2 Commits

Author	SHA1	Message	Date
Marc Schier	cef83550cb	OPC UA Vision: a vision-guided bin-picking demo an LLM agent can see and drive (#4235) ... # Description Implements **OPC UA — Vision** and uses it to build the demo the companion specifications were drafted for: an eye-in-hand camera watches a bin of parts, a language model *looks at the frame*, decides which part matches the instruction, and commands the robot through Robot Intent — with the OpenUSD viewer showing the arm move and the bin empty out. > **Draft, and stacked.** [#4195](https://github.com/OPCFoundation/UA-.NETStandard/pull/4195) has merged and master is merged in, so what is left on top of `master` is the Vision work plus the unmerged `marcschier/ai-model-management` branch, which the demo needs for the model/dataset/deployment provenance. Review the Vision commits; the AI Model Management work lands with its own PR. ## What is here **The specification, implemented.** `Opc.Ua.Vision` is generated from the spec's own `Opc.Ua.Vision.NodeSet2.xml`, vendored byte-identically (SHA-256 `79C592C6…`, verified against a fresh download and stored LF-preserved like the Robot Intent NodeSet). `Opc.Ua.Vision.Server` builds the §4.2 discovery object, sensors, the coordinate-frame tree, calibrations, media endpoints, inference pipelines and the §9 feedback methods, with rendering and inference behind injectable providers. `Opc.Ua.Vision.Client` is built on the generated ObjectType proxies and composes transforms across the frame tree, which is what turns a detection in camera coordinates into a pose a robot can act on. **Two perception paths behind one contract**, which is exactly what `InferenceLocationEnum` is for. `OnServer` is a deterministic detector derived from the stage's ground truth — no model, no network, no GPU, so CI can exercise the whole loop. `EdgeOffServer` is the agent: it sees the frame over MCP and calls `SubmitDetections`, and the Server publishes results it did not compute. A client reads `DetectionResultType` identically either way. **The agent's eyes.** `tools/Opc.Ua.Mcp.Vision` adds 22 `vision_*` tools, which with the four shared connection tools makes the `vision` profile 26; `vision_get_frame` returns the camera image as an MCP `ImageContentBlock`, so the model looks at pixels rather than reading a description of them. Tool profiles now compose — `--profile vision,robotics` yields 62 tools — so one agent can both see and act without dragging in the full 136-tool catalogue. **The cell.** `samples/Robotics/BinPickingCell` follows the spec's own Robotics-Vision Addendum: the `world → robot_base → flange → gripper_tcp` frame tree with `camera_eih` on the flange, the `HandEye` extrinsics and the `Intrinsics612x512` intrinsics — the addendum's 2448×2048 calibration scaled to the 4×4-binned grid the camera actually delivers. `samples/Robotics/BinPickingClient` hosts the composed MCP catalogue and optionally the viewer. **Documentation.** `docs/Vision.md` (~800 lines) covers the model, hosting, the build context, every topology builder, both perception paths, feedback, the MCP profile and the limitations; linked from `docs/README.md`, with the profile table in `docs/McpServer.md` and READMEs for both samples. ## Evidence, not assertions Each of these was measured on a running system rather than assumed: | Claim | Evidence | |---|---| | The camera sees the bin | 612×512 frame with the bin and all five parts in view — red `(212,47,47)`, green `(61,193,71)`, blue `(54,85,218)`, yellow `(215,196,58)`, orange `(209,119,55)` | | An agent can act on what it sees | Over MCP, `vision_get_frame` returns a 612×512 PNG that decodes from base64 to 1,254,051 bytes, the descriptor beside the detections says 612×512, and all five `BoundingBox2D` centres land **inside** the frame on the pixel colour their own detection claims | | The geometry is right | A detection composed `camera_eih → world` lands on the authored position, residual **0.0000 m** — on both the on-server and off-server paths | | The detector tracks the world | Placing a part drops the next scan from 5 detections to 4, with that part absent | | The robot obeys the agent | Scripted run took command authority and completed a pick and a place, both `Succeeded` | | Tool catalogues are honest | Counted from a running server: vision 26, robotics 40, `vision,robotics` 62, full 136, **zero duplicates** | | Every node is browsable by NodeId | Eight representative NodeIds — both folders, a pipeline, a frame, the sensor, its Calibrations folder, the HandEye calibration, the Media folder — all `Good` | That tool-count check matters because the profile tests collapse names into a `HashSet` and *cannot see* a duplicate registration — so the new tests count registrations, not names. ## Things that were wrong, and are not now Writing the tests found three defects of one kind — code that answered instead of refusing. A zero-norm quaternion was silently rewritten to identity, so a caller with an all-zero orientation got a confident wrong grasp pose. `SubmitCorrection` rewrote a null `ResultId` to empty and called the sink anyway. A bad `StageIdentifier` **killed the host process** inside native `UsdStage.Open`, via an exception .NET cannot catch. All three now refuse loudly and are pinned by tests. **Nodes that existed but could not be reached.** The node manager indexes the Vision root before configurators run, so everything the fluent builder grafted on afterwards — both folders, the pipelines with their Results and Feedback, the frames, the sensors with their optics, calibrations and media endpoints — was never added to the index. Browsing *forward from a parent* worked, because that walks `NodeState.Children` in memory; browsing any of those nodes *by its own NodeId* returned `BadNodeIdUnknown`, which is how an ordinary client and the MCP discovery tools navigate. Registration is now deferred and flushed after each configurator. The first fix covered only the DI path and left the documented non-DI fallback broken in exactly the same way, so `ConfigureVisionAsync` now exists and registers what it builds. That defect survived a green suite because every test browsed children. The new tests assert by NodeId, and were checked against the broken code — with the flush removed, they fail. **Every Method was uncallable.