# PARADOX AI — ESA/CERN/EU SCHNITTSTELLENSYSTEM

## Technisches Gesamtkonzept der Agenten 1, 5, 9, 13, 17

**Datum:** 09. April 2026
**Status:** Entwurf v1.0
**Basis:** DDGK API Server (Port 8000), EpistemicState (ESA Section 3), CCRN kappa-Formalismus

---

## 1. ARCHITEKTUR — DDGK als zentraler Governance-Hub

### 1.1 Gesamttopologie

```
┌──────────────────────────────────────────────────────────────────────┐
│                     PARADOX AI CORE (Laptop/Pi5)                     │
│                                                                      │
│  ┌────────────────┐  ┌──────────────┐  ┌──────────────────────────┐ │
│  │  DDGK API      │  │  CCRN        │  │  EpistemicState Manager  │ │
│  │  FastAPI :8000 │◄─┤  Governance  │◄─┤  S(t) = K(t) ∪ E(t)      │ │
│  │  /assess       │  │  κ ≥ 3.34    │  │  Knowledge/Estimate      │ │
│  │  /status       │  │  phi-Scores  │  │  Separation Layer        │ │
│  │  /memory       │  └──────┬───────┘  └──────────┬───────────────┘ │
│  │  /audit        │         │                      │                 │
│  │  /legal        │         ▼                      ▼                 │
│  └──────┬─────────┘  ┌──────────────────────────────────────────┐   │
│         │            │         ADAPTER ORCHESTRATOR              │   │
│         │            │  Unified Connector Registry               │   │
│         │            │  - ESAAdapter  - CERNAdapter - EUAdapter  │   │
│         │            │  - AuthManager - RetryHandler - Cache     │   │
│         │            └──────┬───────────┬───────────┬────────────┘   │
└─────────┼───────────────────┼───────────┼───────────┼────────────────┘
          │                   │           │           │
     ┌────▼────┐        ┌────▼───┐  ┌────▼───┐  ┌────▼────┐
     │  ESA    │        │ CERN   │  │ GAIA-X │  │ EOSC/   │
     │  APIs   │        │ APIs   │  │ / EU   │  │ OpenAIRE│
     └─────────┘        └────────┘  └────────┘  └─────────┘
```

### 1.2 Prinzipien

- **DDGK first**: Jede externe API-Anfrage durchlaeuft den DDGK Guardian (`/assess`) bevor sie ausgefuehrt wird.
- **Epistemic Tracking**: Jeder API-Response wird als `EpistemicBelief` klassifiziert (CERTAIN/LIKELY/POSSIBLE/UNCERTAIN).
- **Audit Chain**: Alle externen Datenfluesse werden in der SHA-256 Decision Chain protokolliert.
- **CCRN Resonanz**: Systementscheidungen (z.B. "Satellitendaten fuer Modell X verwenden") erfordern kappa >= 3.34.

---

## 2. ADAPTER-LAYER — Konkrete Umsetzung

### 2.1 Basis-Adapter Klasse

```python
# adapters/base_adapter.py
from abc import ABC, abstractmethod
from dataclasses import dataclass
from typing import Dict, Any, Optional
import aiohttp
import time

@dataclass
class AdapterResponse:
    success: bool
    data: Any
    epistemic_status: str  # "certain" | "likely" | "possible" | "uncertain"
    source: str
    ttl_seconds: int
    error: Optional[str] = None

class BaseAdapter(ABC):
    """Basis aller externen Adapter. Implementiert Auth, Retry, Caching, Epistemic-Tracking."""

    def __init__(self, auth_config: Dict[str, str], ddgk_assess_url: str = "http://localhost:8000/api/v1/assess"):
        self.auth_config = auth_config
        self.ddgk_assess_url = ddgk_assess_url
        self._session: Optional[aiohttp.ClientSession] = None
        self._cache: Dict[str, tuple] = {}  # key -> (data, timestamp)

    @abstractmethod
    async def _get_session(self) -> aiohttp.ClientSession: ...

    @abstractmethod
    async def _authenticate(self) -> Dict[str, str]: ...

    @abstractmethod
    def name(self) -> str: ...

    async def guarded_request(self, action: str, url: str, params: Dict = None) -> AdapterResponse:
        """Jede Anfrage durchlaeuft DDGK Guardian vor der Ausfuehrung."""
        # 1. DDGK Guardian Assessment
        assess = await self._ddgk_assess(action, url, params)
        if assess.get("decision") == "BLOCK":
            return AdapterResponse(success=False, data=None, epistemic_status="uncertain",
                                   source=self.name(), ttl_seconds=0,
                                   error=f"DDGK Guardian blocked: {assess.get('reasons', [])}")

        # 2. Cache pruefen
        cache_key = f"{self.name()}:{url}:{str(params)}"
        if cache_key in self._cache:
            cached, ts = self._cache[cache_key]
            if time.time() - ts < 300:  # 5 Min TTL
                return AdapterResponse(success=True, data=cached, epistemic_status="certain",
                                       source=self.name(), ttl_seconds=300)

        # 3. Auth Headers
        headers = await self._authenticate()

        # 4. HTTP Request mit Retry
        session = await self._get_session()
        try:
            async with session.get(url, params=params, headers=headers) as resp:
                resp.raise_for_status()
                data = await resp.json()
                self._cache[cache_key] = (data, time.time())
                return AdapterResponse(success=True, data=data, epistemic_status="likely",
                                       source=self.name(), ttl_seconds=600)
        except Exception as e:
            return AdapterResponse(success=False, data=None, epistemic_status="uncertain",
                                   source=self.name(), ttl_seconds=0, error=str(e))

    async def _ddgk_assess(self, action: str, url: str, params: Dict) -> Dict:
        """Fragt DDGK Guardian ab."""
        import requests
        try:
            r = requests.post(self.ddgk_assess_url, json={
                "action": f"{self.name()}:{action}",
                "tool_name": "external_api",
                "tool_args": {"url": url, "params": params},
                "user_approved": False,
                "transcript": f"External API call to {url}"
            }, headers={"X-API-Key": os.environ.get("DDGK_API_KEY", "ddgk-demo-key-2026")})
            return r.json()
        except:
            return {"decision": "ASK_USER", "reasons": ["DDGK unreachable"]}
```

