H3 Semantic Proposal Producer:
The Semantic Proposal Producer prepares semantic proposals for review within the H3 Context workflow, using a locally configured Ollama model to work from validated H3 context information. Its purpose is to support a controlled proposal-generation stage rather than directly generate artwork or automatically apply changes. Before generation, it checks the source information and provider configuration, then runs the model through an execution path that tracks resource usage and supports time limits and cancellation. Successful work is registered for review within the current running process, while failed work has its pending reservation closed. This design separates proposal generation from review and keeps source material, model prompts, and responses out of the returned handle and unexpected error chains. For you, this provides a structured way to introduce model-assisted semantic suggestions into an H3 workflow without treating an unverified model response as an approved result.
H3 Semantic Proposal Producer Input Parameters:
report
Required input of type H3_CONTEXT_REPORT. This supplies the H3 context report used, together with the native wiring, to build the source information for semantic proposal generation. The node validates the source authority before resolving a model under the ollama.local profile, and source-bundle construction also forms part of the preparation for generation. Connect a compatible report from your H3 workflow rather than ordinary text or an unrelated custom object. The report affects the information available to the proposal-generation process; however, the provided code does not expose its internal fields or define how individual report contents change proposals. No default, numeric minimum, numeric maximum, or selectable options are specified.
wiring
Required input of type H3_NATIVE_H3_WIRING. This provides the native H3 wiring information that accompanies the report when the node validates its source and builds the semantic source bundle. In practical terms, it supplies the workflow-side wiring context needed to qualify the material used for proposal generation; it is not a text prompt or a model setting. Use compatible wiring associated with the report you intend to process, rather than assuming that any object with a similar name is interchangeable. The supplied code does not show the internal wiring fields or all validation rules. No default, numeric minimum, numeric maximum, or selectable options are specified.
provider_setup
Required input of type H3_PROVIDER_SETUP. This supplies the provider configuration used to prepare the proposal-generation authority and execute the local Ollama model. For the ollama.local profile, it also participates in resolving and qualifying your selected model. This connection therefore matters both before generation and during the actual model request: choosing a profile and typing a model name alone does not replace a valid provider setup. Connect a compatible H3 provider-setup object appropriate to your local configuration. Its internal settings are not shown in the supplied source, so this node does not expose documented endpoint, credential, or tuning fields through this socket. No default, numeric minimum, numeric maximum, or selectable options are specified.
ollama_profile
Required selection input that determines which local semantic-provider configuration the node uses. The authoritative interface lists two options: ollama.local and ollama.qwen3_8.27b_bf16.local. The source builds this list from the semantic connection and provider catalog, with the first connection profile selected by default; in the supplied interface, that default is ollama.local. Choose ollama.local when you want to specify a local model explicitly through ollama_model; this branch requires a nonempty model name and resolves it against the configured connection and provider setup. Choosing the catalog profile instead uses its predefined execution profile and model ID, and any nonempty model entry must match that model ID exactly. Profile qualification checks include the expected local endpoint and catalog configuration, so this selector is not a way to choose an arbitrary remote service. Numeric minimum and maximum values do not apply.
ollama_model
Optional STRING input with a default of "", an empty string. This chooses the model explicitly when ollama_profile is ollama.local; despite being optional in the interface, it must contain a nonempty string for that profile or execution raises choose_local_model. Enter the exact model identifier available through your configured local Ollama connection, for example by copying the identifier shown in your local model list. When using a catalog profile such as ollama.qwen3_8.27b_bf16.local, leave this input empty to use that profile’s configured model, or supply its exact model ID; a different value raises model_choice_invalid. The optional field preserves compatibility with saved workflows created before explicit model selection was added. No predefined model list, string-length limits, or numeric minimum and maximum values are specified.
H3 Semantic Proposal Producer Output Parameters:
review_authority
Output of type H3_SEMANTIC_PROPOSAL_REVIEW_AUTHORITY. This is a content-free handle to a semantic proposal review seed held inside the current running process. Think of it as a controlled reference that a compatible downstream review stage can use, not as the proposal text itself. The node returns this authority after the required preparation and successful proposal-generation transaction, or when its authority-management step supplies an already qualified review authority. Its presence indicates that the node has supplied authority for the review stage; it does not mean the proposed changes have been reviewed, approved, or applied. Because the underlying review state is process-local, do not treat the handle as an exported document or a portable result that can be reused independently after restarting ComfyUI.