** The Server passed 426 unit tests while not one of its Methods could be invoked over a real session — `RunInference` returned `Bad_TooManyArguments`, and so did every other Method that takes arguments, which is all of them but `StartContinuous` and `Stop`. Three independent faults, each sufficient on its own: the Method nodes were never created (they are Optional children, and the dispatcher guards every attachment with a null check, so it silently attached nothing); nothing declared the arguments, so the stack concluded none were expected; and `MethodDeclarationId` pointed at a synthesised NodeId rather than the NodeSet declaration a client calls with, so the lookup missed the instance Method entirely. The builder now creates the Methods, the `Results` folder and the `Feedback` object when a provider or sink is configured, declares every signature, and names the right declaration. Unit tests could not have caught this: they invoke the handler delegates directly and never go through `MethodState.Call`. It took an end-to-end test over a real session. Relatedly, children created by the generated `CreateOrReplace` helpers carry no `ReferenceTypeId`, so nothing could reach them — fixed once in the build context rather than at the several dozen call sites. **And one where I was wrong.** Earlier in this branch I reversed my own agents' decision and made `SubmitDetections` accept an empty detection set, reasoning that "I looked and the bin is empty" is a correct observation and refusing it forces a correct agent to invent a detection. The reasoning is sound; it is also an argument *against* the specification rather than an implementation of it. §9.5 is normative and explicit — its StatusCode table lists `Bad_InvalidArgument` for "`Detections` empty; or `SubmitCorrection` supplies both or neither corrected array". Worse, **none of §9.5's argument rules were enforced anywhere** — the dispatcher forwarded every submission straight to the feedback sink, so conformance depended on whichever sink a host installed. They are Server obligations, so they now live in the dispatcher, ahead of the sink, proven by strict mocks that would throw if the sink were consulted. The usability problem is real and does not go away by conforming, so it is raised where it can be fixed: [opcua-drafts#70](https://github.com/marcschier/opcua-drafts/issues/70). ## Tracking the updated drafts The Vision, Robot Intent and AI Model Management NodeSets on `opcua-drafts` `main` have moved, partly because of feedback this branch raised. All three are re-vendored byte-identically — verified by comparing git blob hashes against the upstream files rather than by eye — and the implementations follow. **Two statements that were previously inexpressible now have a wire representation.** `SubmitDetections` takes `SceneIsEmpty` and `SubmitCorrection` takes `RetractAll`, so an agent can report "I examined this frame and there is nothing in it" — the terminating condition of the bin-picking task — and can retract a false positive by correcting a result down to nothing. Both are checked in both directions, because the flag is exactly what separates a deliberate empty observation from a lost payload: an empty array without the flag is refused, and the flag with an array attached is refused as two contradictory claims about one frame. Implemented in the dispatcher, the client, the MCP tools the agent drives, and the sample cell. Also absorbed: `LampType` and `LightingMode` became enums (§66), `ProfileName` became `DefaultProfileName` on the endpoint (distinct from the `ProfileName` a caller passes to `GetStreamEndpoint`, which keeps its name), AI `Invoke`/`InvokeAsync` gained `PayloadUri` under the same exactly-one rule, and AI `ListModels` gained a `ContinuationPoint` — because `MaxResults` alone puts everything past the bound permanently out of reach. **OpenUSD is on `0.9.0-alpha`**, verified on the running cell rather than by restore alone: D3D12 backend, 612×512 frame, the bin and all five parts in view and detected — the same result `0.8.0-alpha` produced. The one local workaround that survives is `VisionMethodArguments`, and its reason has changed rather than gone away: the builder materialises these Methods as Optional children through `CreateOrReplace`, which constructs the state object directly and never runs the generated factory that now carries the signatures. Checked rather than assumed — removing the declarations fails five of the eight `VisionMethodSurfaceTests`. ## Review feedback Three threads, all addressed in `0fe2803a3`: - **The AI Model Management work is now a library family, not a sample.** `src/Opc.Ua.AI` (model), `.Inference` (`IInferenceBackend` and backends), `.Server` (node manager) and `.Client` (discovery, reads, calls, artefact transfer), in the same shape as the Robotics and Vision families; the samples become `samples/AI/ModelManagementServer` and `ModelManagementClient`. The client library needed real work rather than relocation: its browsing surface was `private` inside the sample's scenario runner, so moving the file alone would have produced a library nobody could call. - **No cloud-vendor SDK is referenced anywhere.