H3 Product Shell Boundary:
The H3 Product Shell Boundary connects the H3 context backend to its supported ComfyUI sidebar environment, presenting a backend-validated view of your current work while keeping execution authority under backend control. Its purpose is to make the prepared prompt and its readiness information available in a manual-only product workflow, rather than automatically authoring changes or starting generation. It builds a bounded, content-free product-shell representation containing information such as task mode, profile, readiness, supported field bindings, and limitations, and publishes the associated sidebar workspace. Here, “content-free” describes the shell metadata, not the separately available prompt text. When additional workflow information is supplied, the node can also publish transaction transparency, generation-sequence information, and a controlled semantic-proposal review handle. Each execution is tied to the active ComfyUI prompt and node, helping prevent stale interface information from being reused. Importantly, having a prompt ready for export does not mean the native generation setup is qualified to queue, and assisted authoring remains unselected, unauthorized, and not ready in this projection.
H3 Product Shell Boundary Input Parameters:
report
Required connection of type H3_CONTEXT_REPORT. This provides the prepared context report used to construct the product-shell projection and publish the sidebar workspace. Its prompt document is also the source of the returned prompt text, so this node exposes the text already prepared upstream rather than rewriting it. The projection records identifying information about this report, including its revision and fingerprint, allowing the shell to describe the particular report being presented. The value must be an exact backend ContextReport object; plain text, a dictionary, or an unrelated connection is not accepted. There is no numeric minimum, maximum, default value, or selectable option defined for this input.
native_h3_wiring
Required connection of type H3_NATIVE_H3_WIRING. This supplies the native H3 configuration used alongside the report to build the product shell and sidebar workspace. The resulting projection includes information derived from this wiring, such as the task mode, profile, native node identifier, and prompt fingerprint. Its qualification status also determines whether native queuing is considered ready: an exportable prompt does not override a hold caused by unqualified native inputs. This same wiring is used when an optional semantic-proposal review is staged. The value must be an exact backend NativeH3Wiring object, not a manually assembled substitute. There is no numeric minimum, maximum, default value, or selectable option defined for this input.
recompute_plan
Optional connection of type H3_RECOMPUTE_PLAN. When supplied together with pipeline_transaction, this provides the recomputation plan used to build a transaction-transparency projection for the interface. This lets the node present the supplied plan and transaction in the context of the current execution; it does not itself carry out recomputation. Although the socket is individually optional, supplying it requires a valid pipeline_transaction as well. Connecting only one member of this pair causes execution to fail rather than publish incomplete transaction information. The supplied value must be an exact RecomputePlan object. The execution default is None, meaning no plan is supplied; numeric limits and selectable options do not apply.
pipeline_transaction
Optional connection of type H3_PIPELINE_TRANSACTION. This supplies the transaction object that accompanies recompute_plan when transaction-transparency information is published. It adds transaction-related interface information without changing the fact that the returned prompt comes directly from the report. If either transaction-related input is connected, both must be present with the expected backend types; this input must specifically contain an exact PipelineTransaction object. For a basic prompt-and-sidebar workflow, you can leave both transaction-related inputs disconnected. The execution default is None, and no numeric minimum, maximum, or selectable options are defined.
generation_sequence_state
Optional connection of type H3_GENERATION_SEQUENCE_STATE. This supplies generation-sequence state for an additional interface projection associated with the current execution. When present, the node validates that it is an exact GenerationSequenceState object, builds its sequence projection, and associates the backend sequence authority with the published sidebar workspace. This keeps the underlying authority on the backend rather than copying workspace authority or commands into browser-facing information. Supplying this input publishes sequence-related state; it does not, by itself, mean the node starts a generation sequence. The execution default is None, which omits this additional publication. There are no numeric limits or selectable options.
semantic_proposal_review_authority
Optional connection of type H3_SEMANTIC_PROPOSAL_REVIEW_AUTHORITY. This provides backend authority to publish a semantic-proposal review associated with the current wiring, execution, and sidebar workspace. It must be an exact SemanticProposalReviewAuthority object, not proposal text or a manually constructed review payload. The node claims this authority, stages and commits the review publication, and then commits the claim; the interface receives a review handle rather than the private review content or backend commands. The code treats the predecessor authority as one-time: publication is arranged so that a publication failure remains retryable, and a staged publication is rolled back if a later step in that operation fails. The execution default is None, meaning no review publication is requested. Numeric limits and selectable options do not apply.
H3 Product Shell Boundary Output Parameters:
prompt
Output of type STRING containing the exact text stored in the report’s prompt document. This is the ordinary text form of your prepared prompt, suitable for connecting to a compatible downstream text input. The node does not add shell metadata to this string or rewrite it according to the optional transaction, sequence, or review information. Treat it as prompt content, not as proof that generation can safely be queued: the product shell tracks native-input readiness separately, and a prompt can be export-ready while the native setup remains unqualified.
product_shell
Output of type H3_PRODUCT_SHELL containing the backend ProductShellProjection object. This is a structured, bounded representation of the product state rather than an image or another copy of the prompt. It includes report identity and revision information, report and prompt fingerprints, execution correlation, task mode, profile, host information, native node identity, field identifiers, bindings, limitations, and readiness indicators. In the provided builder, prompt export is marked ready, native queue readiness follows native-input qualification, and assisted readiness is false. The readiness reason is manual_only_scoped when native inputs are ready and native_input_unqualified otherwise. Assisted-authoring availability depends on whether the prompt-model catalog contains profiles, but availability does not mean it is selected, ready, or authorized. The node separately publishes sidebar and optional transaction, sequence, and review information to ComfyUI’s interface; those publications are not additional output sockets.
