Abstract Backend

Abstract Backend provides a pluggable backend layer for Python services. Applications code against shared protocols while concrete providers—Redis, Kafka, AWS SQS, or your own implementation—are discovered at runtime. Install a provider with pip, remove it with pip uninstall, and keep your business logic unchanged.

Status & Quality

CI/CD & Testing

Code Coverage & Quality

Code Style & Standards

Package Info

Downloads

Why it exists

• 🚫 Stop copying backend code – Extracted from real MCP server projects so multiple services can share the same queue abstraction.
• 📐 Protocol-first design – Contracts live in abe/types.py, keeping providers honest with structural typing and contract tests.
• 🔍 Runtime discovery – load_backend() selects implementations via Python entry points, driven by environment variables.
• 📊 Operational clarity – Logging helpers, contract suites, and documentation make backends observable and portable.

Features

• 📮 Queue provider protocols covering publish, consume, and lifecycle operations.
• ⚙️ AsyncLoopConsumer helper to execute handlers against any compliant backend.
• 🪵 Logging utilities for consistent configuration across providers and apps.
• 🧪 Contract tests to validate third-party implementations.
• 📚 Documentation & examples detailing architecture, provider lifecycle, and runtime flow.

Installation

Install the core library:

pip install abstract-backend

Then install one or more providers, for example the Redis backend:

pip install abe-redis

Set QUEUE_BACKEND=redis (the entry-point exposed by abe-redis) and load_backend() will resolve the correct provider at runtime.

Quick start ⚡️

from abe.backends.message_queue.loader import load_backend
from abe.backends.message_queue.consumer import AsyncLoopConsumer


async def process(payload: dict[str, object]) -> None:
    ...


async def main() -> None:
    backend = load_backend()
    consumer = AsyncLoopConsumer(backend, group="billing")
    await consumer.run(process)

See docs/contents/development/architecture/ for diagrams explaining the flow and provider relationships.

Building providers 🧩

1. Implement the protocols from abe/types.py (especially MessageQueueBackendProtocol).
2. Expose a from_env() constructor for runtime configuration.
3. Register an entry point under abe.backends.message_queue in pyproject.toml.
4. Run the contract tests in test/contract_test/backends/message_queue/ against your provider.
5. Publish your package to PyPI; users activate it with pip install and QUEUE_BACKEND.

The showcase at /docs/src/pages/showcase.tsx highlights template and reference implementations such as abe-redis.

Development 🛠️

• Install dev dependencies with uv pip install -r pyproject.toml or your preferred tool.
• Run tests: uv run pytest.
• Type checking: uv run mypy.
• Linting: uv run pylint abe tests.

The project follows black formatting and pylint linting (see .pre-commit-config.yaml).

Documentation & showcase 📖

• Developer docs live under docs/contents/development/ and include architecture, provider lifecycle, and layer integration guides.
• A Showcase page (/docs/src/pages/showcase.tsx) lists template and implementation repositories with release badges.
• Run cd docs && pnpm start for local previews.

CI/CD & workflows 🤖

• GitHub Actions definitions reside in .github/workflows/.
• Reusable workflows leverage the logging, testing, and packaging helpers of this project.
• docusaurus.config.ts and docs/contents/development/sidebars.ts manage documentation navigation.

License

This project is licensed under the MIT License.