How to Extract Specs from Framework Modules

Ash.StateMachine

Requires ash_state_machine as a dependency.

mix tlx.gen.from_ash_state_machine MyApp.Order --output specs/order.ex

The extractor uses runtime introspection via AshStateMachine.Info:

• Reads all states, initial states, transitions
• Expands :* wildcards (excluding deprecated states)
• All transitions are :high confidence (declarative DSL)

No source parsing needed — the module must be compiled.

Reactor

Reactor is a workflow orchestration library from the Ash ecosystem. Available as a transitive dependency via ash.

mix tlx.gen.from_reactor MyApp.ProvisionWorkflow --output specs/provision.ex

The extractor uses Spark introspection via Reactor.Info.to_struct!/1:

• Reads the step DAG (inputs, dependencies, async flags, retries)
• Detects compensation and undo callbacks
• Builds a dependency graph with cycle detection

The generated spec models each step's status (pending → completed/failed) with guards enforcing dependency ordering.

Broadway

Broadway is a data processing pipeline library. Add {:broadway, "~> 1.0"} as a dependency.

mix tlx.gen.from_broadway MyApp.IngestPipeline --output specs/ingest.ex

The extractor parses source AST for Broadway.start_link/2 config:

• Producers (module, concurrency, rate limiting)
• Processors (name, concurrency, demand settings)
• Batchers (name, concurrency, batch_size, batch_timeout)
• Callback counts (handle_message, handle_batch)

The generated spec models concurrency bounds and batch size invariants.

Choosing the right extractor