memory2 tf service

[leshy]

we converted recordings to sensor frame, mem now needs a tf service for global mapping

[mintlify]

Preview deployment for your docs. Learn more about Mintlify Previews.

Project	Status	Preview	Updated (UTC)
dimensional	🟢 Ready	View Preview	Jul 3, 2026, 12:34 AM

💡 Tip: Enable Workflows to automatically generate PRs for you.

[codecov]

❌ 1 Tests Failed:

Tests completed	Failed	Passed	Skipped
2551	1	2550	70

View the top 1 failed test(s) by shortest run time

dimos.e2e_tests.test_dimsim_path_replaning::test_path_replanning

Stack Traces | 230s run time

lcm_spy = <dimos.e2e_tests.lcm_spy.LcmSpy object at 0x7ef275f51d90>
start_blueprint = <function start_blueprint.<locals>.set_name_and_start at 0x7ef273915300>
dim_sim = <dimos.e2e_tests.dim_sim_client.DimSimClient object at 0x7ef27631b4d0>
direct_cmd_vel_explorer = <dimos.simulation.mujoco.direct_cmd_vel_explorer.DirectCmdVelExplorer object at 0x7ef2778ef6b0>
spawn_wall_on_pose = <function spawn_wall_on_pose.<locals>.spawn at 0x7ef2739160c0>

    @pytest.mark.self_hosted_large
    def test_path_replanning(
        lcm_spy, start_blueprint, dim_sim, direct_cmd_vel_explorer, spawn_wall_on_pose
    ) -> None:
        start_blueprint(
            "--dimsim-scene=empty",
            "run",
            "unitree-go2-agentic",
            simulator="dimsim",
        )
        lcm_spy.save_topic(".../McpClient/on_system_modules/res")
        lcm_spy.wait_for_saved_topic(".../McpClient/on_system_modules/res", timeout=1200.0)
    
        # robot spawns at (3, 2)
    
        # side wall
        dim_sim.add_wall(2, -2.5, 12, -2.5)
        # other side wall
        dim_sim.add_wall(2, 3.5, 12, 3.5)
        # back wall (behind robot)
        dim_sim.add_wall(2, -2.5, 2, 3.5)
        # forward wall (far end)
        dim_sim.add_wall(12, -2.5, 12, 3.5)
        # dividing wall at x=7 with doors at y=[-1.5,-0.5] and y=[1.5,2.5]
        dim_sim.add_wall(7, -2.5, 7, -1.5)
        dim_sim.add_wall(7, -0.5, 7, 1.5)
        dim_sim.add_wall(7, 2.5, 7, 3.5)
    
        direct_cmd_vel_explorer.linear_speed = 0.8
        direct_cmd_vel_explorer.follow_points([(10, 2), (2.5, 2), (3, 2)])
    
        # When the robot comes within 1.5 m of the left door's centre, drop a wall
        # in the opening so the planner has to bail out and route through the
        # right door at y=-1 instead.
        spawn_wall_on_pose(
            point=(7, 2),
            threshold=1.5,
            wall=(7, 1.5, 7, 2.5),
        )
    
        dim_sim.publish_goal(10.913, 0.588)
    
>       lcm_spy.wait_until_odom_position(10.913, 0.588, threshold=1, timeout=120)

dim_sim    = <dimos.e2e_tests.dim_sim_client.DimSimClient object at 0x7ef27631b4d0>
direct_cmd_vel_explorer = <dimos.simulation.mujoco.direct_cmd_vel_explorer.DirectCmdVelExplorer object at 0x7ef2778ef6b0>
lcm_spy    = <dimos.e2e_tests.lcm_spy.LcmSpy object at 0x7ef275f51d90>
spawn_wall_on_pose = <function spawn_wall_on_pose.<locals>.spawn at 0x7ef2739160c0>
start_blueprint = <function start_blueprint.<locals>.set_name_and_start at 0x7ef273915300>