H3 Semantic Proposal Producer Usage Tips:
- Start with a compatible
report,wiring, andprovider_setupfrom your H3 workflow. These are structured connections, not fields where you can substitute a written description of your scene. - For explicit local model selection, choose
ollama_profile=ollama.localand fill inollama_modelbefore running. An empty model field is not accepted for this profile. - For the catalog profile
ollama.qwen3_8.27b_bf16.local, leaveollama_modelempty unless you know the profile’s exact configured model ID. Do not assume the profile label itself is the model identifier. - Make sure your configured local Ollama service is available and that the selected model is accessible before queuing the node. The adapter checks the configured local endpoint and model qualification; selecting a profile alone does not establish that generation can succeed.
- Choose a model your machine can run within the configured action deadline. The explicit local-model path starts that deadline before model-profile resolution, so preparation and generation share the available time; this node exposes no separate timeout control.
- Expect execution to be reevaluated when the node is queued. Its change-check deliberately disables normal cache replay because provider qualification and process-local authority should not be reused as ordinary cached outputs.
- Pass
review_authorityto a compatible semantic review stage in the same running process. Keep generation and review conceptually separate: producing proposals is not equivalent to approving them. - Run the node through the normal ComfyUI execution flow. It depends on host-provided execution identity, so directly calling its generation method without the expected execution context can fail even when its visible inputs look correct.
H3 Semantic Proposal Producer Common Errors and Solutions:
choose_local_model
- Explanation: The selected profile is
ollama.local, butollama_modelis empty or is not a string. This profile requires an explicit model choice. - Solution: Enter the exact identifier of an available local Ollama model in
ollama_model, or switch to the catalog profile and leave the model field empty to use its configured model.
model_choice_invalid
- Explanation: A catalog profile is selected, but the supplied
ollama_modelis neither empty nor an exact match for that profile’s configured model ID. - Solution: Clear
ollama_modelto use the catalog model automatically. If you want to choose a different model, selectollama.localand enter that model’s exact identifier instead.
missing_correlation
- Explanation: The node could not obtain the required execution identifiers as strings. These identifiers associate its review authority with the current ComfyUI execution and node; they are supplied by the host, not through the visible parameter list.
- Solution: Queue the workflow normally in ComfyUI rather than invoking the node method directly. If normal execution still fails, check that the H3 execution integration is loaded correctly and that its installation is compatible with your ComfyUI environment.
correlation_mismatch
- Explanation: The node identifier supplied to the call does not match the identifier reported by the active execution context. The node stops rather than attaching authority to an inconsistent execution.
- Solution: Run the node through the standard workflow executor without manually overriding execution identifiers. If you use a custom runner or integration, check that it passes the identity of the node actually being executed.
cancellation_invalid
- Explanation: A host or test integration supplied an unsupported cancellation object. The node only accepts its expected local cancellation-probe type. This is not a setting you can correct through the visible model fields.
- Solution: Try execution without the custom cancellation integration. If you maintain the integration, supply a compatible
LocalCancellationProbeor omit the injected probe.
producer_execution_failed
- Explanation: An unexpected failure occurred during source preparation, provider work, generation, or transaction completion. The node deliberately replaces unexpected exception details with this general code so that source objects, prompts, responses, and exception chains are not exposed at the host boundary.
- Solution: Check that the report and wiring are compatible, confirm the provider setup is valid, verify that your local Ollama service and selected model are available, and retry through normal ComfyUI execution. If the problem persists, use safely available local diagnostics to investigate; this code alone does not identify the underlying cause.
Other provider or proposal-producer error codes
- Explanation: The node also forwards error codes raised by its semantic proposal producer and Ollama execution adapter. The supplied source does not list all those codes, so an exhaustive list of their exact messages cannot be established here.
- Solution: Use the specific displayed code to guide troubleshooting of the relevant stage: source compatibility, catalog-profile configuration, local model availability, provider qualification, cancellation, or execution timing. Consult the corresponding H3 provider or proposal-producer documentation for the precise meaning rather than treating every forwarded code as a model-name problem.
