fix(perception): use pos_x/pos_y/pos_z metadata keys in spatial location queries

[hvent90]

Contribution path

• Small, safe change that does not need a tracking issue

Problem

SpatialVectorDB.query_by_location() and get_all_locations() filter stored frames by metadata keys "x"/"y"/"z", but frames are stored with keys pos_x/pos_y/pos_z (see the metadata dict built in SpatialMemory._process_frame). The keys never match, so both methods always return empty results against real spatial-memory data.

Solution

Read pos_x/pos_y/pos_z in both methods.

Added dimos/perception/test_spatial_vector_db.py: stores frames in an in-memory SpatialVectorDB using the exact metadata shape _process_frame writes, and asserts location queries find them. The test fails against the old
code and runs in the default (fast) CI suite.

How to Test

uv run pytest dimos/perception/test_spatial_vector_db.py

Also verified against real recorded data:

First I populated the chromadb: dimos --replay --replay-db go2_china_office run unitree-go2-spatial --disable perceive-loop-skill (--disable because the blueprint's PerceiveLoopSkill requires an agent module it doesn't include — separate issue). Let it log Stored frame at position (…) for ~60s, then stop it.

Then, I ran this script:

uv run python - <<'EOF'
import chromadb
from dimos.perception.spatial_vector_db import SpatialVectorDB

client = chromadb.PersistentClient(path="assets/output/memory/spatial_memory/chromadb_data")
db = SpatialVectorDB(collection_name="spatial_memory", chroma_client=client)
frames = db.image_collection.get(include=["metadatas"])
print(len(frames["ids"]), "frames stored")          # non-zero — the data exists
print("sample metadata:", frames["metadatas"][0])   # keys are pos_x/pos_y/pos_z
print("get_all_locations:", db.get_all_locations()) # [] — the bug
print("query (infinite radius):", db.query_by_location(0, 0, radius=1e9))  # still []
EOF

Running it on main produced empty query results. Running it on this feature branch produced results.

AI assistance

My associate (Fable 5) discovered the issue, proposed a fix, and implemented it. I then reviewed the output and manually ran the verification against the chromadb as outlined in the How to Test section.

[greptile-apps]

Greptile Summary

This PR fixes spatial location queries to use the metadata keys written for stored frames. The main changes are:

• Read pos_x, pos_y, and pos_z in spatial location queries.
• Add tests for radius filtering and missing position metadata.
• Add coverage for retrieving all stored frame locations.

Confidence Score: 5/5

This looks safe to merge.

• The updated keys match the metadata written by both spatial frame ingestion paths.
• The tests cover location filtering, missing coordinates, and full location retrieval.
• No blocking issues were found in the changed code.

Important Files Changed

Filename	Overview
dimos/perception/spatial_vector_db.py	Aligns location-query metadata reads with the keys used by spatial frame producers.
dimos/perception/test_spatial_vector_db.py	Tests nearby and distant frames, missing position metadata, and complete location retrieval.

_{Reviews (2): Last reviewed commit: "test(perception): cover frames stored wi..." | Re-trigger Greptile}

[codecov]

Codecov Report

❌ Patch coverage is 96.66667% with 1 line in your changes missing coverage. Please review.
✅ All tests successful. No failed tests found.

Files with missing lines	Patch %	Lines
dimos/perception/spatial_vector_db.py	83.33%	0 Missing and 1 partial ⚠️

@@            Coverage Diff             @@
##             main    #3123      +/-   ##
==========================================
- Coverage   72.99%   72.77%   -0.23%     
==========================================
  Files        1031     1034       +3     
  Lines       93177    93964     +787     
  Branches     8479     8609     +130     
==========================================
+ Hits        68018    68384     +366     
- Misses      22835    23251     +416     
- Partials     2324     2329       +5     

Flag	Coverage Δ
OS-ubuntu-24.04-arm	66.30% <96.66%> (+0.14%)	⬆️
OS-ubuntu-latest	68.58% <96.66%> (+0.12%)	⬆️
Py-3.10	68.58% <96.66%> (+0.12%)	⬆️
Py-3.11	68.57% <96.66%> (+0.12%)	⬆️
Py-3.12	68.57% <96.66%> (+0.11%)	⬆️
Py-3.13	68.58% <96.66%> (+0.12%)	⬆️
Py-3.14	68.58% <96.66%> (+0.11%)	⬆️
Py-3.14t	?
SelfHosted-macOS	?

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
dimos/perception/test_spatial_vector_db.py	100.00% <100.00%> (ø)
dimos/perception/spatial_vector_db.py	61.94% <83.33%> (+20.35%)	⬆️

... and 42 files with indirect coverage changes

🚀 New features to boost your workflow:

• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[paul-nechifor]

Hi @hvent90 . Thank you for the contribution. However, it looks like query_by_location is dead code. Could you instead please delete that method and the query_by_location method which uses it.