** Inference goes through `Microsoft.Extensions.AI` — an `IChatClient` is implemented by hosted services and on-device runtimes alike, which is exactly the property clause 8.1 asserts. `Azure.Identity` is gone: workload identity is not an Azure feature, every platform projects a token into a file and each SDK reads that file behind its own API, so the resolver reads it directly. The comment claiming `Microsoft.Extensions.AI` would force a `System.Text.Json` bump was stale — it pins 10.0.10, which this repository already centralises on. - **`BinPickingCell` and `BinPickingClient` moved into `samples/Robotics/`**, which is where they belonged: the cell was already linking the arm, its kinematics and both USD assets out of `IntentEnabledRobot`. ## Upstream Work that belonged elsewhere went elsewhere rather than being worked around locally: - **`openusd-dotnet` [#13](https://github.com/marcschier/openusd-dotnet/pull/13)** — merged and shipped in `0.8.0-alpha`. `SilkFrameCapture.Capture` rendered only the first capture of a session and silently returned a blank frame for every one after it. Also filed [#14](https://github.com/marcschier/openusd-dotnet/issues/14) (no camera-by-prim API) and [#15](https://github.com/marcschier/openusd-dotnet/issues/15) (skipped tests reported as passed — which fooled me for several runs). Both are fixed; this branch takes `0.8.0-alpha` and deletes two workarounds, including 143 lines of hand-rolled projection maths. - **`opcua-drafts` [#66](https://github.com/marcschier/opcua-drafts/issues/66)–[#69](https://github.com/marcschier/opcua-drafts/issues/69)** — a close read of the spec produced 22 candidate ambiguities; 12 were discarded as misreadings with clause citations and the remaining 10 filed. One is concrete: five members exist in the NodeId CSV that no clause documents. - **`opcua-drafts` [#70](https://github.com/marcschier/opcua-drafts/issues/70)** — §9.5 could not express an empty observation or a false-positive retraction. **Adopted**, with the explicit-flag resolution argued for: `SubmitDetections` gained `SceneIsEmpty`, `SubmitCorrection` gained `RetractAll`, and the corrected-array rule relaxed from *exactly one* to *at most one*. - **`opcua-drafts` [#71](https://github.com/marcschier/opcua-drafts/issues/71)** — the Vision NodeSet browse-named all 13 `InputArguments` / `OutputArguments` Properties in the Vision namespace rather than the base namespace, so a stack could not find them and every Method in the specification was uncallable. **Fixed**; this branch now tracks the corrected artefact. ## The defect underneath it all The end-to-end suite was removed earlier in this branch because nine of its fourteen tests failed on one symptom: a client could not open a pipeline's `Feedback` object or enumerate its `Results`, even though browsing the same parent listed both. Restoring it meant explaining that, and the explanation was not in Vision. `RelativePathElement` initialised **`IsInverse` to `true`**. A caller that sets a `ReferenceTypeId` and a `TargetName` — which reads as *"the child reached by this Reference"* — was therefore asking the Server for the **inverse** Reference, and got `Bad_NoMatch` with nothing in the request to suggest why. `ObjectTypeClient.ResolveChildNodeIdAsync` is such a caller, and it is what **every source-generated Optional-child accessor in the stack** is built on — the accessor whose own documentation cites `AlarmConditionTypeClient.GetShelvingStateAsync` as an example. Vision is merely where it surfaced, because Vision puts two Optional children on the path an off-Server agent must take. The sibling helper in `StateMachineTypeClientExtensions` already set `IsInverse = false` by hand, which is the shape of a trap rather than a default. The default is now forward, `ResolveChildNodeIdAsync` says so explicitly, and the existing test that had recorded `true` as *expected* now pins `false` with the reasoning. Of the forty `RelativePathElement` construction sites in the tree, that accessor was **the only one relying on the default at all** — every deliberate caller already set it. Two further defects were found underneath it, both real and both fixed: - The generated `CreateOrReplace` helpers construct a child state object directly and leave **`TypeDefinitionId`** unset as well as `ReferenceTypeId`. A child referenced by nothing cannot be browsed; one with no type definition is a malformed Object that a client filtering by type silently skips — which is exactly what `EnumerateResultsAsync` does. Normalisation moved to `VisionNodeManager`, so it also covers the case the builder cannot see: a result published **at runtime** by an inference provider, long after the address space was built. That path produced results a client could list but not read. - `Opc.Ua.AI` shipped without an `AssemblyInfo`, so the solution built with six `CA1014` warnings. I had initially "fixed" this in the Vision client with Browse-based fallbacks, on the strength of a probe that appeared to show `TranslateBrowsePathsToNodeIds` failing even for `Server → ServerStatus`. That probe was wrong: it used an unqualified `ReferenceTypeIds`, which inside a `Opc.Ua.Vision.*` namespace binds to *Vision's* `ReferenceTypeIds` and yields a nonsense NodeId. Chasing that to the end is what turned up the real cause. **The fallbacks are gone** — they would have hidden a stack-wide bug behind a Vision-shaped workaround, and one of them matched on browse name while ignoring the namespace index. `tests/Opc.Ua.Vision.Intent.Tests` is restored and back in `UA.slnx`, passing **14/14** on its own merit rather than through a client-side workaround. A registration test now pins the pipeline's `Feedback` and `Results` children, which nothing covered before. ## Running it, and what that found The demo's premise is that a language model **looks at the bin and reasons about what it sees**. Running it end to end showed it could not, for reasons no test covered, because every one of them lived between components. **The sample did not compile.** Two `CS8604` nullable errors with `TreatWarningsAsErrors`, so neither the cell server nor the demo could be started at all. Worth recording that my first explanation was wrong: CI *does* build this project - the Linux leg's `net10.0` pass compiles it, log-confirmed - so this is not a coverage gap in the workflow, and no workflow change was made. **`Pick` and `Place` never moved the arm.