chore(blueprints): remove aliases

[paul-nechifor]

Problem

Blueprint aliases are annoying because there are two ways to do the same thing.

Closes DIM-538

Solution

Removed my_moduel(). Now, you can only do MyModule.blueprint().

Breaking Changes

You can no longer import blueprint aliases. Just import the actual Module.

How to Test

Contributor License Agreement

• I have read and approved the CLA.

[greptile-apps]

Greptile Summary

This PR removes ~100 lowercase_alias = ClassName.blueprint shim variables that were previously exported from module files, replacing all call sites with the canonical ClassName.blueprint(...) pattern directly. It also refactors the test_all_blueprints_generation.py scanner to auto-discover Module subclasses via AST transitive-closure analysis rather than relying on those alias assignments.

Key changes:

• Every module file (agents, perception, navigation, hardware, manipulation, teleop, etc.) drops its foo = Foo.blueprint / __all__ boilerplate — ~5 lines removed per file across ~40 files.
• dimos/control/coordinator.py removes control_coordinator = ControlCoordinator.blueprint and its __all__.
• test_all_blueprints_generation.py adds _build_module_class_set (transitive closure of Module subclasses), _camel_to_snake conversion, _is_production_module_file filtering, and collision-resolution logic so modules no longer need an alias variable to be registered.
• all_blueprints.py is regenerated: all_modules keys are now derived from class names via _camel_to_snake, causing several renames (e.g. realsense-camera → real-sense-camera, mapper → map, unitree-skills → unitree-skill-container) that are breaking CLI changes.
• pyproject.toml adds -s to addopts, disabling output capture globally — likely a debugging leftover.
• _ends_with_blueprint_method is not updated to recognise ClassName.blueprint(...) calls, so bare .blueprint(...) assignments (all 18 coordinator_* entries in dimos/control/blueprints.py) are not detected by the scanner, which would cause test_all_blueprints_is_current to fail in CI.

Confidence Score: 2/5

• Not safe to merge — the blueprint scanner misses the new ClassName.blueprint(...) call pattern, and a debug flag was left in pytest config.
• The bulk of the alias-removal changes across ~100 files are mechanical and correct. However, two issues lower confidence significantly: (1) _ends_with_blueprint_method does not recognise ClassName.blueprint(...) as a blueprint-producing call, meaning all 18 coordinator_* entries in dimos/control/blueprints.py would be dropped from the auto-generated registry and test_all_blueprints_is_current would fail in CI; (2) the -s flag left in pyproject.toml will produce very noisy CI output.
• dimos/robot/test_all_blueprints_generation.py (scanner logic) and pyproject.toml (debug flag) require attention before merging.

Important Files Changed

Filename	Overview
dimos/robot/test_all_blueprints_generation.py	Major refactor of blueprint scanner: replaces variable-alias detection with class-hierarchy analysis using AST transitive closure, but _ends_with_blueprint_method doesn't handle the new SomeClass.blueprint(...) call pattern, likely causing CI failures for coordinator blueprints.
pyproject.toml	Adds -s (no output capture) flag to default pytest addopts — this is a debugging artifact that should be removed before merging.
dimos/control/coordinator.py	Removes control_coordinator = ControlCoordinator.blueprint alias and __all__ export; callers now use ControlCoordinator.blueprint(...) directly. Clean change.
dimos/control/blueprints.py	Replaces all control_coordinator(...) calls with ControlCoordinator.blueprint(...) across 14 coordinator blueprints. Mechanical and correct.
dimos/robot/all_blueprints.py	Auto-generated registry updated: all_modules keys renamed to match _camel_to_snake class names (e.g. realsense-camera → real-sense-camera, mapper → map). Contains several breaking CLI name changes but reflects the new auto-generation logic.
dimos/robot/unitree/go2/blueprints/agentic/_common_agentic.py	Replaces aliased blueprint factories with direct class imports and .blueprint() calls across all five modules in the common agentic stack. Correct.
dimos/manipulation/blueprints.py	Migrates all aliased blueprint factories to direct ClassName.blueprint(...) calls throughout the manipulation stack. Consistent with the rest of the PR.
dimos/visualization/rerun/bridge.py	Removes rerun_bridge = RerunBridgeModule.blueprint alias and its associated comment. All call sites updated to RerunBridgeModule.blueprint(...) directly.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A["_scan_for_blueprints(root)"] --> B["_build_module_class_set(root)\n(AST transitive closure)"]
    B --> C["known Module subclass names"]
    A --> D["Iterate Python files"]
    D --> E["_find_blueprints_in_file(file, module_classes)"]
    E --> F{"Top-level AST node?"}
    F -->|"ast.Assign"| G{"_is_autoconnect_call?\n_ends_with_blueprint_method?"}
    G -->|"Yes"| H["blueprint_vars ← var_name"]
    G -->|"No"| I["❌ SomeClass.blueprint(...)\nnot detected!"]
    F -->|"ast.ClassDef"| J{"base in module_classes?"}
    J -->|"Yes"| K["module_vars ← _camel_to_snake(class_name)"]
    H --> L["all_blueprints[cli_name] = module:var"]
    K --> M{"_is_production_module_file?"}
    M -->|"Yes"| N["all_modules[cli_name] = module_path"]
    M -->|"No"| O["Skipped"]
    L --> P{"Key collision?"}
    N --> P
    P -->|"Collision"| Q["Blueprint takes priority\ndel all_modules[key]"]
    P -->|"No collision"| R["Final registry"]
    Q --> R

Loading

Comments Outside Diff (1)

1. dimos/robot/test_all_blueprints_generation.py, line 301-306 (link)

   SomeClass.blueprint(...) call pattern is never detected as a blueprint variable

   _ends_with_blueprint_method checks if the outermost call's attribute name is in BLUEPRINT_METHODS, but "blueprint" is not in that set. As a result, top-level assignments of the form coordinator_basic = ControlCoordinator.blueprint(tick_rate=100.0, ...) (and all other coordinator_* variables in dimos/control/blueprints.py) are not picked up as blueprint variables.

   The old code detected module_alias = SomeClass.blueprint (attribute-only, no call) via the now-removed _is_blueprint_factory. With all aliases removed, the scanner has no path to detect the new direct-call pattern SomeClass.blueprint(...).

   The committed all_blueprints.py currently includes 18 coordinator-* entries (lines 19–36). If the scanner does not regenerate them, test_all_blueprints_is_current will fail in CI because the generated content won't match.

   The fix is to also recognise func.attr == "blueprint" as a valid blueprint-producing call:

_{Last reviewed commit: "chore(blueprints): r..."}