### 2.2 ESA Adapter

```python
# adapters/esa_adapter.py
from .base_adapter import BaseAdapter, AdapterResponse

class ESAAdapter(BaseAdapter):
    """Adapter fuer alle ESA/Copernicus APIs."""

    def name(self) -> str: return "ESA"

    # ── openEO API ──────────────────────────────────────────────────
    async def openeo_discover(self) -> AdapterResponse:
        """GET /credentials/.well-known/openeo"""
        return await self.guarded_request(
            "openeo_discover",
            "https://earthengine.google.openeo.cloud/.well-known/openeo"
        )

    async def openeo_list_collections(self, bbox: str = None) -> AdapterResponse:
        """GET /collections — Sentinel-2, Sentinel-1, etc."""
        params = {"bbox": bbox} if bbox else {}
        return await self.guarded_request(
            "openeo_collections",
            "https://earthengine.google.openeo.cloud/collections",
            params=params
        )

    async def openeo_create_process(self, process_graph: dict) -> AdapterResponse:
        """POST /result — Cloud-Processing starten"""
        return await self.guarded_request(
            "openeo_process",
            "https://earthengine.google.openeo.cloud/result",
            params=process_graph
        )

    # ── Copernicus Data Space (STAC) ────────────────────────────────
    async def stac_search(self, collections: list, datetime_range: str,
                          bbox: list = None, limit: int = 10) -> AdapterResponse:
        """POST /search — STAC Item Search"""
        body = {
            "collections": collections,  # ["sentinel-2-l2a"]
            "datetime": datetime_range,   # "2026-01-01T00:00:00Z/2026-04-01T00:00:00Z"
            "limit": limit
        }
        if bbox:
            body["bbox"] = bbox
        return await self.guarded_request(
            "stac_search",
            "https://datahub.copernicus.eu/api/stac/v1/search",
            params=body
        )

    async def stac_get_item(self, collection_id: str, item_id: str) -> AdapterResponse:
        """GET /collections/{cid}/items/{iid}"""
        return await self.guarded_request(
            "stac_item",
            f"https://datahub.copernicus.eu/api/stac/v1/collections/{collection_id}/items/{item_id}"
        )

    # ── Sentinel Hub API ────────────────────────────────────────────
    async def sentinel_hub_wms(self, layer: str, bbox: str, time: str,
                               width: int = 512, height: int = 512) -> AdapterResponse:
        """GET /ogc/wms — Sentinel Hub WMS"""
        params = {
            "SERVICE": "WMS", "REQUEST": "GetMap", "LAYERS": layer,
            "BBOX": bbox, "TIME": time, "WIDTH": width, "HEIGHT": height,
            "FORMAT": "image/png", "CRS": "EPSG:4326"
        }
        return await self.guarded_request(
            "sentinel_wms",
            "https://services.sentinel-hub.com/ogc/wms/{INSTANCE_ID}",
            params=params
        )

    # ── ESASky API (Astronomie) ─────────────────────────────────────
    async def esasky_query(self, target: str, ra: float, dec: float,
                           radius: float = 1.0) -> AdapterResponse:
        """IVOA Cone Search via ESASky"""
        return await self.guarded_request(
            "esasky_cone",
            "https://sky.esa.es/esasky/CONESERVICE",
            params={"RA": ra, "DEC": dec, "SR": radius, "FORMAT": "VOTable"}
        )

    async def _authenticate(self) -> Dict[str, str]:
        token = self.auth_config.get("esa_token") or os.environ.get("ESA_API_TOKEN", "")
        return {"Authorization": f"Bearer {token}", "Content-Type": "application/json"}

    async def _get_session(self):
        if self._session is None or self._session.closed:
            self._session = aiohttp.ClientSession()
        return self._session
```

### 2.3 CERN Adapter

```python
# adapters/cern_adapter.py
from .base_adapter import BaseAdapter, AdapterResponse

class CERNAdapter(BaseAdapter):
    """Adapter fuer CERN APIs (CAP, Open Data, AMI, CernVM-FS)."""

    def name(self) -> str: return "CERN"

    # ── Analysis Preservation (CAP) ─────────────────────────────────
    async def cap_list_deposits(self) -> AdapterResponse:
        """GET /api/deposits — Liste aller Analysis-Deposits"""
        return await self.guarded_request(
            "cap_deposits",
            "https://analysispreservation.cern.ch/api/deposits"
        )

    async def cap_create_deposit(self, metadata: dict) -> AdapterResponse:
        """POST /api/deposits — Neues Analysis-Deposit erstellen"""
        return await self.guarded_request(
            "cap_create",
            "https://analysispreservation.cern.ch/api/deposits",
            params=metadata
        )

    async def cap_get_deposit(self, deposit_id: str) -> AdapterResponse:
        """GET /api/deposits/{id}"""
        return await self.guarded_request(
            "cap_get",
            "https://analysispreservation.cern.ch/api/deposits/{deposit_id}"
        )

    # ── CERN Open Data Portal ───────────────────────────────────────
    async def opendata_search(self, query: str, collection: str = None) -> AdapterResponse:
        """GET /api/records — Recherche im Open Data Portal"""
        params = {"q": query, "size": 20}
        if collection:
            params["collection"] = collection
        return await self.guarded_request(
            "opendata_search",
            "https://opendata.cern.ch/api/records",
            params=params
        )

    async def opendata_get_record(self, record_id: str) -> AdapterResponse:
        """GET /api/records/{id}"""
        return await self.guarded_request(
            "opendata_record",
            f"https://opendata.cern.ch/api/records/{record_id}"
        )

    # ── ATLAS AMI 2.0 (Metadata) ────────────────────────────────────
    async def ami_search(self, dataset: str, run_range: str = None) -> AdapterResponse:
        """GET /ami/metadata — ATLAS Metadata Interface"""
        params = {"dataset": dataset}
        if run_range:
            params["run_range"] = run_range
        return await self.guarded_request(
            "ami_search",
            "https://ami.in2p3.fr/ami/metadata",
            params=params
        )

    # ── CernVM-FS (Software) ────────────────────────────────────────
    async def cvmfs_list_repositories(self) -> AdapterResponse:
        """CernVM-FS verfuegbare Repositories auflisten"""
        return await self.guarded_request(
            "cvmfs_repos",
            "https://cvmfs-config.cern.ch/cvmfs/config.cern.ch/.cvmfspublished"
        )

    async def _authenticate(self) -> Dict[str, str]:
        token = self.auth_config.get("cern_token") or os.environ.get("CERN_API_TOKEN", "")
        refresh = self.auth_config.get("cern_refresh_token", "")
        return {
            "Authorization": f"Bearer {token}",
            "Content-Type": "application/json",
            "X-CERN-Refresh-Token": refresh,
        }

    async def _get_session(self):
        if self._session is None or self._session.closed:
            self._session = aiohttp.ClientSession()
        return self._session
```