** They were a delay plus a gripper action, so the operations reported `Succeeded` while nothing travelled. They now solve IK once and interpolate in joint space. **And the parts never moved**: the arm swung but the bin stayed full, because nothing tied the held object to the world state. The connector now creates 11 monitored items where it created 6, and a scripted run delivers `SeqNo=151` across 10 messages where it delivered `SeqNo=3` across 1. **The frame was smuggled through a string field.** The media provider put a base64 *data URI* of the PNG into `VisionImageReferenceDataType.Uri` - a field meant to be a *reference*, while the bytes were already returned in the inline `ByteString`. Every response therefore shipped the image twice, and a 1,311,400-byte PNG became a 1,748,536-character string against a `MaxStringLength` of 65,535: **26.7x over**. That is why raising `MaxByteStringLength`, `MaxArrayLength` and `MaxMessageSize` never helped - none of them govern a `String`. Separately the MCP tool assigned raw bytes where `ImageContentBlock` requires base64, so the wire carried `"data":"�PNG..."` - every non-UTF-8 byte replaced by U+FFFD, destroying the image and inflating the response to 6.8 MB. **The camera was aimed at nothing.** The arm's home pose, the flange scan pose the Vision model declared and the joint angles in the USD stage were three independent claims about where the camera was, and they disagreed - the declared flange orientation pointed the view 86 degrees away from the projection camera. Solving them as one thing exposed two further constraints that are easy to miss: the camera has to sit off the tool axis or it photographs its own gripper, and the IK branch has to be elbow-back or a link parks under the camera and fills the frame with the arm's own upper arm. A third was a regression this branch introduced and `Opc.Ua.Robotics.Tests` caught - aiming a straight-down camera from a point on the base's own X-Z plane puts the wrist on the J4/J6 singularity, so every motion away from home fails to solve. A 15-degree camera roll, free because the camera still looks straight down, gets 25 degrees clear, which is the margin the poses before this branch had. **The frame and the detections were different images.** The sensor declared 2448x2048, the clip endpoint 1280x1024, and the renderer produced 640x512 - not even the same aspect ratio. The ground-truth detector projects through the declared intrinsics, so its boxes lived in 2448x2048 space while an agent was handed a 640x512 picture: *"pick the red cube you can see"* arrived with coordinates that pointed off the image. Everything now agrees at 612x512, an exact 4x4 bin of the native sensor with the intrinsics scaled to match; the native size stays in the model and serial number, where it describes the hardware rather than the image. **`LatestClip` was created and never written**, so it reported `Bad_NoDataAvailable` for the life of the Server and a consumer following the model - read the published frame, call the method only if there is none - never got a frame. The dispatcher now publishes a clip it has just encoded, for every Vision server rather than this sample alone. One genuine API gap turned up on the way: `OpcUaServerOptions` exposes `MaxByteStringLength`, `MaxArrayLength` and `MaxMessageSize` and **no `MaxStringLength`**, so a Server that legitimately needs a long string could not raise the limit through the hosting API at all. Added, with tests. It is not the fix for anything above - that was a design bug - but the missing knob was real. ## Not finished - **Two projects sit below the repository's 80% line coverage.** `Opc.Ua.Vision.OpenUsd` is at 71.1% — what is left only executes with a real OpenUSD stage and plugin tree, which is not on every build agent. `Opc.Ua.AI.Inference` is now at **86.7%**, up from 12.9%: the code this branch *wrote* is covered by 25 tests, and `RestChatCompletionsBackend` — previously the whole of the shortfall — is covered by 18 more against a stubbed `HttpMessageHandler`, exercising success, refusal, throttling, retry-after, timeout and cancellation without a network. Everything else clears the bar — `Opc.Ua.Vision.Server` 92.1%, `Opc.Ua.AI` 91.4%, `.Vision.Client` 88.1%, `Opc.Ua.Vision` 85.7%, `Opc.Ua.AI.Server` 80.1%, all measured with coverlet. The CI coverage gate reports itself as advisory and non-blocking. - **`SamplesCollected` is not counted here, and `docs/Vision.md` now says so.** Section 9.4 requires a Server to count a negative example (`SceneIsEmpty` / `RetractAll` carrying a `GroundTruthLabel`) exactly as it counts one carrying geometry. `VisionNodeManager` does not, and structurally cannot: `SamplesCollected` is a property of `LearningJobType`, which the *AI Model Management* companion defines, and Vision reaches the job through a `NodeId` value rather than a Reference **precisely so this model takes no dependency on the model that defines it**. What the Server does guarantee is that the negative example survives the hop intact — both flags are carried verbatim to `IVisionFeedbackSink` — so a host that binds a learning job has everything it needs to satisfy the clause on the counter it owns. Stated in the limitations rather than left implied. ## Related Issues Was stacked on #4195, now merged. No tracking issue yet for the Vision implementation itself — happy to open one if you would like the design recorded as an ADR before this goes further. ## Checklist - [x] I have signed the [CLA](https://opcfoundation.org/license/cla/ContributorLicenseAgreementv1.0.pdf) and read the [CONTRIBUTING](https://github.com/OPCFoundation/UA-.NETStandard/blob/master/CONTRIBUTING.md) doc. - [x] I have added tests that prove my fix is effective or that my feature works and increased code coverage. — *Types 8562, Server 4072, Client 2123, InformationModel 1063, Robotics 540, Tools 466, Vision 442, AI 74 and the restored Vision+Intent end-to-end suite 14/14, all green on net10.0. Only `.OpenUsd` (71.1%) is below the 80% bar — see Not finished.* - [x] I have added all necessary documentation. - [x] I have verified that my changes do not introduce (new) build or analyzer warnings. — *`dotnet build UA.slnx -c Release`: 0 errors, 0 warnings.