(CODUX-105) Schedule node-flow runs

[numnx]

🧩 Task: Schedule node-flow runs

Automated task PR opened by Code UX for T04.

📋 Summary

Sprint: Sprint 173 · Repeatable node flows · Task: T04

Implement repeatable n8n-style node workflows in Code UX with durable flow definitions, node invocation execution, MCP automation, agent skill attachment, scheduler support, an editable dashboard canvas, and current documentation.

🤖 Model & Provider

Provider	Codex CLI
Model	gpt-5.5

⏱️ Timing

Started	2026-07-07 11:13:07 UTC
Finished	2026-07-07 11:53:37 UTC
Duration	40m 29s

📊 Token Usage

Metric	Value
Input tokens	171,120
Cached input tokens	5,686,272
Output tokens	25,982
Total tokens	5,883,374
Invocations	2
Tool calls	92
Active time	10m 30s

Included usage estimate (subscription/local login): $4.48

✅ QA Review

⏳ QA review runs after this PR is opened. Check the task card in the dashboard for the outcome once it lands.

📝 Full Task Prompt

## Objective
Make node workflows schedulable through the existing scheduler engine, including backend contracts, MCP scheduler normalization, due-run execution, and persisted target payloads.

## Scope
- Edit `src/contracts/scheduler-types.ts`
- Edit `src/repositories/scheduler-repository.ts`
- Edit `src/services/scheduler-service.ts`
- Edit `src/mcp/management/scheduler-actions.ts`
- Edit `src/contracts/internal-management-types.ts`
- Edit `src/contracts/mcp-tool-definitions.ts`
- Edit `src/server/scheduler-routes.ts` only if request/response typing or route validation needs updates
- Edit `src/server/code-ux-server.ts` only if `SchedulerServiceDeps` requires `NodeFlowRuntimeService`
- Add or update tests under `tests/backend/repositories/`, `tests/backend/services/`, and `tests/backend/mcp/`

## Implementation Requirements
1. Add `node_flow` to `ScheduleTargetType` and define `ScheduleNodeFlowTarget` with `flowId`, optional JSON `input`, and optional `flowVersion` or version pin if T01 introduced version history.
2. Extend scheduler target serialization/deserialization in `SchedulerRepository` so `target_json` can store and round-trip `nodeFlowTarget` without breaking existing sprint, quicksprint, chat, or memory remediation rows.
3. Update `SchedulerService.executeEntry` to call `NodeFlowRuntimeService.runFlow` for `node_flow` entries and include scheduler metadata in the run options.
4. Add project ownership validation so only node flows from the selected project can be scheduled or executed by due-run processing.
5. Extend `SchedulerActions` and `manage_scheduler` schemas to accept `schedule_node_flow`, generic `create` with `targetType: "node_flow"`, flattened `flowId`, nested `nodeFlowTarget`, and optional JSON input.
6. Preserve recurrence, pause/resume, run count, failure marking, and `run_due` semantics for all existing scheduler targets.
7. Add tests for create/update/list/run-due behavior for node-flow entries and regression tests proving existing sprint/quicksprint/chat/memory remediation scheduler targets still round-trip.

## Constraints
- Do not add a second scheduling table or scheduler loop.
- Do not execute a node flow before `next_run_at` is due.
- Do not mark a schedule entry succeeded if `NodeFlowRuntimeService.runFlow` throws before creating or completing the run.
- Keep target details inside `target_json` to preserve row compatibility.

## Verification
- Run focused scheduler repository, service, and MCP action tests.
- Run `pnpm run test:backend`.
- Run `pnpm run lint` and confirm scheduler target unions are exhaustive.

🌿 Branch Info

Base: feature/CODUX-173-repeatable-node-flows
Head: task/feature-codux-173-repeatable-node-flows-t04-codex-mrajvqua

---

🤖 Generated by Code UX