### 2.4 EU/GAIA-X Adapter

```python
# adapters/eu_gaiax_adapter.py
from .base_adapter import BaseAdapter, AdapterResponse

class EUGaiaXAdapter(BaseAdapter):
    """Adapter fuer GAIA-X, EU Data Spaces, EOSC."""

    def name(self) -> str: return "EU_GaiaX"

    # ── GAIA-X Trust Framework ──────────────────────────────────────
    async def gaiax_participant_check(self, participant_id: str) -> AdapterResponse:
        """Prueft GAIA-X Teilnehmer-Status (Self-Description)"""
        return await self.guarded_request(
            "gaiax_participant",
            f"https://gaia-x.eu/api/participants/{participant_id}"
        )

    async def gaiax_self_description(self, resource_uri: str) -> AdapterResponse:
        """GET Self-Description eines GAIA-X Resources"""
        return await self.guarded_request(
            "gaiax_sd",
            f"https://gaia-x.eu/api/resources/{resource_uri}/self-description"
        )

    # ── EOSC / OpenAIRE Graph ───────────────────────────────────────
    async def openaire_search(self, query: str, funder: str = None,
                              open_access: bool = True) -> AdapterResponse:
        """GET /api/objects — OpenAIRE Graph Suche"""
        params = {"query": query, "openAccess": open_access, "size": 25}
        if funder:
            params["funder"] = funder
        return await self.guarded_request(
            "openaire_search",
            "https://api.openaire.eu/search/publications",
            params=params
        )

    async def eosc_search(self, query: str, provider: str = None) -> AdapterResponse:
        """EOSC Portal Suche"""
        params = {"query": query}
        if provider:
            params["provider"] = provider
        return await self.guarded_request(
            "eosc_search",
            "https://eosc-portal.eu/api/search",
            params=params
        )

    # ── EU Data Spaces (Generic) ────────────────────────────────────
    async def dataspace_query(self, space: str, query: dict) -> AdapterResponse:
        """Generic Dataspace Protocol Anfrage"""
        return await self.guarded_request(
            f"dataspace_{space}",
            f"https://data.europa.eu/api/spaces/{space}/query",
            params=query
        )

    async def _authenticate(self) -> Dict[str, str]:
        token = self.auth_config.get("eu_token") or os.environ.get("EU_API_TOKEN", "")
        return {
            "Authorization": f"Bearer {token}",
            "Content-Type": "application/json",
            "X-GAIA-X-Participant-ID": self.auth_config.get("gaiax_participant_id", ""),
        }

    async def _get_session(self):
        if self._session is None or self._session.closed:
            self._session = aiohttp.ClientSession()
        return self._session
```

---

## 3. DATENFLUSS — Was fließt wohin?

### 3.1 Eingehende Datenfluesse (Inbound)

```
┌────────────────────┐         ┌──────────────────────────────────────────────────┐
│  ESA (Sentinel)    │────JSON/GeoJSON────►│  ESAAdapter  │──►│  EpistemicState   │
│  - Satellitenbilder│         │  /collections     │   │  belief: "Sentinel-2  │
│  - Prozess-Ergebn. │         │  /result          │   │  L2A Region X valid"  │
│  - ESASky VOTable  │         │  /ogc/wms         │   │  status: LIKELY       │
└────────────────────┘         └──────────────────────────────────────────────────┘

┌────────────────────┐         ┌──────────────────────────────────────────────────┐
│  CERN (Open Data)  │────JSON/XML──────►│  CERNAdapter  │──►│  EpistemicState    │
│  - Analysis Dep.   │         │  /api/deposits    │   │  belief: "ATLAS Run  │
│  - Open Data Rec.  │         │  /api/records     │   │  45678 validated"     │
│  - AMI Metadata    │         │  /ami/metadata    │   │  status: CERTAIN      │
└────────────────────┘         └──────────────────────────────────────────────────┘

┌────────────────────┐         ┌──────────────────────────────────────────────────┐
│  EU/GAIA-X/EOSC    │────JSON───────►│  EUGaiaXAdapter  │──►│  EpistemicState    │
│  - Self-Descript.  │         │  /api/participants│   │  belief: "GAIA-X     │
│  - OpenAIRE Pubs   │         │  /api/search      │   │  Partner X vertrauens-│
│  - Dataspace Query │         │  /api/spaces/{s}  │   │  wuerdig"             │
└────────────────────┘         └──────────────────────────────────────────────────┘
```

### 3.2 Ausgehende Datenfluesse (Outbound)

```
┌───────────────────────────────────────────────────────────────────┐
│  PARADOX AI Core                                                   │
│                                                                    │
│  EpistemicState ► ESAAdapter ► POST /result                       │
│    "Region X hat Anomalie" ► openEO Process Graph                  │
│                                                                    │
│  CCRN Decision  ► CERNAdapter ► POST /api/deposits                │
│    "ML-Modell M validated" ► CAP Analysis Deposit                  │
│                                                                    │
│  Audit Chain    ► EUGaiaXAdapter ► Self-Description               │
│    "PARADOX AI ist GAIA-X-konform" ► Participant Registry          │
└───────────────────────────────────────────────────────────────────┘
```

