Yimo/feat/graphviz rpc

[poorwym]

Problem

Blueprint topology was difficult to inspect from the CLI. The existing Graphviz renderer only showed stream connections and did not visualize RPC dependencies.

Solution

• Added dimos graph to generate a Blueprint topology SVG.
• Added RPC Spec/function relationships using dashed edges.
• Added a dark SVG background for standalone viewing. This needs more discussions. I'm not sure whether should add a background instead of transparent background.
• Kept static rendering isolated from runtime transports.
• Updated Rerun parsing to support long RPC labels.

How to Test

dimos graph unitree-go2-agentic --output unitree-go2-agentic.svg

uv run pytest dimos/cli/commands/test_graph.py -q

[mintlify]

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

Project	Status	Preview	Updated (UTC)
dimensional	🔴 Failed	–	Aug 26, 2026, 4:58 AM

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

[greptile-apps]

Greptile Summary

This change adds a dimos graph command for rendering Blueprint stream and RPC relationships. A runtime reproduction with two namespaced instances of the same caller and provider classes showed that the graph merges their separate RPC relationships into a single class-name-based topology. Update the DOT RPC identities to retain Blueprint atom names before merging.

Confidence Score: 4/5

The graph command is not safe to merge as-is for deployments that contain multiple named instances of the same module class, because it can display incorrect RPC provider relationships.

A focused executed reproduction exercised the exact multi-instance resolution path and directly observed the renderer discard the instance identities in its emitted DOT.

Files Needing Attention: dimos/core/introspection/blueprint/dot.py needs to use Blueprint atom names for RPC caller and provider identities, with a regression test for same-class namespaced instances.

T-Rex Logs

What T-Rex did

• Produced a proof for the first posted P1 finding and linked it to the review comment.
• Produced a proof for the second posted P1 finding and linked it to the review comment.
• Before capture, per-instance RPC dependencies were resolved separately, with robot0/caller --service--> robot0/provider and robot1/caller --service--> robot1/provider.
• After capture, both RpcConnection records became identical (caller='Caller', provider='Provider'), and the DOT output collapsed the two relationships into a single Caller/Provider module.
• Uploaded artifacts including the executable reproduction source and the before/after RPC logs and DOT output to support the proofs.

_{Ran code and verified through T-Rex}

Comments Outside Diff (1)

1. General comment

   DOT RPC renderer collapses distinct same-class Blueprint instances

   • Bug
     • Two independent namespaced RPC relationships (robot0/caller to robot0/provider and robot1/caller to robot1/provider) render as duplicate edges through one rpc_Caller_service contract between one Caller and one Provider DOT node.
   • Cause
     • _resolve_rpc_connections stores atom.module.__name__ and provider.module.__name__ in RpcConnection, discarding BlueprintAtom.name, which is the unique instance identity used for ModuleRef resolution.
   • Fix
     • At dimos/core/introspection/blueprint/dot.py:106-107, use caller=atom.name and provider=provider.name. The renderer's existing RPC node ID and edge construction will then retain unique namespaced identities; add a regression test with two namespaced same-class caller/provider pairs.

   _{Ran code and verified through T-Rex}

_{Reviews (1): Last reviewed commit: "add rpc connections" | Re-trigger Greptile}

[greptile-apps]

RPC instance identities collapse

_resolve_rpc_connections stores atom.module.__name__ and provider.module.__name__, discarding the Blueprint atom names that distinguish instances. Two independently resolved same-class relationships, such as robot0/caller → robot0/provider and robot1/caller → robot1/provider, therefore render as one Caller → rpc_Caller_service → Provider topology with duplicate edges. Use atom.name and provider.name for the connection identities, and add a regression case with multiple namespaced instances of the same caller/provider classes.

Artifacts

Executable reproduction source for two namespaced same-class RPC pairs

• Creates two independent same-class Blueprint instance pairs, resolves their ModuleRef providers, renders DOT, and asserts the observed collapse.

Resolved per-instance RPC dependencies before DOT rendering

• The executed script shows `robot0/caller` resolves to `robot0/provider` and `robot1/caller` resolves to `robot1/provider`, establishing the distinct relationships the renderer must preserve.