dimos/e2e_tests/test_dimsim_path_replaning.py:60: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
dimos/e2e_tests/lcm_spy.py:182: in wait_until_odom_position
    self.wait_for_message_result(
        predicate  = <function LcmSpy.wait_until_odom_position.<locals>.predicate at 0x7ef273916a20>
        self       = <dimos.e2e_tests.lcm_spy.LcmSpy object at 0x7ef275f51d90>
        threshold  = 1
        timeout    = 120
        x          = 10.913
        y          = 0.588
dimos/e2e_tests/lcm_spy.py:168: in wait_for_message_result
    self.wait_until(
        event      = <threading.Event at 0x7ef275eb3b30: unset>
        fail_message = 'Failed to get to position x=10.913, y=0.588'
        listener   = <function LcmSpy.wait_for_message_result.<locals>.listener at 0x7ef273916ac0>
        predicate  = <function LcmSpy.wait_until_odom_position.<locals>.predicate at 0x7ef273916a20>
        self       = <dimos.e2e_tests.lcm_spy.LcmSpy object at 0x7ef275f51d90>
        timeout    = 120
        topic      = '/odom#geometry_msgs.PoseStamped'
        type       = <class 'dimos.msgs.geometry_msgs.PoseStamped.PoseStamped'>
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

self = <dimos.e2e_tests.lcm_spy.LcmSpy object at 0x7ef275f51d90>

    def wait_until(
        self,
        *,
        condition: Callable[[], bool],
        timeout: float,
        error_message: str,
        poll_interval: float = 0.1,
    ) -> None:
        start_time = time.time()
        while time.time() - start_time < timeout:
            if condition():
                return
            time.sleep(poll_interval)
>       raise TimeoutError(error_message)
E       TimeoutError: Failed to get to position x=10.913, y=0.588

condition  = <bound method Event.is_set of <threading.Event at 0x7ef275eb3b30: unset>>
error_message = 'Failed to get to position x=10.913, y=0.588'
poll_interval = 0.1
self       = <dimos.e2e_tests.lcm_spy.LcmSpy object at 0x7ef275f51d90>
start_time = 1783078207.7808752
timeout    = 120

dimos/e2e_tests/lcm_spy.py:105: TimeoutError

To view more test analytics, go to the Test Analytics Dashboard
_{📋 Got 3 mins? Take this short survey to help us improve Test Analytics.}

[greptile-apps]

Greptile Summary

This PR introduces StreamTF, a read-only TF service backed by a recorded tf stream, so the mapping CLI can register sensor-frame point clouds into a world frame during offline replay. It also replaces the --use-tf boolean with automatic world-frame detection (_detect_world) and a new --frame / --tf-tolerance pair, moves get_pose up to TFSpec, and adds a TFLookup protocol for type-safe loose coupling.

• dimos/memory2/tf.py — new StreamTF class: demand-loads a sliding window of TF data from the store into an in-memory MultiTBuffer cache; eviction is manual on window miss.
• dimos/mapping/utils/cli/map.py — world-frame auto-detection probes world/map/odom via the dataset's tf stream; sensor-frame clouds are registered per-frame, world-frame clouds pass through verbatim; dedup and trajectory now use the TF-derived position instead of the stored obs.pose.
• dimos/protocol/tf/tf.py — TFLookup protocol added; get_pose promoted from PubSubTF to TFSpec; None tolerance bug in MultiTBuffer.get_transform fixed by defaulting to buffer_size.

Confidence Score: 5/5

Safe to merge; all changed paths are well-tested and the two observations raised are optimization notes without correctness impact.

The core caching and eviction logic in StreamTF is correct for the single-threaded CLI use case, the world-frame detection handles all edge cases (null tf_buf, user-provided --frame, cloud already in a world frame), and the test suite now parametrizes over live, memory-backed, and sqlite-backed backends covering chain composition, time tolerance, eviction, and protocol conformance. The two open points — I/O under lock and redundant register calls — are non-blocking quality concerns.

No files require special attention for correctness; dimos/memory2/tf.py and dimos/mapping/utils/cli/map.py carry the non-blocking quality notes.

Important Files Changed

Filename	Overview
dimos/memory2/tf.py	New StreamTF class wrapping a recorded tf stream for offline replay; caching and eviction logic is correct for single-threaded use, with minor concerns around holding the lock during I/O and the 10 s lookback fallback for time_tolerance=None.
dimos/mapping/utils/cli/map.py	Replaces the --use-tf boolean with automatic world-frame detection and per-frame tf registration; logic is correct but each kept frame incurs a redundant second register() call during accumulation.
dimos/protocol/tf/tf.py	Adds TFLookup protocol and moves get_pose from PubSubTF to TFSpec; also fixes the None tolerance passed to TBuffer.get by defaulting to buffer_size instead.
dimos/protocol/tf/test_tf.py	Tests refactored into a parametrized grid fixture covering live, memory-backed, and sqlite-backed StreamTF; good coverage of caching, eviction, chain composition, and the TFLookup protocol conformance check.
docs/usage/transforms.md	Minor doc fix: corrects the relative path to tf.py in the Markdown link.