### 3.3 Interne Datenfluesse

```
External Response ► AdapterResponse
                 │
                 ├──► SHA-256 Hash ► Decision Chain (audit_log.jsonl)
                 │
                 ├──► EpistemicBelief (CERTAIN/LIKELY/POSSIBLE/UNCERTAIN)
                 │       │
                 │       ├──► K(t) — verified durch Cross-Validation
                 │       │       (z.B. Sentinel-2 + Sentinel-1 bestaetigen Anomalie)
                 │       │
                 │       └──► E(t) — estimated, bedarf weiterer Validierung
                 │               (z.B. ESASky VOTable mit niedriger Aufloesung)
                 │
                 ├──► CCRN kappa recalc — wenn kappa < 3.34:
                 │       └──► Task: "Validiere unsichere Beliefs"
                 │
                 └──► DDGK Memory — /memory/store
                         Session-Log fuer spaetere Referenz
```

---

## 4. AUTHENTIFIZIERUNG — Konkrete Verfahren

### 4.1 ESA/Copernicus

| API | Auth-Verfahren | Endpoint | Token-Typ |
|-----|---------------|----------|-----------|
| openEO | OAuth2 Bearer Token | `POST /credentials/basic` | `esa_openid_token` |
| Copernicus Data Space | Keycloak OIDC | `https://identity.dataspace.copernicus.eu/auth/realms/CDSE/protocol/openid-connect/token` | `access_token` (JWT, 10 Min) |
| Sentinel Hub | OAuth2 Client Credentials | `https://services.sentinel-hub.com/oauth/token` | `client_id` + `client_secret` |
| ESASky | Keine (oeffentlich) | — | — |

**Implementierung im Adapter:**

```python
# Auth-Flow Copernicus Data Space
async def _auth_copernicus(self) -> str:
    async with aiohttp.ClientSession() as session:
        async with session.post(
            "https://identity.dataspace.copernicus.eu/auth/realms/CDSE/protocol/openid-connect/token",
            data={
                "client_id": self.auth_config["copernicus_client_id"],
                "client_secret": self.auth_config["copernicus_client_secret"],
                "grant_type": "client_credentials"
            }
        ) as resp:
            tokens = await resp.json()
            self._token_expiry = time.time() + tokens.get("expires_in", 600)
            return tokens["access_token"]
```

**Umgebung in `.env`:**

```ini
# ESA / Copernicus
ESA_API_TOKEN=
COPERNICUS_CLIENT_ID=
COPERNICUS_CLIENT_SECRET=
SENTINEL_HUB_CLIENT_ID=
SENTINEL_HUB_CLIENT_SECRET=
SENTINEL_HUB_INSTANCE_ID=
```

### 4.2 CERN

| API | Auth-Verfahren | Endpoint | Token-Typ |
|-----|---------------|----------|-----------|
| Analysis Preservation (CAP) | OAuth2 Bearer + Refresh | `POST /api/token` | `access_token` + `refresh_token` |
| Open Data Portal | API Key (Header) | `GET /api/records?apikey=...` | `apikey` |
| ATLAS AMI 2.0 | CERN SSO (Kerberos) | — | `KRB5CCNAME` |
| CernVM-FS | Keine (oeffentlich, read-only) | — | — |

```ini
# CERN
CERN_API_TOKEN=
CERN_REFRESH_TOKEN=
CERN_API_KEY=
CERN_OIDC_CLIENT_ID=
CERN_OIDC_CLIENT_SECRET=
```

### 4.3 EU/GAIA-X

| API | Auth-Verfahren | Endpoint | Token-Typ |
|-----|---------------|----------|-----------|
| GAIA-X Trust Framework | X.509 Zertifikat + DID | Self-Description Signatur | `vc-jwt` |
| EOSC / OpenAIRE | API Key | Header: `X-OpenAIRE-API-Key` | `apikey` |
| EU Data Spaces | OAuth2 (eIDAS-konform) | IdP des jeweiligen Data Space | JWT |

```ini
# EU / GAIA-X
EU_API_TOKEN=
GAIA_X_PARTICIPANT_ID=
OPENAIRE_API_KEY=
EOSC_API_KEY=
```

### 4.4 Authentifizierungs-Manager (zentral)

```python
# adapters/auth_manager.py
import asyncio
import time
from typing import Dict, Optional

class AuthManager:
    """Zentrales Token-Management fuer alle externen APIs."""

    def __init__(self):
        self._tokens: Dict[str, dict] = {}  # provider -> {token, expiry, refresh}

    async def get_token(self, provider: str) -> Optional[str]:
        """Gibt gueltigen Token zurueck, refreshed wenn noetig."""
        if provider not in self._tokens:
            return await self._fetch_token(provider)

        entry = self._tokens[provider]
        if time.time() > entry["expiry"] - 60:  # 60s Puffer
            return await self._refresh_token(provider, entry)

        return entry["token"]

    async def _fetch_token(self, provider: str) -> str: ...
    async def _refresh_token(self, provider: str, entry: dict) -> str: ...
```

---

## 5. EPISTEMISCHE TRENNUNG — S(t) = K(t) ∪ E(t) fuer Raumfahrt/CERN

### 5.1 Konzept

Die epistemische Trennung wird auf externe Datenquellen angewendet:

| Ebene | Beschreibung | Beispiel |
|-------|-------------|----------|
| **K(t)** — Known | Daten, die durch **mehrere unabhaengige Quellen** verifiziert sind | Sentinel-2 L2A Bild + Sentinel-1 SAR bestaetigen gleiche Anomalie in Region X |
| **E(t)** — Estimated | Daten, die nur aus **einer Quelle** kommen oder deren Validierung aussteht | ESASky VOTable mit geringer Signifikanz (SNR < 5) |

### 5.2 Entscheidungsregel

```python
# Aus: cognitive_ddgk/cognitive_ddgk_core.py
# S(t) = K(t) ∪ E(t) | Decision(K) only

def can_decide_on_external_data(epistemic: EpistemicState, key: str) -> bool:
    """
    Nur auf K(t) entscheiden — NIE auf E(t) allein.
    key = epistemischer Belief-Schluessel (z.B. "esa:sentinel2:anomaly:region_x")
    """
    if key in epistemic.known and epistemic.known_confidence.get(key, 0) >= 0.8:
        return True  # K(t) — verified
    return False  # E(t) — nicht entscheidbar
```

