feat(workflow): expose prompt and completion tokens in workflow run responses

[priyansh19]

Summary

Fixes #34315.

Workflow apps previously returned only total_tokens in their run results, while
chatflow apps already expose prompt_tokens (input) and completion_tokens (output)
through the usage metadata. Because most model providers price input and output
tokens differently, workflow consumers had no way to break down or estimate cost from
the run result.

The graph runtime already aggregates the input/output split across all LLM-backed nodes
in graph_runtime_state.llm_usage (maintained alongside the scalar total_tokens).
This PR surfaces that existing data without changing how it is computed:

• WorkflowFinishStreamResponse.Data (the streaming workflow_finished event) now
  includes prompt_tokens and completion_tokens, populated from
  graph_runtime_state.llm_usage.
• WorkflowAppBlockingResponse.Data (the blocking workflow run response) carries the
  same two fields through from the finish event.
• The OpenAPI WorkflowRunData schema documents the new fields.

The new fields default to 0, so behavior is unchanged for runs without LLM usage and
the change is backward compatible for existing API consumers.

Note on persistence scope

This change exposes the split in the live run responses (streaming + blocking), which is
where the issue reports the gap. Persisting prompt_tokens / completion_tokens on the
workflow_runs table (so historical reads via the console/service log APIs also show the
split) would require carrying the fields on the WorkflowExecution domain entity, which
currently lives in the external graphon package and only holds total_tokens. That can
follow as a separate change once the upstream entity supports it; happy to do it in a
follow-up if maintainers prefer.

Checklist

• This change requires a documentation update, included: Dify Document
• I understand that this PR may be closed in case there was no previous discussion or issues. (This doesn't apply to typos!)
• I've added a test for each change that was introduced, and I tried as much as possible to make a single atomic change.
• I've updated the documentation accordingly.
• I ran make lint && make type-check (backend) and cd web && pnpm exec vp staged (frontend) to appease the lint gods

From Cursor

[github-actions]

Pyrefly Diff

base → PR

--- /tmp/pyrefly_base.txt	2026-06-16 14:18:18.592313562 +0000
+++ /tmp/pyrefly_pr.txt	2026-06-16 14:18:09.689232697 +0000
@@ -3023,6 +3023,10 @@
   --> tests/unit_tests/core/app/apps/common/test_workflow_response_converter_resumption.py:31:37
 ERROR Argument `SimpleNamespace` is not assignable to parameter `user` with type `Account | EndUser` in function `core.app.apps.common.workflow_response_converter.WorkflowResponseConverter.__init__` [bad-argument-type]
   --> tests/unit_tests/core/app/apps/common/test_workflow_response_converter_resumption.py:32:14
+ERROR Argument `SimpleNamespace` is not assignable to parameter `application_generate_entity` with type `AdvancedChatAppGenerateEntity | WorkflowAppGenerateEntity` in function `core.app.apps.common.workflow_response_converter.WorkflowResponseConverter.__init__` [bad-argument-type]
+  --> tests/unit_tests/core/app/apps/common/test_workflow_response_converter_token_usage.py:34:37
+ERROR Argument `SimpleNamespace` is not assignable to parameter `user` with type `Account | EndUser` in function `core.app.apps.common.workflow_response_converter.WorkflowResponseConverter.__init__` [bad-argument-type]
+  --> tests/unit_tests/core/app/apps/common/test_workflow_response_converter_token_usage.py:35:14
 ERROR Object of class `NoneType` has no attribute `data` [missing-attribute]
    --> tests/unit_tests/core/app/apps/common/test_workflow_response_converter_truncation.py:588:16
 ERROR Argument `dict[str, dict[str, str]]` is not assignable to parameter `override_config_dict` with type `AppModelConfigDict | None` in function `core.app.apps.completion.app_config_manager.CompletionAppConfigManager.get_app_config` [bad-argument-type]

[github-actions]

Pyrefly Type Coverage

Metric	Base	PR	Delta
Type coverage	48.59%	48.59%	-0.00%
Strict coverage	48.10%	48.10%	-0.00%
Typed symbols	27,994	27,994	0
Untyped symbols	29,922	29,925	+3
Modules	2892	2893	+1