built-in document expiration
  • ✅ Event Hooks - Before/after insert, update, delete, save

Custom Encoders System

  • Author - Claude Code Assistant
  • Custom type encoding/decoding registry
  • Decorator and function APIs
  • Auto-registration for NumPy and PyTorch types
  • Type metadata storage for runtime type resolution
  • Example: Store NumPy arrays, PyTorch tensors, custom classes

Performance Optimizations

  • msgspec for JSON serialization (2x faster than orjson)
  • Redis pipelines for batch operations
  • Lazy validation - Skip validation on reads by default
  • 8% overhead vs vanilla Redis (benchmarked)

Documentation

  • Complete tutorial system (8 core tutorials)
  • Getting started guide
  • Custom encoders guide
  • Side-by-side code comparisons (vanilla Redis vs Beanis)
  • Removed MongoDB-only features documentation

API Changes

  • init_beanis() instead of init_beanie()
  • Redis client instead of Motor client
  • Indexed(type) for indexable fields
  • Removed: Link, BackLink, migrations, aggregations, views

Tests

  • 72 passing tests
  • Comprehensive document operations tests
  • Custom encoder tests
  • FastAPI integration tests
  • Migration tests (legacy, kept for reference)

0.0.8 - 2024-06-05

Initial Fork from Beanie

  • Forked from Beanie ODM
  • Changed package name to Beanis
  • Updated imports and references
  • Initial PyPI release

Credits

Beanis is inspired by Beanie - the amazing MongoDB ODM by Roman Right.

We took the Beanie philosophy (elegant API, Pydantic models, async/await) and adapted it for Redis, creating a simple yet powerful ODM that works with vanilla Redis.

Migration from Beanie

If you're migrating from Beanie (MongoDB) to Beanis (Redis), here are the key changes:

Initialization

# Before (Beanie)
from motor.motor_asyncio import AsyncIOMotorClient
from beanie import init_beanie

client = AsyncIOMotorClient("mongodb://localhost:27017")
await init_beanie(database=client.db_name, document_models=[Product])

# After (Beanis)
from redis.asyncio import Redis
from beanis import init_beanis

client = Redis(decode_responses=True)
await init_beanis(database=client, document_models=[Product])

Document Definition

# Before (Beanie)
from beanie import Document
import pymongo

class Product(Document):
    name: str
    price: Indexed(float, index_type=pymongo.DESCENDING)

    class Settings:
        name = "products"

# After (Beanis)
from beanis import Document, Indexed

class Product(Document):
    name: str
    price: Indexed(float)  # Sorted Set index

    class Settings:
        name = "products"

Features Not Available

  • ❌ Relations (Link/BackLink) - Use embedded documents
  • ❌ Aggregation pipelines - Use Python for data processing
  • ❌ Migrations - Not needed (schema-less)
  • ❌ Views - Not applicable to Redis
  • ❌ Time Series - Use Redis TimeSeries module or TTL

New Features

  • ✅ TTL support - Document expiration
  • ✅ Custom encoders - Store any Python type
  • ✅ Atomic operations - increment_field() for counters
  • ✅ Performance - 8% overhead vs vanilla Redis