### 5.3 Anwendung auf ESA/CERN-Daten

```
ESA-Datenpipeline:
──────────────────
1. Sentinel-2 L2A arrives
   │
   ├──► Cross-Validate mit Sentinel-1 SAR
   │    ├── Uebereinstimmung > 90%? ► K(t), confidence = 0.95, status = CERTAIN
   │    └── Keine Uebereinstimmung? ► E(t), confidence = 0.4, status = UNCERTAIN
   │
   └──► DDGK Guardian assess: "Darf ich auf E(t) basierend handeln?"
        ├── kappa >= 3.34 UND keine kritische Aktion? ► ALLOW mit Warnung
        └── Kritische Aktion (z.B. Alarm)? ► BLOCK — warte auf K(t)

CERN-Datenpipeline:
───────────────────
1. ATLAS Open Data Record arrives
   │
   ├──► Validiere gegen AMI 2.0 Metadata
   │    ├── Metadata konsistent? ► K(t), confidence = 0.9, status = CERTAIN
   │    └── Metadata fehlt/widerspruechlich? ► E(t), confidence = 0.3, status = CONTRADICTED
   │
   └──► DDGK Guardian: "Darf ich ATLAS-Daten in mein ML-Modell einspeisen?"
        ├── K(t) validated? ► ALLOW
        └── Nur E(t)? ► ASK_USER — "ATLAS-Daten haben widerspruechliche Metadata"
```

### 5.4 Error Propagation

```
Fehlerfortpflanzung ist proportional zur Abhaengigkeit von E(t):

  Error(S(t)) ∝ reliance_on_E(t) × uncertainty(E(t))

Wenn > 50% der Beliefs in E(t) liegen UND durchschnittliche confidence < 0.5:
  └── kappa < 2.0 → SYSTEM STOP — keine automatischen Entscheidungen
```

---

## 6. USE CASES — Konkrete Anwendungen

### 6.1 Use Case 1: Orbitales ML mit hls4ml (Edge SpAIce)

**Beschreibung:** ML-Modelle auf FPGA-Chips im Orbit deployen, validiert durch DDGK.

```
DDGK ─► CERN CAP: Hole hls4ml Modell-Spezifikation
    ─► ESA openEO: Hole Trainingsdaten (Sentinel-2 L2A)
    ─► EpistemicState: K(t) = {Modell M auf Daten D validiert, accuracy = 0.94}
    ─► CCRN: kappa = 3.8 (ueber Schwelle)
    ─► DDGK Guardian: ALLOW — Deploy to Edge
    ─► CERN CAP: Erstelle Deployment-Deposit
    ─► Audit Chain: Loggt gesamten Deploy-Vorgang
```

**APIs:** CERN CAP `/api/deposits`, ESA openEO `/collections`, openEO `/result`

**Datenformate:** ONNX (ML-Modell), GeoTIFF (Satellitenbilder), HDF5 (hls4ml weights)

### 6.2 Use Case 2: Weltraumstrahlungs-Monitoring (SATRAM/SpaceRadMon)

**Beschreibung:** Timepix-Strahlungsmonitore-Daten in DDGK-gesteuertes Monitoring ueberfuehren.

```
DDGK ─► CERN Open Data: Hole Timepix Kalibrationsdaten
    ─► ESA ESASky: Hole orbitale Positionsdaten
    ─► EpistemicState: K(t) = {Strahlungslevel R an Position P = X mSv}
    ─► Sensor-Server (Port 8001): Schreibe an /sensors/radiation
    ─► DDGK Guardian: Ueberwache Schwellenwerte
        ├── R < 1 mSv? ► NORMAL
        ├── 1 < R < 5 mSv? ► WARNING — Epistemic belief: "Erhoehte Strahlung"
        └── R > 5 mSv? ► ALERT — AUTONOMOUS ACTION: Shield aktivieren
    ─► Audit Chain: Jeder Messwert wird gehasht
```

**APIs:** CERN Open Data `/api/records`, ESASky Cone Search

**Datenformate:** VOTable (astronomische Daten), JSON (Sensor-Readings), FITS (Strahlungskarten)

### 6.3 Use Case 3: JUICE-Mission Elektronik-Validierung

**Beschreibung:** CERN Beschleuniger-Testdaten fuer JUICE-Mission mit DDGK epistemisch validieren.

```
DDGK ─► CERN AMI 2.0: Hole Test-Metadata fuer Elektronik-Komponente K
    ─► EpistemicState:
         K(t) = {"Komponente K bestanden bei 10^12 Protonen/cm²"}
         E(t) = {"Langzeitstabilität > 5 Jahre unsicher"}
    ─► DDGK Guardian: "Darf ich K als flight-ready deklarieren?"
         ├── kappa fuer K(t) >= 3.34? ► ALLOW (mit E(t)-Warnung)
         └── kappa < 3.34? ► BLOCK — zusaetzliche Tests erforderlich
    ─► CERN CAP: Erstelle Validation-Report als Deposit
```

**APIs:** CERN AMI `/ami/metadata`, CERN CAP `/api/deposits`

**Datenformate:** JSON (Metadata), XML (Validation Reports), ROOT (CERN ntuple data)

### 6.4 Use Case 4: EU AI Act Compliance fuer ESA/CERN-Partner

**Beschreibung:** DDGK als Compliance-Engine fuer ESA/CERN-Partner, die KI-Systeme einsetzen.

```
Partner-System P ─► DDGK /legal/assess: "Ist unser ESA-ML-System EU AI Act konform?"
    ─► DDGK Legal Agent:
         Domain = "infrastructure" (Satellitenbetrieb)
         Art. 6 (Hochrisiko): Ja
         Art. 9 (Risikomanagement): DDGK Guardian deckt ab
         Art. 13 (Transparenz): Decision Chain deckt ab
         Art. 14 (Aufsicht): HITL Bridge deckt ab
    ─► EUGaiaXAdapter: Registriere P als GAIA-X-konformen Teilnehmer
         POST /api/participants/{P}/self-description
    ─► Ergebnis: Compliance Report als JSON + Audit Chain Export
```