H3 Product Shell Boundary Usage Tips:
- For a basic setup, connect valid
reportandnative_h3_wiringvalues and leave the optional inputs disconnected. This gives you the prepared prompt and product-shell projection while also publishing the sidebar workspace. - Find the node under
h3_context/productwith the display name “H3 Product Shell Boundary.” It is an output node, so its interface-publication role matters even when you do not connect its results farther downstream. - Use the
promptoutput when you need ordinary text. Useproduct_shellonly with components that understandH3_PRODUCT_SHELL; it is structured backend state, not a text prompt. - Connect
recompute_planandpipeline_transactiontogether when you need transaction transparency. If you do not need that interface information, leave both disconnected rather than connecting only one. - Check native-input readiness before attempting to queue generation through the surrounding workflow. A
native_input_unqualifiedreason points to a wiring qualification issue; editing or exporting the prompt alone does not remove that hold. - Add
generation_sequence_stateonly when you have a valid backend-generated sequence state to present. This input publishes sequence information and its workspace association rather than acting as a start-generation switch. - Treat semantic-proposal review authority as action-specific and one-time. For a new review action, obtain the appropriate authority from the upstream producer instead of assuming a previously committed authority can be reused.
- Run the node within an active ComfyUI execution so it can obtain the host prompt and execution-node identifiers automatically. These are execution details, not settings you need to add as visible inputs.
- Expect the node to execute again for each prompt: its change-checking behavior intentionally prevents reuse of cached interface information. There are no sampling, resolution, or speed controls on this node; configure generation performance in the relevant upstream or native generation components.
H3 Product Shell Boundary Common Errors and Solutions:
invalid_authority — product shell requires exact report and wiring values
- Explanation: One or both required inputs are not the exact backend object types expected by the node. Matching-looking data, plain strings, dictionaries, and even substitute object types are not accepted.
- Solution: Reconnect
reportto a component that produces a genuineH3_CONTEXT_REPORTandnative_h3_wiringto a component that produces genuineH3_NATIVE_H3_WIRING. If the problem started after an extension update, refresh the affected upstream results and check that the installed H3 components are compatible.
missing_correlation — product shell requires host prompt and execution-node IDs
- Explanation: The node could not obtain valid string identifiers for the active host prompt and execution node. These identifiers are necessary to associate the shell and sidebar information with the correct execution.
- Solution: Run the workflow through ComfyUI rather than calling the node directly outside its execution environment. If it already runs in ComfyUI, check that the H3 execution integration is available and functioning; manually adding visible prompt-ID inputs is not part of this node’s interface.
correlation_mismatch — hidden execution-node authority does not match the active host context
- Explanation: The automatically supplied execution-node identifier disagrees with the identifier reported by the active host execution. The node refuses to publish information under a conflicting execution identity.
- Solution: Start a fresh normal execution and avoid integrations that override or reuse execution identifiers. If the mismatch continues, check the host adapter or custom execution integration rather than changing your prompt text.
projection_failed — product-shell projection failed closed
- Explanation: A projection, workspace publication, optional publication, type check, or other protected operation failed. “Failed closed” means the node stops instead of returning a successful result with potentially invalid authority. Most errors inside the protected publication process are deliberately wrapped in this general message, and unexpected failures do not expose private review content or identifiers.
- Solution: First verify the two required upstream objects. Then test with the optional inputs disconnected and reconnect each optional feature, keeping the transaction pair together, to identify the failing path. Check available backend diagnostics for the underlying cause without sharing private review data. If a specific product-shell validation code accompanies the same message, use that code to guide further investigation.
incomplete_transaction_authority
- Explanation: This is the underlying transaction error raised when
recompute_planorpipeline_transactionis supplied without a valid matching input. At the node boundary, it is wrapped asprojection_failedwith the message “product-shell projection failed closed.” - Solution: Supply an exact
RecomputePlanand an exactPipelineTransactiontogether, or disconnect both inputs if transaction transparency is unnecessary. Do not replace the missing object with a string or dictionary.
generation_sequence_state_type
- Explanation: This underlying error means the connected generation-sequence value is not an exact
GenerationSequenceStateobject. The node normally reports it through the generalprojection_failederror. - Solution: Reconnect
generation_sequence_stateto the appropriate backend sequence-state producer, or disconnect it if you do not need sequence publication. Regenerate stale upstream state rather than attempting to construct the value manually.
review_authority
- Explanation: This underlying error is raised when
semantic_proposal_review_authorityis not an exactSemanticProposalReviewAuthorityobject. It is wrapped asprojection_failedat this node’s boundary. Failures while claiming, staging, or committing a valid authority are also handled through the protected review-publication path. - Solution: Supply the authority object produced for the intended review action, not proposal text or a browser-facing review handle. If an earlier review completed successfully, obtain a fresh authority for a new action. If review publication is not needed, disconnect this optional input.
