feat(steering): K-Ext-2 Attention-Registry, Gate/Intake-Proposals, generische UI
Deploy Development / deploy (push) Successful in 46s
Test Suite / pytest-backend (push) Successful in 4m13s
Test Suite / lint-backend (push) Successful in 3s
Test Suite / compose-smoke (push) Has been skipped
Test Suite / k6 /api/health Baseline (push) Successful in 19s
Test Suite / playwright-smoke (push) Successful in 12s

Methodenuebergreifende Provider auf Kernel v0.3; verbindliche Coding Rules fuer Agenten in ADP und .cursor/rules.

Co-authored-by: Cursor <cursoragent@cursor.com>
This commit is contained in:
LarsandCursor committed 2026-07-27 11:48:10 +02:00
1 parent 1bbc5dbfce
commit b747200f3a
29 files changed
+1364 -182

No files matched your search

@@ -0,0 +1,134 @@
# ADP — Steering Kernel Coding Rules v0.1
**Status:** PO-freigegeben — **verbindlich für alle Coding-Agenten**
**Stand:** 2026-07-27
**Bezug:** ADP Steering Kernel Spine v0.1, ADP Steering Kernel Extension Model v0.1, ADP AP2.4, `.cursor/rules/kairo-steering-kernel.mdc`
---
## 1. Geltungsbereich
Diese Regeln gelten für **jede** Änderung an:
- `backend/steering/**` (Kernel, Registries, Provider, Ranker, Contributors)
- `backend/data_layer/initiative_snapshot.py` (Steuerungs-Read-Models)
- Frontend-Steuerungs-UI (`SteeringProposalsPanel`, Operating Context, Snapshot-Konsum)
**Ziel:** Ein gemeinsamer Steuerungskern für alle Methoden — keine Agile-Sonderlogik in Pages, kein Methoden-Fork im Kernel.
---
## 2. Herzmaschine (Stop-the-line)
| Regel | Detail |
|-------|--------|
| **SK-01** | Initiative-Steuerung (Next Work, Attention, Read Models, Proposals) **nur** über `steering.kernel.evaluate_steering()`. |
| **SK-02** | Snapshot/API spiegeln Kernel-Output — **keine** parallele Berechnung in `initiative_snapshot.py` (Ausnahme: Legacy-Signals bis Migration). |
| **SK-03** | Keine Steuerungsheuristik in Routers, Services außerhalb `backend/steering/`, oder React-Pages. |
---
## 3. Extension Registry (Pflicht-Muster)
Neue Steuerungsfähigkeit = **Provider registrieren**, nicht inline codieren.
| Typ | Registry | Aktivierung |
|-----|----------|-------------|
| Read Model | `steering/read_models/registry.py` | `requires_elements`, optional `requires_any_element`, `requires_data_slices` |
| Proposal | `steering/proposals/registry.py` | wie oben + optional `excludes_elements` |
| Attention | `steering/attention/contributors/registry.py` | `requires_elements` / `requires_any_element` |
| Ranker (Proposal) | `register_sprint_commit_ranker()` o. Ä. | Operating Context `proposal_rankers.{key}` |
**SK-04:** Provider-Schlüssel sind stabil (`sprint_commit`, `gate_next_actions`, `intake_triage`) — UI bindet an Keys, nicht an Archetyp.
**SK-05:** Keine feste Produktpolitik in Rankern (z. B. „Bugs immer zuerst“). Heuristik = Fallback mit `ranker_key: heuristic_v0`; Agent = `agent_v1` nach Principle Gate.
---
## 4. Proposal-DTO (Minimalvertrag)
Jeder Proposal-Eintrag liefert mindestens:
```python
{
"proposal_key": str,
"scope_type": "backlog_item" | "roadmap_item" | "action" | ...,
"scope_id": str,
"rank": int,
"reason_code": str,
"summary": str,
"ranker_key": str,
"confidence": "heuristic" | "agent",
"factors": [{"code", "weight", "label"}],
"dependency_refs": [...],
"dependency_blocked": bool,
"data_source": "steering_kernel",
}
```
**SK-06:** Proposals **mutieren nie** OM — Accept erfolgt explizit in UI/API.
**SK-07:** Abhängigkeiten (`parent_action_id`, Execution Graph, Epic) in `dependency_refs` / `dependency_blocked` — nicht stillschweigend ignorieren.
---
## 5. Methoden-Parität
| Methode / Element | Read Models | Proposals |
|-------------------|-------------|-----------|
| `work_cycle_scope` | epic_rollup (mit backlog_epic_hierarchy) | `sprint_commit` |
| `gate_fulfillment` | `planning_debt`, `execution_graph` | `gate_next_actions` |
| `backlog` ohne Sprint | — | `intake_triage` (`excludes work_cycle_scope`) |