* - [x] I ran **all** tests locally using the **UA.slnx** solution against at least .net **framework** and .net **10**, and all passed. — *net48: Types 8555, Client 2126, Vision 385 green. Two net48 Server tests fail (`CreateUserManagementSeamBindsTheModelToTheServerAsync`, and `DurableDataValueQueueVerifyReferenceBatchingAsync` only under parallel load); both were confirmed **pre-existing** by re-running them with the `IsInverse` change reverted, which fails identically.* - [ ] I fixed **all** failing and flaky tests in the CI pipelines and **all** CodeQL warnings. — *The net48 build legs failed because the restored end-to-end project used `ValueTask.FromResult` / `string.Create`, which do not exist on .NET Framework; fixed in `fb57522da` and verified building on every TFM. `DisposeDrainsHeldConnectionCallbackBeforeListenerDisposal` failed once on the Ubuntu Client leg — a 13 ms timing-sensitive reverse-connect disposal test that passes locally 2123/2123 and touches no browse paths.* - [ ] I have addressed **all** PR feedback received. ## Typed Vision-guided Robotics MCP workflows The branch now also makes the demo workflow first-class in the Full MCP profile: - replaces stringified Robotics intent and mission inputs with typed request schemas and exact name-or-NodeId selectors; - adds bounded operation/mission paging, stable step-to-operation identity, retained mission discovery, and `robotics_wait_mission`; - makes `vision_run_inference` return bounded typed detection, inspection, or segmentation summaries in one call; - adds same-session `robotics_vision_pick` to select a detection and submit Pick or Pick/Place work without implicitly taking authority, retrying, cancelling, or waiting; and - starts Publish workers for classic subscriptions so an independently launched OpenUSD viewer receives live robot and part updates. The migration guide documents the intentionally breaking MCP request-schema changes. --------- Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Copilot-Session: 1dab1302-19c5-4a9b-a50c-d97d389713aa Copilot-Session: a248a589-9a20-4372-868e-5d347e57001b Copilot-Session: 99d1ab42-f83a-41ab-bf3e-85bf81fadace Copilot-Session: bcbc8ac9-5abd-42a8-8bac-e10533790e2e	2026-08-29 19:34:15 +02:00
Marc Schier	466c5cb982	Robot intent follow-ups: multi-controller executors, spec fixes, and an MCP tool library (#4195) ... Follow-ups to #4165, plus the work that followed from re-checking this branch against the specification. Everything started as a defect found by review or by probing the running code; the specification is unchanged except where noted. The TLS certificate lifetime fix (#4196) and the OpenUSD work (#4198) were both merged into this branch. #4196 has since landed on master, so it no longer appears in this diff. #4198 was closed in favour of this PR and its content is here. The MCP trim/AOT work (#4228), the `DataGenerator` null `NodeId` fix (#4227) and the NodeSet external-parent import fix (#4229) have since been merged in as well. All three are closed in favour of this PR and their content is here, described under *Making the MCP libraries trim and AOT clean*, *A generator that returned the value you opted out of* and *A NodeSet parent that lived outside the import batch* below. ## A two-arm cell could not be built `IntentExecution` carried no controller identity, and there was effectively one `IIntentExecutor` per server. An executor handed an execution therefore could not tell **which arm** it was for, so a cell with two `IntentControllerType` instances — an ordinary configuration — was not expressible through the builder or DI. Both halves are addressed: - `IntentExecution` now carries `ControllerId` (`NodeId`, the OPC UA identity) and `ControllerName`, so one shared executor can discriminate; - DI gained per-controller registration — `AddRobotIntentExecutor<TExecutor>(controllerBrowseName)` and an instance overload — so two genuinely different executors can coexist. Resolution order is `WithExecutor` → per-controller DI registration → the single DI executor → a build-time throw. There is still **no silent fallback** to a rejecting executor; that was a Critical defect fixed in #4165 and it stays fixed. `IIntentExecutor` is unchanged, so this is purely additive. ## Declarations that were quietly untrue - **`Accepts<T>` silently discarded a redeclaration.** Declaring an intent type twice — say to narrow `cancelSupported` on the second call — kept the first, wider declaration and said nothing. That is the same class of defect as the capability-honesty problems #4165 fixed (`PauseSupported` declaring a pause the host could not perform; `Aborting` degrading silently to `Buffered`). Identical redeclarations remain idempotent; conflicting ones now throw `BadInvalidArgument`. - **`RI-Interop-40010` assumed Annex B item 3 rather than checking it.** It now requires exact set equality between the OPC 40010 `TaskProgramName` values and the Robot Intent `ProgramType.ProgramId` values. Both sides are readable from the address space without commanding the robot, which is what makes the requirement structural under §12.1. ## Two things the specification requires that this did not do Re-checking the implementation against the merged draft revision turned up two genuine violations. Both are fixed, and the NodeSet is untouched by either. - **Every graphed mission reported `Failed` on success.** `MissionRules.SelectTransition` returned `null` for two different situations — a graph *terminus*, which is the normal success path, and a branch point whose conditions all evaluated false — and the caller collapsed both into `Failed`/`NoTransition`. A three-way result (`Terminus` / `Selected` / `NoConditionMatched`) separates them, so a mission that runs to its end now completes. The failure reason surfaces through the inherited Part 10 result container rather than a new member. - **Facet URIs were never published.