DOT output with same-class RPC instances collapsed

• The executed renderer produces two identical class-name RpcConnections and one repeated `rpc_Caller_service` DOT identity, confirming the topology collapse.

_{Ran code and verified through T-Rex}

[codecov]

❌ 1 Tests Failed:

Tests completed	Failed	Passed	Skipped
4912	1	4911	74

View the full list of 1 ❄️ flaky test(s)

dimos.e2e_tests.test_dimsim_spatial_memory::test_go_to_the_bed

Flake rate in main: 15.43% (Passed 137 times, Failed 25 times)

Stack Traces | 318s run time

lcm_spy = <dimos.e2e_tests.lcm_spy.LcmSpy object at 0x724ae07355b0>
start_blueprint = <function start_blueprint.<locals>.set_name_and_start at 0x724ae0450180>
human_input = <function human_input.<locals>.send_human_input at 0x724ae0450900>
dim_sim = <dimos.e2e_tests.dim_sim_client.DimSimClient object at 0x724be8e03ef0>
explore_house = <function explore_house.<locals>.explore at 0x724ae0450fe0>

    @pytest.mark.self_hosted_large
    def test_go_to_the_bed(lcm_spy, start_blueprint, human_input, dim_sim, explore_house) -> None:
        start_blueprint(
            "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)
    
        explore_house()
    
        human_input("go to the bed")
    
>       lcm_spy.wait_until_odom_position(-3.567, -1.332, threshold=2, timeout=180)

dim_sim    = <dimos.e2e_tests.dim_sim_client.DimSimClient object at 0x724be8e03ef0>
explore_house = <function explore_house.<locals>.explore at 0x724ae0450fe0>
human_input = <function human_input.<locals>.send_human_input at 0x724ae0450900>
lcm_spy    = <dimos.e2e_tests.lcm_spy.LcmSpy object at 0x724ae07355b0>
start_blueprint = <function start_blueprint.<locals>.set_name_and_start at 0x724ae0450180>

dimos/e2e_tests/test_dimsim_spatial_memory.py:32: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
dimos/e2e_tests/lcm_spy.py:167: in wait_until_odom_position
    self.wait_for_message_result(
        predicate  = <function LcmSpy.wait_until_odom_position.<locals>.predicate at 0x724ae0467740>
        self       = <dimos.e2e_tests.lcm_spy.LcmSpy object at 0x724ae07355b0>
        threshold  = 2
        timeout    = 180
        x          = -3.567
        y          = -1.332
dimos/e2e_tests/lcm_spy.py:153: in wait_for_message_result
    wait_until(
        event      = <threading.Event at 0x724ae12b9730: unset>
        fail_message = 'Failed to get to position x=-3.567, y=-1.332'
        listener   = <function LcmSpy.wait_for_message_result.<locals>.listener at 0x724ae0451300>
        predicate  = <function LcmSpy.wait_until_odom_position.<locals>.predicate at 0x724ae0467740>
        self       = <dimos.e2e_tests.lcm_spy.LcmSpy object at 0x724ae07355b0>
        timeout    = 180
        topic      = '/odom#geometry_msgs.PoseStamped'
        type       = <class 'dimos.msgs.geometry_msgs.PoseStamped.PoseStamped'>
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

predicate = <bound method Event.is_set of <threading.Event at 0x724ae12b9730: unset>>

    def wait_until(
        predicate: Callable[[], bool],
        *,
        timeout: float,
        interval: float = 0.1,
        message: str | None = None,
    ) -> None:
        """Poll ``predicate`` until it returns truthy or ``timeout`` elapses."""
        deadline = time.monotonic() + timeout
        while time.monotonic() < deadline:
            if predicate():
                return
            time.sleep(interval)
>       raise TimeoutError(message or f"Timed out after {timeout}s waiting for condition")
E       TimeoutError: Failed to get to position x=-3.567, y=-1.332

deadline   = 4218138.460349792
interval   = 0.1
message    = 'Failed to get to position x=-3.567, y=-1.332'
predicate  = <bound method Event.is_set of <threading.Event at 0x724ae12b9730: unset>>
timeout    = 180

.../utils/testing/waiting.py:35: 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.}

[leshy]

looks great can we make rpc render an optional boolean then ready to merge