Sequence Diagram

%%{init: {'theme': 'neutral'}}%%
sequenceDiagram
    participant CLI as map.py CLI
    participant STF as StreamTF
    participant Store as SqliteStore / Stream
    participant MTB as MultiTBuffer (cache)

    CLI->>Store: stream("tf", TFMessage)
    CLI->>STF: StreamTF.from_store(store)
    CLI->>STF: "get("world", cloud_frame, ts=first_obs.ts)"
    STF->>STF: _ensure(ts-10, ts+0)
    STF->>Store: stream.at(center, radius)
    Store-->>STF: TFMessage observations
    STF->>MTB: receive_transform(...)
    STF->>MTB: super().get(...)
    MTB-->>STF: "Transform | None"
    STF-->>CLI: world frame detected

    loop for each obs in lidar (dedup)
        CLI->>STF: register(obs) via _position()
        STF->>STF: _ensure(obs.ts-10, obs.ts) — cache hit or evict+reload
        STF->>MTB: super().get(world, frame_id, obs.ts)
        MTB-->>STF: "Transform | None"
        STF-->>CLI: "position tuple | None"
    end

    loop for each kept obs in _accumulate()
        CLI->>STF: register(obs)
        STF->>MTB: super().get(...) — typically cache hit
        MTB-->>STF: Transform
        STF-->>CLI: Transform (used to register cloud into world frame)
    end

Loading

%%{init: {'theme': 'base', 'themeVariables': {"darkMode": true, "background": "#0d1117", "primaryColor": "#21262d", "primaryTextColor": "#e6edf3", "primaryBorderColor": "#8b949e", "lineColor": "#8b949e", "textColor": "#e6edf3", "edgeLabelBackground": "#161b22", "actorBkg": "#21262d", "actorBorder": "#8b949e", "actorTextColor": "#e6edf3", "actorLineColor": "#8b949e", "signalColor": "#8b949e", "signalTextColor": "#e6edf3", "noteBkgColor": "#373320", "noteBorderColor": "#d4a72c", "noteTextColor": "#f0e6c0", "labelBoxBkgColor": "#21262d", "labelBoxBorderColor": "#8b949e", "labelTextColor": "#e6edf3", "loopTextColor": "#e6edf3", "activationBkgColor": "#30363d", "activationBorderColor": "#8b949e"}}}%%
sequenceDiagram
    participant CLI as map.py CLI
    participant STF as StreamTF
    participant Store as SqliteStore / Stream
    participant MTB as MultiTBuffer (cache)

    CLI->>Store: stream("tf", TFMessage)
    CLI->>STF: StreamTF.from_store(store)
    CLI->>STF: "get("world", cloud_frame, ts=first_obs.ts)"
    STF->>STF: _ensure(ts-10, ts+0)
    STF->>Store: stream.at(center, radius)
    Store-->>STF: TFMessage observations
    STF->>MTB: receive_transform(...)
    STF->>MTB: super().get(...)
    MTB-->>STF: "Transform | None"
    STF-->>CLI: world frame detected

    loop for each obs in lidar (dedup)
        CLI->>STF: register(obs) via _position()
        STF->>STF: _ensure(obs.ts-10, obs.ts) — cache hit or evict+reload
        STF->>MTB: super().get(world, frame_id, obs.ts)
        MTB-->>STF: "Transform | None"
        STF-->>CLI: "position tuple | None"
    end

    loop for each kept obs in _accumulate()
        CLI->>STF: register(obs)
        STF->>MTB: super().get(...) — typically cache hit
        MTB-->>STF: Transform
        STF-->>CLI: Transform (used to register cloud into world frame)
    end

Loading

_{Reviews (4): Last reviewed commit: "Merge branch 'feat/ivan/memtf' of github..." | Re-trigger Greptile}