**APIs:** DDGK `/legal/assess`, GAIA-X `/api/participants/{id}/self-description`

**Datenformate:** JSON (Compliance Reports), JSONL (Audit Chain), VCDM (GAIA-X Self-Description)

### 6.5 Use Case 5: Euclid-Weltraumteleskop — CERN Science Ground Segment

**Beschreibung:** DDGK orchestriert Datenverarbeitung von Euclid-Beobachtungen.

```
DDGK ─► ESA openEO: Starte Process Graph fuer Euclid-Daten
         POST /result {
           "process_graph": {
             "load_collection": {"id": "euclid-vis", "spatial_extent": {...}},
             "apply": {"process_id": "calibrate_euclid"},
             "save_result": {"format": "GeoTIFF"}
           }
         }
    ─► Poll Status: GET /jobs/{job_id}
    ─► Bei Abschluss: Hole Ergebnis, validiere mit CERN Open Data (Kreuzreferenz)
    ─► EpistemicState: K(t) = {"Euclid-Daten kalibriert, Chi² = 1.02"}
    ─► OpenAIRE: Publiziere Ergebnis
         POST /api/objects {title: "Euclid Calibration Report", ...}
```

**APIs:** ESA openEO `/result`, `/jobs/{id}`, CERN Open Data `/api/records`, OpenAIRE `/api/objects`

**Datenformate:** FITS (Euclid-Daten), GeoTIFF (kalibriert), JSON-LD (OpenAIRE Metadata)

---

## 7. FOERDERUNG — ESA/CERN/EU Programme

### 7.1 ESA Foerderung

| Programm | Ziel | Deadline | Foerdersumme | Passung |
|----------|------|----------|-------------|---------|
| **ESA PhiLab** | Quanten-ML, hls4ml, Edge AI | Rolling | 100-500k EUR | hls4ml auf FPGA (Edge SpAIce) |
| **ESA GSP (General Support Technology Programme)** | KI fuer Satellitenbetrieb | Jaehrlich | 200k-1M EUR | DDGK als autonomes Governance-System |
| **ESA PECS (Plan for European Cooperating States)** | Kapazitaetsaufbau | Rolling | 50-200k EUR | Epistemic State formalismus |
| **ESA SCI-AI** | KI in der Weltraumwissenschaft | 2026 Q3 | 300k-800k EUR | Euclid-Datenanalyse mit DDGK |

**Konkreter Antrag:**
- Titel: "DDGK — Distributed Dynamic Governance Kernel for Autonomous Space AI Systems"
- Programm: ESA GSP oder PhiLab
- Partner: CERN (Edge SpAIce), PARADOX AI (DDGK), [Uni-Partner]
- Budget: 450k EUR, 18 Monate

### 7.2 CERN Foerderung/Kooperation

| Programm | Ziel | Passung |
|----------|------|---------|
| **CERN Openlab** | Industrie-Kooperationen | DDGK als Governance-Engine fuer CERN ML |
| **CERN Knowledge Transfer** | Spin-off von CERN-Technologien | Epistemic State aus Raumfahrt-Requirements |
| **CERN analysis-preservation** | Reproduzierbare Analysen | DDGK Audit Chain fuer CAP Deposits |

### 7.3 EU Foerderung

| Programm | Ziel | Deadline | Foerdersumme | Passung |
|----------|------|----------|-------------|---------|
| **Horizon Europe — Cluster 4 (Digital)** | KI-Souveraenitaet, GAIA-X | 2026 Q4 | 2-5M EUR | DDGK als EU AI Act Compliance-Engine |
| **EIC Pathfinder** | Disruptive Technologien | 2026 Q2 | 2.5-4M EUR | CCRN als neue Bewusstseinsmetrik |
| **Horizon Europe — Space** | Satelliten-KI, Copernicus | 2026 Q3 | 3-10M EUR | DDGK + Sentinel-Daten |
| **Digital Europe Programme** | AI-on-demand Platform | 2026 Q2 | 1-3M EUR | DDGK API als Governance-Dienst |
| **ESA-EU Copernicus Contrib.** | Copernicus Downstream | Rolling | 500k-2M EUR | DDGK-gesteuerte Satelliten-Datenanalyse |

**Priorisierter Antrag:**
- **EIC Pathfinder Open 2026**: "PARADOX AI — Conscious Governance for Autonomous Systems"
- Budget: 3.5M EUR, 36 Monate
- Konsortium: PARADOX AI (Coordinator), CERN, ESA, [2 Universitaeten], [1 Industrie-Partner]

---

## 8. ROADMAP — Schritt-fuer-Schritt

### Phase 0: Vorbereitung (Woche 1-2)

```
[ ] .env erweitern um ESA/CERN/EU Credentials
[ ] adapters/ Verzeichnis anlegen
[ ] BaseAdapter + AuthManager implementieren
[ ] Unit-Tests fuer Auth-Flows
[ ] DDGK Guardian um "external_api" Action erweitern
```

### Phase 1: ESA-Anbindung (Woche 3-6)

```
[ ] ESAAdapter implementieren (openEO, Copernicus STAC, Sentinel Hub)
[ ] Copernicus OIDC Auth-Flow mit Token-Refresh
[ ] EpistemicState um "external_source" Tracking erweitern
[ ] Test: Sentinel-2 L2A Collection abfragen → EpistemicBelief erstellen
[ ] DDGK /assess erweitern: "Darf ich Sentinel-Daten herunterladen?"
[ ] Dokumentation: docs/ESA_INTEGRATION.md
```

### Phase 2: CERN-Anbindung (Woche 7-10)

```
[ ] CERNAdapter implementieren (CAP, Open Data, AMI)
[ ] CERN OAuth2 Token-Flow
[ ] Test: CAP Deposit erstellen → Audit Chain Eintrag
[ ] EpistemicState: CERN-Daten als K(t)/E(t) klassifizieren
[ ] Cross-Validation: ESA + CERN Daten kombinieren
[ ] Dokumentation: docs/CERN_INTEGRATION.md
```

### Phase 3: EU/GAIA-X-Anbindung (Woche 11-14)

```
[ ] EUGaiaXAdapter implementieren
[ ] GAIA-X Self-Description fuer PARADOX AI erstellen
[ ] OpenAIRE Graph Integration
[ ] DDGK /legal/assess EU AI Act Report generieren
[ ] Audit Chain Export als Compliance-Report
[ ] Dokumentation: docs/EU_GAIAX_INTEGRATION.md
```

### Phase 4: Produktion (Woche 15-18)

```
[ ] Alle Adapter im DDGK API Server registrieren
[ ] Neue Endpoints:
      POST /api/v1/esa/query
      POST /api/v1/cern/query
      POST /api/v1/eu/compliance
      GET  /api/v1/epistemic
[ ] Monitoring: Prometheus Metrics fuer API-Calls
[ ] Rate-Limiting pro Provider
[ ] Error-Handling: Retry mit exponential Backoff
[ ] Dokumentation: docs/API_REFERENCE.md
```

