A runtime you can build on

Melder wires ordinary Python objects into a live dependency graph: services, configuration, connections, and the objects your application works with. Start with a small application. Grow into connected subsystems, isolated worlds, and agent-operated infrastructure when you need them.

bind       register what exists
conjure    compile and validate the graph
meld       resolve an instance

pip install melder

Python 3.14+. Zero runtime dependencies. The saved example curriculum targets free-threaded Python 3.14.

Start with a working example

Bind a plain class and resolve your first instance.

Hello Melder

Browse the examples

Learn through the runnable scripts saved with Melder.

Learn by running examples

Explore the full contents

Jump directly to a level, lesson, or reference.

Full Contents

Choose your level

The four levels follow the same progression as the README. Read in order or enter where your question belongs. Later capabilities stay optional.

🟢 Beginner

The core runtime. Bind, resolve, categorize, scope, and clean up.

For: anyone building their first Melder application.

🟢 Beginner

🟡 Intermediate

Connection. Register modules, configure scopes, and link independent subsystems.

For: applications growing beyond one scope.

🟡 Intermediate

🟠 Advanced

Isolation and inspection. Work with separate worlds, runtime views, and precise overrides.

For: multi-tenant and multi-world systems.

🟠 Advanced

🔵 Expert

Agent rooms, transactions, checkpoints, research, and governed change.

For: runtime and agent infrastructure builders.

🔵 Expert

Build something real

🟢 A complete first application

Combine registration, categories, lifetimes, and a working application graph.

Build a complete beginner application

🟡 Connect three subsystems

Wire platform → services → workflows through explicit contracts.

Compose platform → services → workflows

🟠 Inspect an isolated world

Choose the world, open its view, and understand what is visible.

Inspect an explicitly selected world

🔵 Generate a working application

Turn source strings into a tokenizer, counter, reporter, and independent workers.

Validate → materialize → import → bind → meld

Browse all saved examples by level, topic, or name.

Three verbs, a high ceiling

Your classes stay ordinary Python. A spellbook describes what can be built; a conduit is the runtime you resolve through. Choose how long instances live and who cleans them up.

That graph also gives deeper tooling something concrete to inspect and operate. The higher levels explain those capabilities without making them prerequisites for your first application.

Continue with Hello Melder, inspect the system picture, or use the API reference.