memory2 StreamModules, cleanup

[leshy]

No description provided.

[Copilot]

Pull request overview

This PR extends the memory2 streaming architecture with (1) a StreamModule base that wires In→memory2.Stream pipelines→Out, (2) support for unbound stream pipelines via Stream() + .chain(), and (3) voxel-map accumulation refactored into a reusable VoxelGrid + VoxelMap transformer, alongside typing improvements and new tool-only replay/visualization tests.

Changes:

• Add unbound memory2.Stream pipelines (Stream()), .chain() to bind them to real streams, and .publish() to drive Out ports.
• Introduce StreamModule (defaulting to a live-only NullStore) and update voxel mapping to a framework-agnostic VoxelGrid + VoxelMap transformer.
• Add embedding API overloads + new MemoryStore retention controls (max_size, NullStore), and add tool-marked replay DB tests (with LFS artifact).

Reviewed changes

Copilot reviewed 19 out of 19 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
dimos/simulation/unity/test_unity_sim.py	Add stop() to test transport mock to support new module shutdown behavior.
dimos/robot/all_blueprints.py	Register new "stream-module" module entry.
dimos/models/embedding/treid.py	Add overloads for embed/embed_text return typing.
dimos/models/embedding/mobileclip.py	Add overloads for embed/embed_text return typing.
dimos/models/embedding/clip.py	Add __future__ annotations + overloads for embed APIs.
dimos/models/embedding/base.py	Define overloads on the abstract EmbeddingModel interface.
dimos/memory2/vectorstore/sqlite.py	Validate identifiers; handle missing vec tables by returning empty results.
dimos/memory2/transform.py	Add Batch transformer and stride() utility.
dimos/memory2/test_voxel_map.py	New tool-marked tests for voxel-map accumulation + replay integration.
dimos/memory2/test_visualizer.py	New tool-marked “visualizer” workflow tests over replay DB.
dimos/memory2/test_save.py	Update expected error message substring for .save() restrictions.
dimos/memory2/test_module.py	New tests for unbound streams, .chain(), StreamModule, and NullStore behavior.
dimos/memory2/test_e2e.py	Add tool-marked embed-and-search workflow using CLIP + EmbedImages.
dimos/memory2/stream.py	Add unbound streams, .chain(), .publish(), improved errors, and stricter live/save/append guards.
dimos/memory2/store/null.py	Add NullStore (live-only, max_size=0) convenience store.
dimos/memory2/store/memory.py	Implement MemoryStoreConfig.max_size and wire it into backend creation.
dimos/memory2/store/README.md	Update docs to reflect Store/Backend/ObservationStore roles and new layout.
dimos/memory2/observationstore/memory.py	Add max_size retention via deque(maxlen=...).
dimos/memory2/module.py	Add StreamModule that bridges Module ports to memory2 streams.
dimos/memory2/conftest.py	Add CLIP fixture for tool tests and tighten fixture typing via cast().
dimos/memory/embedding.py	Remove now-unneeded cast() thanks to embed() overload typing.
dimos/mapping/voxels.py	Refactor voxel mapping into VoxelGrid + VoxelMapTransformer; reintroduce mapper as StreamModule.
dimos/mapping/test_voxels.py	Update tests to use VoxelGrid directly and dispose resources explicitly.
dimos/mapping/pointclouds/test_occupancy_speed.py	Switch benchmark to use VoxelGrid instead of the old mapper module.
dimos/hardware/sensors/lidar/fastlio2/fastlio_blueprints.py	Update blueprint wiring to new VoxelGridMapper config (no publish_interval).
dimos/core/stream.py	Add Stream.stop() to stop transports on module shutdown.
dimos/core/module.py	Stop transports for all In/Out ports during module close to avoid leaks/cycles.
data/.lfs/go2_bigoffice.db.tar.gz	Add LFS pointer for replay DB artifact used by tool tests.

Comments suppressed due to low confidence (2)

dimos/memory2/transform.py:91

• Batch.init should validate batch_size > 0 (raise ValueError) to avoid surprising behavior (e.g., batch_size=0 causes every element to be flushed immediately, negative values behave similarly).

    def __init__(self, fn: Callable[[list[T]], Sequence[R]], batch_size: int = 16) -> None:
        self._fn = fn
        self._batch_size = batch_size

dimos/memory2/test_visualizer.py:55

• There are several spelling mistakes in these comments (e.g., "dissaper" -> "disappear"). Please fix to keep docs/comments readable.

# these need to be functions that are easily called to render latest query results in different ways
# I can imagine querying for multiple things and I want previous visualization to dissaper, being replaced
# by latest results. we might do this transparently so agent queries are also visible in real time
#

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[paul-nechifor]

You're not keeping track of the subscription.

[leshy]

I am very smart and shutdown of module now shuts down transports which unsubscribes these things

https://github.com/dimensionalOS/dimos/pull/1682/changes#diff-2a78d9a0954c76ba1473eb7f8c2a22c05ea722416c4f15869e46435eed1dde54R166

https://github.com/dimensionalOS/dimos/pull/1682/changes#diff-106cdf9cd8240a8e0e920a92eb61398e93a5f2b33f85f8062403c3fbd5938cb6R138

I think this manual resource handling is annoying and fragile, stopping an owner object should propagade down the tree

[paul-nechifor]

stream.live() creates a new Stream. It also needs to be closed.

[leshy]

same as above, shutdown of store propagades to it's streams

store = self.register_disposable(NullStore())