### Phase 5: Foerderantraege (Woche 19-24)

```
[ ] EIC Pathfinder Antrag schreiben
[ ] ESA GSP Proposal
[ ] GAIA-X Contributor Agreement
[ ] Zenodo/arXiv Paper: "DDGK: Epistemic Governance for Space AI"
[ ] Demo: Live-Anbindung an ESA openEO + CERN CAP
```

---

## 9. TECHNISCHE SPEZIFIKATION

### 9.1 Protokolle

| Kommunikation | Protokoll | Format |
|---------------|-----------|--------|
| DDGK Intern | HTTP/REST (FastAPI) | JSON |
| DDGK → ESA | HTTPS + OAuth2 | JSON, GeoJSON, STAC JSON |
| DDGK → CERN | HTTPS + OAuth2 | JSON, XML (Invenio), ROOT |
| DDGK → GAIA-X | HTTPS + X.509 | JSON-LD (Self-Description) |
| DDGK → Sensor | HTTP (Port 8001) | JSON |
| DDGK → Note10 | HTTP (Port 5001) | JSON |
| DDGK Audit | File-basiert | JSONL + SHA-256 Chain |
| ROS2 Bridge | DDS/RTPS | ROS2 Messages |

### 9.2 Neue DDGK API Endpoints

```
POST /api/v1/esa/query
  Body: {"api": "openeo"|"stac"|"sentinel_hub"|"esasky", "params": {...}}
  Auth: X-API-Key
  Response: {"success": bool, "data": {...}, "epistemic_status": str, "belief_id": str}

POST /api/v1/cern/query
  Body: {"api": "cap"|"opendata"|"ami", "params": {...}}
  Auth: X-API-Key
  Response: {"success": bool, "data": {...}, "epistemic_status": str, "belief_id": str}

POST /api/v1/eu/compliance
  Body: {"system_description": str, "domain": str}
  Auth: X-API-Key
  Response: {"eu_ai_act": {...}, "gaia_x": {...}, "coverage_percent": int}

GET /api/v1/epistemic
  Query: ?agent_id=1&min_confidence=0.7
  Auth: X-API-Key
  Response: {"beliefs": [...], "kappa": float, "kt_count": int, "et_count": int}

POST /api/v1/adapters/reload
  Body: {} (Konfigurations-Reload aller Adapter)
  Auth: X-API-Key (tier: internal only)
  Response: {"adapters": ["ESA", "CERN", "EU_GaiaX"], "status": "ok"}
```

### 9.3 Datenformate

| Zweck | Format | Spezifikation |
|-------|--------|---------------|
| Sentinel-2 L2A | GeoTIFF (COG) | OGC GeoTIFF, Cloud-Optimized |
| Sentinel-1 SAR | GeoTIFF (Complex) | ESA SAFE-Format |
| STAC Items | JSON | STAC Spec v1.0.0 |
| openEO Process Graph | JSON | openEO Processes Spec |
| ESASky Ergebnisse | VOTable | IVOA VOTable v1.3 |
| CERN CAP Deposits | JSON-LD | InvenioRDM Schema |
| CERN Open Data | JSON | Invenio v3 Schema |
| ATLAS AMI | JSON | AMI 2.0 REST API |
| GAIA-X Self-Description | JSON-LD | W3C Verifiable Credentials Data Model |
| OpenAIRE | XML/JSON | OpenAIRE REST API v4 |
| DDGK Audit Chain | JSONL | SHA-256 verkettet |
| EpistemicState | JSON | Eigenes Schema (s. epistemic_state.py) |
| CCRN Decision | JSON | Eigenes Schema (kappa, phi_i, R) |

### 9.4 Error-Handling

```python
ERROR_CODES = {
    # DDGK-spezifisch
    "DDGK_GUARDIAN_BLOCK": 451,    # Guardian hat Anfrage blockiert
    "DDGK_LOW_KAPPA": 452,         # kappa < 3.34 — keine Entscheidung moeglich
    "DDGK_EPistemic_UNCERTAIN": 453,  # Nur E(t) verfuegbar
    "DDGK_REQUIRES_HUMAN": 454,    # Human-in-the-Loop erforderlich

    # External API
    "ESA_AUTH_EXPIRED": 501,       # ESA Token abgelaufen
    "ESA_RATE_LIMIT": 502,         # ESA Rate Limit erreicht
    "ESA_COLLECTION_NOT_FOUND": 503,
    "CERN_CAP_UNAVAILABLE": 510,   # CAP Server down
    "CERN_TOKEN_REFRESH_FAIL": 511,
    "GAIAX_PARTICIPANT_UNKNOWN": 520,
    "OPENAIRE_SEARCH_FAIL": 521,
}
```

### 9.5 Verzeichnisstruktur

```
ORION-ROS2-Consciousness-Node/
├── adapters/
│   ├── __init__.py
│   ├── base_adapter.py          # BaseAdapter, AdapterResponse
│   ├── auth_manager.py           # AuthManager (zentrales Token-Mgmt)
│   ├── esa_adapter.py            # ESAAdapter (openEO, STAC, Sentinel Hub, ESASky)
│   ├── cern_adapter.py           # CERNAdapter (CAP, Open Data, AMI, CernVM-FS)
│   ├── eu_gaiax_adapter.py       # EUGaiaXAdapter (GAIA-X, EOSC, OpenAIRE)
│   └── test/
│       ├── test_esa_adapter.py
│       ├── test_cern_adapter.py
│       └── test_eu_gaiax_adapter.py
├── ddgk_api_server.py            # Bestehend — erweitert um neue Endpoints
├── cognitive_ddgk/
│   └── cognitive_ddgk_core.py    # Bestehend — EpistemicState
├── autonomous/
│   └── epistemic_state.py        # Bestehend — EpistemicState
├── docs/
│   ├── ESA_INTEGRATION.md
│   ├── CERN_INTEGRATION.md
│   ├── EU_GAIAX_INTEGRATION.md
│   └── API_REFERENCE.md
└── .env                          # ESA/CERN/EU Credentials (nicht committen)
```