** An upstream `may` became a `shall` in the revision this branch adopts. `RobotIntentConformanceUris.Facets` held 31 URIs with **zero** production references — the set existed but nothing emitted it. The published set is now derived from `RobotIntentFacetCalculator`, the same authority that already computed `SupportedFacets`, rather than from a second hand-maintained list that could drift; a reflection test pins the two together. ## An easy client mistake `RequestAuthorityAsync` returns a lease the caller must remember to inspect. Forgetting makes every later submission fail with `ControlNotOwned` for reasons that are invisible at the call site. `RequireAuthorityAsync` throws instead, and names the current owner. `RequestAuthorityAsync` is untouched — inspecting and branching is legitimate. ## Letting an agent drive the robot New `tools/Opc.Ua.Mcp.Robotics` gives a language model the intent API as MCP tools: 36 across discovery, monitoring, direct control and missions, behind a `Robotics` tool profile, composable into any MCP host exactly as the existing OPC UA tool packages are. The `IntentViewerClient` sample hosts them behind `--mcp` (stdio by default, HTTP when `--view`, since MCP stdio and the in-process viewer would otherwise share stdout), and the sample gained a pallet scene so an agent can pick and stack parts and watch the result in the OpenUSD viewer. All the scene logic is server-side — no viewer or connector changes. Four gaps in the client API had to be filled first, since the tools are a thin layer over it and could not have been written honestly without them: `ReadStateAsync`, `ListOperationsAsync`/`ListMissionsAsync`, controller-level `CancelIntent`/`CancelAll`/`Pause`/`Resume`/`Retry`/`ReleaseControl`, and `WaitForCompletionAsync`. The tools refuse rather than paper over: a refusal from the server is returned with its `IntentFailureEnum` and never retried, and command authority is never acquired as a side effect of a tool call. **One defect this surfaced:** `--profile robotics` shipped 36 tools that could not do anything. Every robotics tool resolves a named OPC UA session and only the connection tools can open one, so with no `Connect` in the catalogue the whole profile failed. A test had pinned the defect in place, asserting the profile did *not* contain `Connect` on the line below the `diagnostics` assertion requiring it. Each package now declares its own requirement in its registration extension rather than the host special-casing it, so an application embedding `Opc.Ua.Mcp.Robotics` directly gets a usable catalogue without reproducing the host's pairing. `Full` still takes the connection tools from the core package, so they are never registered twice. The replacement test states the invariant — a profile exposing session-scoped tools must expose the tools that open a session — rather than the instance. ## Making the MCP libraries trim and AOT clean *Merged from #4228.* `Opc.Ua.Mcp.Core` is now `IsAotCompatible` on `net10.0`, and the reflection the analysers then reported is gone. Turning the analysers on is the point: until now they were not running on these libraries at all — including on the tool library this branch adds. - **`OpcUaJsonHelper.Serialize`** went through `JsonSerializer` (IL2026 / IL3050). The conversion helpers in that class already reduce every OPC UA value to a closed set — `null`, the primitive scalars, `string`, `Dictionary<string, object?>` and lists thereof — so the JSON is now written directly with a `Utf8JsonWriter`. Output is unchanged, and that is asserted rather than assumed: `OpcUaJsonHelperTests.SerializeMatchesTheReflectionBasedSerializer` compares the new writer byte for byte against `JsonSerializer` for every shape the helpers emit, including all numeric widths, the custom-formatted scalars, nested sequences and strings that force the encoder to escape. One deliberate contract change: arbitrary object graphs are no longer accepted. Serializing them needs reflection, so `Serialize` now throws `NotSupportedException` naming the type instead of silently stringifying it. No call site in the repository passes anything outside the closed set. - **`McpSchemaFilters`** used `JsonSerializer.SerializeToElement` on a `JsonObject`; it now parses the node's own text, which needs no reflection. `Opc.Ua.Mcp.Diagnostics` is deliberately **not** marked. It depends on SharpPcap and on the reflective service-call dissection in `Opc.Ua.Core.Diagnostics`, neither of which is trim safe — which is why that package does not opt in either. Two improvements are kept regardless: `PacketDecodeTools` serializes `DecodedServiceCall` through a source-generated `JsonSerializerContext`, and `PacketCaptureTools.ListInterfaces` now carries the `RequiresUnreferencedCode` / `RequiresDynamicCode` annotations that `NetworkInterfaceEnumerator.ListLocalInterfaces` already declares, so the constraint is visible to callers instead of implicit. Nothing is suppressed. Marking an assembly `IsAotCompatible` only proves its own IL is clean; it says nothing about whether the library still functions once reflection is gone. So `Opc.Ua.Aot.Tests` references `Opc.Ua.Mcp.Core` and exercises it with JSON reflection disabled, stating what is and is not true today: - tool **results** serialize without reflection — this would fail outright if the `Utf8JsonWriter` change regressed; - tool **registration** still requires reflection, because the MCP SDK builds each tool's input schema by asking `JsonSchemaExporter` for a `JsonTypeInfo` of every parameter type. Without a source-generated context covering the tool signatures — and with the injected `OpcUaSessionManager` being described rather than treated as a service — that throws `NotSupportedException`. The second test asserts the throw. It documents the remaining gap and turns red the moment someone closes it, rather than leaving it to be found in a published binary. The package README says the same thing in prose. **Two pre-existing issues this surfaced, both outside this PR:** pulling `Opc.Ua.Mcp.Diagnostics` into an AOT publish reports `IL2046` / `IL3051` on `NicCaptureSource.StartAsync`, which is annotated `RequiresUnreferencedCode` / `RequiresDynamicCode` while the `ICaptureSource.StartAsync` it implements is not, and `IL2075` on `ServiceCallReassembler.GetCollectionCount` / `GetPropertyString`, which call `Type.GetProperty` on the result of `GetType()`. Happy to file these separately if useful. ## A generator that returned the value you opted out of *Merged from #4227. Fixes #4226.