aekasitt / cachette

Cache extension for ASGI frameworks

Cachette

Features

This is an extension aiming at making cache access on the server By configuration at startup of the FastAPI App instance, you can set the backend and other configuration options and have it remain a class constant when using FastAPI's intuitive Dependency Injection system.

The design has built-in limitations like fixed codec and backend once the app has been launched and encourage developers to design their applications with this in mind.

Most of the Backend implementation is directly lifted from fastapi-cache by @long2ice excluding the MongoDB backend option.

Configuration Options

The following are the current available configuration keys that can be set on this FastAPI extension on startup either by using a method which returns a list of tuples or a Pydantic BaseSettings object (See examples below or in examples/ folder)

backend -- optional; must be one of ["inmemory", "memcached", "mongodb", "pickle", "redis"];
  defaults to using inmemory option which required no extra package dependencies. To use
  other listed options; See installation guide on the README.md at
  [Repository Page](https://github.com/aekasitt/cachette).
codec -- optional; serialization and de-serialization format to have cache values stored in
  the cache backend of choice as a string of selected encoding. once fetched, will have their
  decoded values returned of the same format. must be one of ["feather", "msgpack", "parquet",
  "pickle"]; if none is defined, will vanilla codec of basic string conversion will be used.
database_name -- required when backend set to "mongodb"; the database name to be automatically
  created if not exists on the MongoDB instance and store the cache table; defaults to
  "cachette-db"
memcached_host -- required when backend set to "memcached"; the host endpoint to the memcached
  distributed memory caching system.
mongodb_url -- required when backend set to "mongodb"; the url set to MongoDB database
  instance with or without provided authentication in such formats
  "mongodb://user:password@host:port" and "mongodb://host:port" respectively.
pickle_path -- required when backend set to "pickle"; the file-system path to create local
  store using python pickling on local directory
redis_url -- required when backend set to "redis"; the url set to redis-server instance with
  or without provided authentication in such formats "redis://user:password@host:port" and
  "redis://host:port" respectively.
table_name -- required when backend set to "mongodb"; name of the cache collection in case of
  "mongodb" backend to have key-value pairs stored; defaults to "cachette". 
ttl -- optional; the time-to-live or amount before this cache item expires within the cache;
  defaults to 60 (seconds) and must be between 1 second to 1 hour (3600 seconds).
valkey_url -- required when backend set to "valkey"; the url set to valkey-server instance
  with or without provided authentication in such formats "valkey://user:password@host:port"
  and "valkey://host:port" respectively.

Examples

The following shows and example of setting up FastAPI Cachette in its default configuration, which is an In-Memory cache implementation.

from cachette import Cachette
from fastapi import FastAPI, Depends
from fastapi.responses import PlainTextResponse
from pydantic import BaseModel

app = FastAPI()

### Routing ###
class Payload(BaseModel):
  key: str
  value: str

@app.post('/', response_class=PlainTextResponse)
async def setter(payload: Payload, cachette: Cachette = Depends()):
  await cachette.put(payload.key, payload.value)
  return 'OK'

@app.get('/{key}', response_class=PlainTextResponse, status_code=200)
async def getter(key: str, cachette: Cachette = Depends()):
  value: str = await cachette.fetch(key)
  return value

And then this is how you set up a FastAPI Cachette with Redis support enabled.

from cachette import Cachette
from fastapi import FastAPI, Depends
from fastapi.responses import PlainTextResponse
from pydantic import BaseModel

app = FastAPI()

@Cachette.load_config
def get_cachette_config():
  return [('backend', 'redis'), ('redis_url', 'redis://localhost:6379')]

class Payload(BaseModel):
  key: str
  value: str

@app.post('/', response_class=PlainTextResponse)
async def setter(payload: Payload, cachette: Cachette = Depends()):
  await cachette.put(payload.key, payload.value)
  return 'OK'

@app.get('/{key}', response_class=PlainTextResponse, status_code=200)
async def getter(key: str, cachette: Cachette = Depends()):
  value: str = await cachette.fetch(key)
  return value

Roadmap

1. Implement flush and flush_expired methods on individual backends (Not needed for Redis & Memcached backends)

2. Memcached Authentication (No SASL Support) Change library?

3. Add behaviors responding to "Cache-Control" request header

4. More character validations for URLs and Database/Table/Collection names in configuration options

Installation

The easiest way to start working with this extension with pip

pip install cachette
# or
uv add cachette

When you familiarize with the basic structure of how to Dependency Inject Cachette within your endpoints, please experiment more of using external backends with extras installations like

# Install FastAPI Cachette's extra requirements to Redis support
pip install cachette --install-option "--extras-require=redis"
# or Install FastAPI Cachette's support to Memcached
uv add cachette[memcached]
# or Special JSON Codec written on Rust at lightning speed
uv add cachette[orjson]
# or Include PyArrow package making DataFrame serialization much easier
pip install cachette --install-option "--extras-require=dataframe"

Getting Started

This FastAPI extension utilizes "Dependency Injection" (To be continued)

Configuration of this FastAPI extension must be done at startup using "@Cachette.load_config" decorator (To be continued)

These are all available options with explanations and validation requirements (To be continued)

Examples

The following examples show you how to integrate this extension to a FastAPI App (To be continued)

See "examples/" folders

To run examples, first you must install extra dependencies

Do all in one go with this command...

pip install aiomcache motor uvicorn redis
# or
uv sync --extra examples
# or
uv sync --all-extras

Do individual example with this command...

pip install redis
# or
uv sync --extra redis

Contributions

Prerequisites

• python version 3.9 or above
• uv

Set up local environment

The following guide walks through setting up your local working environment using pyenv as Python version manager and uv as Python package manager. If you do not have pyenv installed, run the following command.

Install using Homebrew (Darwin)

brew install pyenv --head

Install using standalone installer (Darwin and Linux)

curl https://pyenv.run | bash

If you do not have uv installed, run the following command.

Install using Homebrew (Darwin)

brew install uv

Install using standalone installer (Darwin and Linux)

curl -LsSf https://astral.sh/uv/install.sh | sh

Once you have pyenv Python version manager installed, you can install any version of Python above version 3.9 for this project. The following commands help you set up and activate a Python virtual environment where uv can download project dependencies from the PyPI open-sourced registry defined under pyproject.toml file.

Set up environment and synchronize project dependencies

pyenv install 3.9.19
pyenv shell 3.9.19
uv venv  --python-preference system
source .venv/bin/activate
uv sync --dev

Test Environment Setup

This project utilizes multiple external backend services namely AWS DynamoDB, Memcached, MongoDB and Redis as backend service options as well as a possible internal option called InMemoryBackend. In order to test viability, we must have specific instances of these set up in the background of our testing environment. Utilize orchestration file attached to reposity and docker-compose command to set up testing instances of backend services using the following command...

docker-compose up --detach

When you are finished, you can stop and remove background running backend instances with the following command...

docker-compose down

Now that you have background running backend instances, you can proceed with the tests by using pytest command as such...

pytest

Or you can configure the command to run specific tests as such...

pytest -k test_load_invalid_configs
# or
pytest -k test_set_then_clear

All test suites must be placed under tests/ folder or its subfolders.

License

This project is licensed under the terms of the MIT license.