| `queue_inbox` | — | `intake_triage` (Queue-Profil) |
| `critical_path` | `execution_graph` | (später path-basiert) |
**SK-08:** Neue Methode = Elemente im `MethodDefinition` + Provider — **kein** `if method_key == …` in Kernel.
**SK-09:** `agile_iteration` ist **Modifier**, keine Parallel-Welt — komponiert mit Primary (Product, Linear, …).
---
## 6. Frontend
**SK-10:** Steuerungs-UI liest `steeringSnapshot.steering_kernel.proposals` / `read_models` — nicht selbst sortieren.
**SK-11:** `SteeringProposalsPanel` + `PROPOSAL_UI_CONFIG` erweitern — **keine** neue Page pro Proposal-Typ.
**SK-12:** Capabilities aus `operatingContext.steering_elements` / `hasSteeringElement` — keine Archetyp-Ifs.
---
## 7. KI / Agent (eingefroren bis Principle Gate)
**SK-13:** Keine Prompts hardcodieren. Agent-Ranker implementieren als `ranker_key: agent_v1` + Actor-Slot — Kontext aus Snapshot/Operating Context.
**SK-14:** Bis Freigabe: Fallback auf `heuristic_v0` mit transparentem `factors[]`.
---
## 8. Tests & Abnahme
**SK-15:** Jeder neue Provider: Unit-Test (reine Logik) + Integration über Snapshot wenn DB verfügbar.
**SK-16:** Truth Table + ADP Extension Model §6 aktualisieren.
---
## 9. Verboten (Anti-Patterns)
- Bug-zuerst / Story-zuerst als globale Sortierregel ohne `ranker_key`-Kennzeichnung
- Proposal-Logik in `BacklogSection` / `PlanSprintPage`
- Duplicate Read Model (Snapshot + Kernel)
- Neue OM-Tabelle für Vorschläge
- Methoden-spezifische Kernel-Forks
---
## 10. Referenzen
| Dokument | Pfad |
|----------|------|
| Kernel Extension Model | `docs/architecture/ADP_Steering_Kernel_Extension_Model_v0.1.md` |
| Kernel Spine | `docs/architecture/ADP_Steering_Kernel_Spine_v0.1.md` |
| Cursor Rule | `.cursor/rules/kairo-steering-kernel.mdc` |
| Plugin-Architektur | `.cursor/rules/kairo-plugin-architecture.mdc` |
@@ -210,7 +210,9 @@ class AgentSlotProvider:
| Extension | agile_iteration | continuous_product | sequential_dependency | program_delivery | care_navigation |
|-----------|-----------------|--------------------|-----------------------|------------------|-----------------|
| `epic_rollup` | ✓ | — | — | ◐ (Epic=Liefercontainer) | — |
| `sprint_commit` proposal | ✓ | — | — | — | — |
| `sprint_commit` proposal | ✓ (work_cycle_scope) | ✓ | — | ◐ | ◐ |
| `gate_next_actions` proposal | — | ◐ | ✓ | ✓ | — |
| `intake_triage` proposal | — | ✓ (excl. sprint) | ✓ | ◐ | ✓ |
| `planning_debt` | ◐ | ◐ | ✓ | ✓ | — |
| `execution_graph` | ◐ | ◐ | ✓ | ✓ | ◐ |
| `intake_triage` proposal | ✓ | ✓ | ◐ | ◐ | ✓ |
@@ -226,7 +228,7 @@ class AgentSlotProvider:
| Phase | Inhalt | Entkoppelt |
|-------|--------|------------|
| **K-Ext-1** | `read_models/registry.py` + Kernel ruft Provider; P4 `epic_rollup` migriert | ✓ Kernel v0.3 |
| **K-Ext-2** | `attention/contributors/registry.py`; inline-Attention refactoren | offen |
| **K-Ext-2** | `attention/contributors/registry.py`; inline-Attention refactoren | ✓ AP2.2j |
| **K-Ext-3** | `proposals/registry.py` + `SteeringEvaluation.proposals` | ✓ AP2.2i (P6) |
| **K-Ext-4** | `agent_slots/registry.py` + Operating Context Feld `agent_slots` | P8 Reviews; Principle Gate für KI |
| **K-Ext-5** | Portfolio-Ebene: Workspace aggregiert Attention/Proposals über Initiativen | Program Director multi-initiative |
@@ -228,7 +228,8 @@ Details: `Kairo_Status_Review_and_Next_Steps_v0.1.md` §2.1
|----|-----------|
| AP2.3/4 | ✓ Plugin-Architektur + Element-Registry (2026-07-25) |
| AP2.2a | Starter-Kits ◐→✓ |
| ADP Backlog/Epic | P1–P4 ✓ · P6 Sprint-Vorschlag ✓ · P5 Agent-Tasks offen · K-Ext-2…4 offen |
| ADP Backlog/Epic | P1–P4 ✓ · P6 Sprint-Vorschlag ✓ · K-Ext-1/2/3 ✓ · Agent-Slots offen |
| Steering Kernel Proposals | ◐ | sprint_commit · gate_next_actions · intake_triage (Kernel v0.3) |
| AP2.2b–e | Referenz-Archetypen End-to-End ◐→✓ |
| AP2.1 | MVP-Abnahfe ✗→✓ |
| AP1.7b | Op-API Parität |