Pause-time token emission now draws only from the recipient set each API
surface is allowed to act on (emit ⊆ validate), so the CLI/OpenAPI caller is
never handed a token the resume endpoint would reject as 404 (WTA-867).
A form's recipients are partitioned once, per surface, into a single
FormDisposition: the surface-actionable recipient yields `form_token`, while
the rest are reported as `approval_channels` (e.g. ["email", "console"]) so the
caller is told where approval actually happens. Token and channels are two
projections of one decision (disposition_for_surface) loaded by one recipient
query (load_form_dispositions_by_form_id); the live pause path and the
reconnect snapshot path consume the same FormDisposition so they cannot drift.
RecipientType carries its user-facing approval-channel label as an enum tuple
value, set in __new__, so a new recipient type cannot be declared without one.
Tests: consolidate recipient/disposition/enrich tests into parametrized
matrices, add CONSOLE-surface and empty-token coverage, extract a shared fake
session for the pause-event tests.
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Co-authored-by: autofix-ci[bot] <114827586+autofix-ci[bot]@users.noreply.github.com>
The frontend and backend implementation for the human input node.
Co-authored-by: twwu <twwu@dify.ai>
Co-authored-by: JzoNg <jzongcode@gmail.com>
Co-authored-by: yyh <92089059+lyzno1@users.noreply.github.com>
Co-authored-by: zhsama <torvalds@linux.do>
The backend part of the human in the loop (HITL) feature and relevant architecture / workflow engine changes.
Signed-off-by: yihong0618 <zouzou0208@gmail.com>
Co-authored-by: autofix-ci[bot] <114827586+autofix-ci[bot]@users.noreply.github.com>
Co-authored-by: -LAN- <laipz8200@outlook.com>
Co-authored-by: 盐粒 Yanli <yanli@dify.ai>
Co-authored-by: CrabSAMA <40541269+CrabSAMA@users.noreply.github.com>
Co-authored-by: Stephen Zhou <38493346+hyoban@users.noreply.github.com>
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
Co-authored-by: yihong <zouzou0208@gmail.com>
Co-authored-by: Joel <iamjoel007@gmail.com>
Certain metadata (including but not limited to `InvokeFrom`, `call_depth`, and `streaming`) is required when resuming a paused workflow. However, these fields are not part of `GraphRuntimeState` and were not saved in the previous
implementation of `PauseStatePersistenceLayer`.
This commit addresses this limitation by introducing a `WorkflowResumptionContext` model that wraps both the `*GenerateEntity` and `GraphRuntimeState`. This approach provides:
- A structured container for all necessary resumption data
- Better separation of concerns between execution state and persistence
- Enhanced extensibility for future metadata additions
- Clearer naming that distinguishes from `GraphRuntimeState`
The `WorkflowResumptionContext` model makes extending the pause state easier while maintaining backward compatibility and proper version management for the entire execution state ecosystem.
Co-authored-by: autofix-ci[bot] <114827586+autofix-ci[bot]@users.noreply.github.com>
This pull request introduces a feature aimed at improving the debugging experience during workflow editing. With the addition of variable persistence, the system will automatically retain the output variables from previously executed nodes. These persisted variables can then be reused when debugging subsequent nodes, eliminating the need for repetitive manual input.
By streamlining this aspect of the workflow, the feature minimizes user errors and significantly reduces debugging effort, offering a smoother and more efficient experience.
Key highlights of this change:
- Automatic persistence of output variables for executed nodes.
- Reuse of persisted variables to simplify input steps for nodes requiring them (e.g., `code`, `template`, `variable_assigner`).
- Enhanced debugging experience with reduced friction.
Closes#19735.