* `GetRandomNodeId(useBoundaryValues: false)` could still return `NodeId.Null` — one of the boundary values the same method offers when boundary values *are* requested. A caller that opted out should not receive one, and `GetRandomExpandedNodeId` builds on it and inherited the same behaviour. The `IdType.Opaque` branch is the practical path: `GetRandomByteString()` draws its length with `NextInt32(MaxStringLength)`, which is inclusive of zero, and a zero-length opaque identifier in namespace 0 satisfies `NodeId.IsNull`. With three namespaces and `MaxStringLength` at 10 that is roughly **one call in 130**. A numeric identifier of `0` in namespace 0 is null too, at around one in four billion. The draw is now repeated while the result is null and boundary values were not requested. Guarding the result rather than the individual branches covers every identifier type, including the numeric case, rather than only the one that happens to be reachable today. The `useBoundaryValues: true` path is untouched, so boundary values remain available when they are asked for. Four tests in `tests/Opc.Ua.Core.Tests/Types/DataGeneratorTests.cs` loop 20,000 draws, because at one call in 130 a single draw proves nothing. Against the **unmodified** generator they fail on draw 93 and draw 1346. Two of the four exist to keep the fix honest rather than to test the bug: `GetRandomNodeIdStillProducesEveryIdentifierType` catches a "fix" that simply stopped emitting opaque identifiers, which would satisfy the null assertions while silently narrowing the generator, and `GetRandomNodeIdWithBoundaryValuesStillYieldsThem` confirms the leak was closed rather than the feature removed. **How it was found:** an intermittent downstream test failure. A test server initialising a static `NodeId`-valued variable via `TestDataSystem.ReadValue` occasionally stored `NodeId.Null`; the encoder correctly writes that as JSON `null`, and a consumer asserting a value was present failed about one run in five — never reproducibly in isolation. It only surfaced under load because `UnsecureRandom.Shared` is a fixed-seed, process-global generator shared between `DataGenerator` and OPC UA transport jitter, so the number of draws preceding address-space initialisation depends on what else is running in the process. ## A NodeSet parent that lived outside the import batch *Merged from #4229.* `UANodeSetHelpers.LinkParentChildRelationships` builds its lookup table from **only the nodes in the current import batch**, links a child when the parent is found — and then cleared `instance.Handle` **unconditionally**, which is exactly where `Import` had stashed the authored `ParentNodeId`. A node whose parent lives in another NodeSet, or is owned by another NodeManager, therefore lost that parent **silently** at import, and nothing downstream could recover it. `RuntimeNodeSetNodeManager.CreateAddressSpaceAsync` already has a step that wires references to external node managers; the parent link never reached it. Such a parent is now recorded rather than discarded, and exposed through `UANodeSet.TryGetUnresolvedParentNodeId`. A parent that resolved normally inside the batch is **not** recorded, so a caller does not re-wire references that already exist. The record lives in a `ConditionalWeakTable`, so it does not keep an imported node alive and does not widen `NodeState` — whose `Handle` is a general-purpose slot a caller may legitimately use for its own purposes, and so is not a safe place to leave the value. `RuntimeNodeSetNodeManager` consumes it to report the one case that cannot be repaired automatically. In NodeSet2 the `ParentNodeId` attribute is a *hint*; the authoritative edge is the entry under `<References>`, which the reverse-reference pass already turns into an external reference. A node that names an out-of-manager parent but carries **no such Reference** materializes with no path from that parent, and no ReferenceType can be inferred to synthesise one — so it is reported as an authoring defect rather than guessed at or silently accepted. This is deliberately **not** a fix for a specific address-space symptom. It removes a silent loss of authored information and gives callers the means to act on it. Whether any particular orphaned-node report is caused by this path, or by namespace-index translation elsewhere, still needs separate diagnosis. **Changed while merging.** As authored in #4229 the accessor was `GetUnresolvedParentNodeId`, returning `NodeId?`. `NodeId` is a readonly struct implementing `INullable`, so that was a `System.Nullable<NodeId>` wrapped around a type that already carries its own null representation — which this repository excludes in favour of an `IsNull` check, and it asks for `TryGet` over casting on struct types. The tests correspondingly asserted `Is.Null` against a struct, which the testing rules also exclude. It is now `bool TryGetUnresolvedParentNodeId(NodeState, out NodeId)`, yielding `NodeId.Null` when there is no unresolved parent. The API is unshipped — it arrives with this branch — so nothing depended on the previous shape. Both tests were re-verified by mutation after the rewrite: disabling the record in `LinkParentChildRelationships` still fails `ImportKeepsAParentThatIsNotInTheSameBatch`. ## Dependencies - **OpenUSD to `0.7.0-alpha`.