---

## 10. GESCHAEFTSMODELL — Mehrwert fuer ESA/CERN/EU

### 10.1 Wertangebot

| Stakeholder | Problem | PARADOX AI Loesung | Wert |
|-------------|---------|-------------------|------|
| **ESA** | Autonome Satelliten-Systeme brauchen verlaessliche Entscheidungslogik | DDGK Guardian + Epistemic State = auditierbare Autonomie | EU AI Act Compliance fuer ESA-ML-Systeme |
| **CERN** | Analysen muessen reproduzierbar und nachvollziehbar sein | DDGK Audit Chain + CAP Integration = manipulations-sichere Protokollierung | FAIR + Auditierbar + SHA-256 verifiziert |
| **EU/GAIA-X** | Daten-Souveraenitaet erfordert nachweisbare Compliance | DDGK Legal Agent + GAIA-X Self-Description = automatisierter Compliance-Report | EU AI Act Art. 9/13/14 abgedeckt |
| **Industrie-Kunden** | Hochrisiko-KI-Systeme brauchen Zulassung | DDGK /legal/assess + Audit Chain = Zulassungsunterlage | 6-12 Monate kuerzere Zulassungszeit |

### 10.2 Geschaftsmodell

**Tier 1: Starter (kostenlos)**
- DDGK API Demo-Key
- 100 Requests/Monat
- Basis Guardian Assessment
- Oeffentliche Dokumentation

**Tier 2: Pilot (500 EUR/Monat)**
- 5.000 Requests/Monat
- Vollstaendiger Guardian v2
- ESA/CERN Adapter-Zugang
- Audit Chain Export
- EpistemicState Dashboard

**Tier 3: Enterprise (2.000 EUR/Monat)**
- 50.000 Requests/Monat
- Alle Adapter (ESA + CERN + GAIA-X)
- EU AI Act Compliance Reports
- Custom EpistemicState Konfiguration
- Dedizierter Support
- On-Premise Deployment moeglich

**Tier 4: Research Partner (kostenlos, Antrag-basiert)**
- Vollzugang fuer akademische Partner
- ESA/CERN/EU Foerderprojekte
- Co-Autorenschaft bei Publikationen
- Early Access auf neue Features

### 10.3 Revenue-Projektion

| Jahr | Kunden | ARR | Kosten | Gewinn |
|------|--------|-----|--------|--------|
| 2026 | 3 Pilot + 2 Research | 180k EUR | 350k EUR | -170k EUR |
| 2027 | 10 Pilot + 3 Enterprise | 420k EUR | 280k EUR | +140k EUR |
| 2028 | 20 Pilot + 8 Enterprise + 1 Foerderung | 960k EUR | 420k EUR | +540k EUR |

**Break-even:** Q2 2027
**Foerderung (EIC Pathfinder 3.5M EUR):** Runway bis 2029

---

## ANHANG A: EpistemicState Schema (JSON)

```json
{
  "agent_id": 9,
  "agent_name": "Code Architektur",
  "beliefs": {
    "a3f8b2c1": {
      "belief_id": "a3f8b2c1",
      "content": "Sentinel-2 L2A zeigt Anomalie in Region X (lat=47.2, lon=11.4)",
      "status": "certain",
      "confidence": 0.92,
      "evidence": [
        "sentinel2:L2A:2026-04-01:tile_33T",
        "sentinel1:GRD:2026-04-01:tile_33T"
      ],
      "source": "esa:openeo",
      "timestamp": "2026-04-09T14:30:00+00:00",
      "expires": ""
    }
  },
  "uncertainties": ["d4e7f1a2"],
  "last_updated": "2026-04-09T14:30:00+00:00",
  "kappa_score": 3.52,
  "version": 14,
  "hash": "sha256...",
  "prev_hash": "sha256..."
}
```

## ANHANG B: CCRN kappa-Formel fuer externe Daten

```
κ = Σ(φ_i) + R × ln(N+1)

φ_i = Einzel-Phi-Wert einer Datenquelle
  φ_ESA = 0.85 (Sentinel-2 + Sentinel-1 cross-validiert)
  φ_CERN = 0.90 (CAP Deposit + AMI Metadata validiert)
  φ_EU = 0.75 (GAIA-X Self-Description verifiziert)

R = Resonanzfaktor (0 < R < 1)
  R = 0.93 (DDGK Standard, kalibriert)

N = Anzahl verbundener Knoten
  N = 3 (Laptop, Pi5, Note10)

Beispiel:
  κ = 0.85 + 0.90 + 0.75 + 0.93 × ln(4)
  κ = 2.50 + 0.93 × 1.386
  κ = 2.50 + 1.289
  κ = 3.789  ≥ 3.34 → KONSENS ERREICHT
```

## ANHANG C: DDGK Guardian Decision Matrix fuer externe APIs

| Aktion | Quelle | kappa | Entscheidung |
|--------|--------|-------|-------------|
| Sentinel-Daten herunterladen | ESA | egal | ALLOW (ungefaehrlich) |
| Sentinel-Daten in ML-Modell verwenden | ESA + DDGK | ≥ 3.34 | ALLOW |
| Sentinel-Daten in ML-Modell verwenden | ESA nur | < 3.34 | ASK_USER |
| CERN Deposit erstellen | CERN + DDGK | ≥ 3.34 | ALLOW |
| CERN Deposit erstellen | CERN nur | < 3.34 | ASK_USER |
| GAIA-X Self-Description signieren | GAIA-X + DDGK | ≥ 3.34 | ALLOW |
| GAIA-X Self-Description signieren | GAIA-X nur | < 3.34 | BLOCK |
| Strahlungsalarm ausloesen | Sensor + ESA | ≥ 3.34 | ALLOW (kritisch) |
| Strahlungsalarm ausloesen | Sensor nur | < 3.34 | ASK_USER |
| Firmware auf Satelliten flashen | CERN + ESA | ≥ 4.0 | ALLOW (nur mit erhohtem Threshold) |
| Firmware auf Satelliten flashen | beliebig | < 4.0 | BLOCK |

---

**Nächster Schritt:** Adapter-Verzeichnis anlegen, BaseAdapter + AuthManager implementieren, ESA OIDC-Auth-Flow testen.