feat: add workflow graph and run detail view

[ralphschuler]

Parent

PRD.md

## Summary

Render workflow graph/run details with step states.

## What to build

- Add graph data API.

• Add graph UI for pending/running/waiting/failed/complete states.

• Add failed-step drilldown.

  Slice Type

  AFK

  Blocked by

  • feat: implement fan out fan in dependency scheduler #34 (feat: implement fan out fan in dependency scheduler #34)

• feat: add wait signal timer approval restart and idempotency tests #40 (feat: add wait signal timer approval restart and idempotency tests #40)

• feat: add stream reconnect disconnect tests #63 (feat: add stream reconnect disconnect tests #63)

  Evidence/Context

  PRD.md Phase 14 and implementation/testing decisions describe this behavior as part of the durable distributed execution platform roadmap.

  Decisions

  Follow already-approved PRD decisions: Postgres is authoritative, Kafka/Redpanda is derived, all durable resources are tenant/project scoped, public APIs are versioned REST/JSON/SSE contracts, and implementations must be behavior-tested.

  Tasks

  • Add graph data API.

• Add graph UI for pending/running/waiting/failed/complete states.

• Add failed-step drilldown.

• Add behavior-focused tests through public contracts or service boundaries.

• Update relevant docs or command references.

  Proposed Solution

  Implement this as a narrow vertical slice that proves one observable behavior through the stable contract/API/service path. Keep module interfaces small and hide internal complexity behind tests.

  Acceptance Criteria

  • Operator can identify failed/waiting step from graph.

• Fan-out/fan-in graph renders accurately.

• Access is authorized.

  Relevant Files/Commands

  • PRD.md

• packages/contracts/ where public contracts change

• apps/control-plane/ for web/API/control-plane work

• services/broker/ or services/scheduler/ for runtime loops when applicable

• workspace test commands listed below

  Validation

  • yarn workspace @helix/control-plane test -- workflow-graph

• yarn validate

  Risks/Rollback

  Risk: graph view can mislead operators if states are stale; source from persisted state/events.

  Source Conversation Context

  User requested issues for everything in PRD.md after the first 12 foundation issues were created. This issue is part of the remaining full-roadmap breakdown confirmed via human-in-loop.