** All 11 issues we filed upstream are fixed, so the workarounds they justified are gone and the parts of the pump and generator samples that were blocked on them are implemented. - **ModelContextProtocol to `2.1.0`** across all five MCP tool libraries and the executable. No code changes were needed — the surface we use is unchanged in 2.x — but central package management required a cascade of pins, each commented where it sits. Both transports were verified against a running server rather than assumed. MCP 2.x is stateless (no `initialize`, no `Mcp-Session-Id`); an OPC UA session is application state addressed by a caller-supplied name, which fits that model but does assume process affinity, and that assumption is now written down rather than implied. ## Infrastructure The `code coverage` gate was measuring nothing on this branch and reporting success. It is fixed and now enforced, which immediately exposed real gaps — most usefully `RoboticsIntentJson`, which parses agent-supplied JSON (untrusted input from a language model) and had no tests at all. It is now at full line coverage in CI. Patch coverage is **90.20%** against a 75% floor. The `netstandard2.1` build leg was also failing. The MCP projects build as empty shells under a legacy `CustomTestTarget`, and the viewer sample referenced them on the target framework alone, so it tried to resolve a net10.0 shell from net8.0. The sample now derives MCP support from whether those projects are actually being built. ## From review A review pass over the branch found two real defects, both fixed. `RoboticsIntentJson` parses JSON written by a language model, so every malformed shape it can be handed is reachable input; it indexed pose arrays without checking their length and enumerated properties without checking their kind, so a short or wrongly-typed array escaped as `IndexOutOfRangeException` or `InvalidOperationException` rather than the argument error the tool descriptions promise the agent. Separately, `IntentOperationHandle` treated a zero `QueuePosition` as "not observed" and discarded it, so a handle kept reporting the last queued position after the server had said the operation left the queue. The pose validation immediately caught a latent test defect: the pallet scenario interpolated doubles into JSON with the current culture, so under a comma-decimal culture `[0.12, 0.16, 0.28]` became a six-element array. It passed only because that submission is expected to be refused for want of authority, so the wrong coordinates were never used. A third review finding — that a default-constructed `RobotIntentControllerState` hands out a null `NodeId` — was checked and rejected: `NodeId`, `QualifiedName` and `LocalizedText` are readonly structs here, so `default` is the null value and `IsNull` is `true`. The tests asserting that invariant are kept. A second review pass over the merged branch found two more, both fixed and both about an error surfacing as the wrong kind of failure rather than about a wrong result: - **Malformed agent JSON did not arrive as an argument error.** `RoboticsIntentJson` turns every *validation* failure into an `ArgumentException` — a wrong root kind, a missing property, a short pose array — but three `JsonDocument.Parse` calls and the NodeId parse were unguarded. A syntax error escaped as `JsonException` and an unparseable NodeId as `ServiceResultException`, which are the two shapes a language model is most likely to produce. That is the same defect class as the pose/kind checking above: a caller separating bad input from a genuine fault had to know to catch three unrelated exception types. Parsing now goes through one helper that reports a syntax error as an `ArgumentException` naming the argument, keeping the `JsonException` as the inner exception so the position is not lost, and the NodeId goes through `NodeId.TryParse` rather than `Parse`, collapsing its two failure modes into one. `MalformedIntentJsonIsRejected` had asserted `InstanceOf<JsonException>`, pinning the behaviour rather than the intended contract; it now asserts `ArgumentException` like every neighbouring test, and three tests cover the newly guarded paths. - **`ConditionalWeakTable.Add` throws on a duplicate key.** Not reachable today, but `TryGetUnresolvedParentNodeId` documents `Handle` as a general-purpose slot a caller may use for its own purposes — and a caller doing exactly that before re-importing into the same collection would take a hard throw from inside the import pipeline. Removing first makes the most recent import win. `AddOrUpdate` would say this more directly but does not exist on `netstandard2.0`. ## Deliberately not fixed here `FastenIntentDataType.Joint` is refused when the Server exposes no OPC 40450/40451 model, and a client cannot discover that before calling. That is **not fixable in this repository**: §5.4.2 makes `Joint` the defining mechanism while §11.3 says the parameters otherwise stand alone, and `IntentCapabilityDataType` declares capabilities at intent-type granularity with no member-level mechanism anywhere in the model. It is a specification gap, raised as S4 on [marcschier/opcua-drafts#60](https://github.com/marcschier/opcua-drafts/pull/60) rather than worked around with a bespoke mechanism the spec does not define. ## Validation `dotnet build UA.slnx -c Release` on the merged branch — **0 errors, 0 warnings**, with the AOT analysers now running over `Opc.Ua.Mcp.Core`. | Suite (net10.0) | Tests | |---|---| | `Opc.Ua.Core.Tests` | 4247 (86 skipped) | | `Opc.Ua.Robotics.Tests` | 540 | | `Opc.Ua.OpenUsd.Tests` | 909 | | `Opc.Ua.Tools.Tests` | 419 | | `Opc.Ua.Server.Tests` — `RuntimeNodeSet` | 94 | | `Opc.Ua.Robotics.Intent.Tests` | 22 | | `Opc.Ua.Aot.Tests` — `McpAotTests` | 3 | 0 failures. `Opc.Ua.Tools.Tests` rose from 409 to 419 with the serializer-equivalence tests from #4228 and the malformed-JSON tests from the second review pass, and `Opc.Ua.Core.Tests` covers the `DataGenerator` tests from #4227 and the import tests from #4229. `UANodeSetHelpers` also passes on **net48** (5/5). The MCP agent-usability grader scores Core 100 / Full 100. --------- Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com> Copilot-Session: a248a589-9a20-4372-868e-5d347e57001b Copilot-Session: 9e6a5abf-3299-4cd1-9855-010fedbf0ad8 Copilot-Session: a0718e8a-a158-45c5-929d-e1d6c0982431	2026-08-11 13:03:36 +02:00