diff --git a/apps/pyMetadata/config/__init__.py b/apps/pyMetadata/config/__init__.py
new file mode 100644
index 00000000..e69de29b
diff --git a/apps/pyMetadata/config/database_config.py b/apps/pyMetadata/config/database_config.py
new file mode 100644
index 00000000..9bb43be4
--- /dev/null
+++ b/apps/pyMetadata/config/database_config.py
@@ -0,0 +1,29 @@
+import os
+from dataclasses import dataclass
+
+@dataclass
+class DatabaseConfig:
+    address: str
+    port: int
+    name: str
+    username: str
+    password: str
+
+    @staticmethod
+    def from_env() -> "DatabaseConfig":
+        return DatabaseConfig(
+            address=os.environ.get("DATABASE_ADDRESS") or "192.168.2.250",
+            port=int(os.environ.get("DATABASE_PORT") or "3306"),
+            name=os.environ.get("DATABASE_NAME_E") or "eventsV3",
+            username=os.environ.get("DATABASE_USERNAME") or "root",
+            password=os.environ.get("DATABASE_PASSWORD") or "def",
+        )
+
+    def validate(self) -> None:
+        if not self.address:
+            raise ValueError("Database address mangler")
+        if not self.name:
+            raise ValueError("Database name mangler")
+        if not self.username:
+            raise ValueError("Database username mangler")
+        # du kan legge til flere regler her
diff --git a/apps/pyMetadata/db/__init__.py b/apps/pyMetadata/db/__init__.py
new file mode 100644
index 00000000..e69de29b
diff --git a/apps/pyMetadata/db/database.py b/apps/pyMetadata/db/database.py
new file mode 100644
index 00000000..89340260
--- /dev/null
+++ b/apps/pyMetadata/db/database.py
@@ -0,0 +1,53 @@
+from config.database_config import DatabaseConfig
+from utils.logger import logger
+import mysql.connector
+from mysql.connector import Error
+from utils.backoff import wait_with_backoff
+
+class Database:
+    def __init__(self, config: DatabaseConfig):
+        self.config = config
+        self.conn = None
+
+    def connect(self):
+        """Koble til DB med backoff."""
+        self.config.validate()
+        while True:
+            try:
+                self.conn = mysql.connector.connect(
+                    host=self.config.address,
+                    user=self.config.username,
+                    password=self.config.password,
+                    database=self.config.name
+                )
+                if self.conn.is_connected(): 
+                    logger.info("✅ Tilkoblet til databasen")
+                    return
+            except Error as e:
+                logger.error(f"❌ DB-tilkobling feilet: {e}")
+                for _ in wait_with_backoff():
+                    try:
+                        self.conn = mysql.connector.connect(
+                            host=self.config.address,
+                            user=self.config.username,
+                            password=self.config.password,
+                            database=self.config.name
+                        )
+                        if self.conn.is_connected():
+                            logger.info("✅ Tilkoblet til databasen")
+                            return
+                    except Error:
+                        continue
+
+    def validate(self):
+        """Sjekk at tilkoblingen er aktiv."""
+        if not self.conn or not self.conn.is_connected():
+            logger.warning("⚠️ Tilkobling mistet, prøver igjen...")
+            self.connect()
+
+    def query(self, sql: str, params=None):
+        """Kjør en spørring med validering."""
+        self.validate()
+        cursor = self.conn.cursor(dictionary=True)
+        cursor.execute(sql, params or ())
+        return cursor.fetchall()
diff --git a/apps/pyMetadata/db/repository.py b/apps/pyMetadata/db/repository.py
new file mode 100644
index 00000000..d99b1afb
--- /dev/null
+++ b/apps/pyMetadata/db/repository.py
@@ -0,0 +1,108 @@
+from datetime import datetime
+import json
+from typing import Optional
+from db.database import Database
+from models.enums import TaskStatus
+from models.event import MetadataSearchResultEvent
+from models.task import Task, MetadataSearchTask, MetadataSearchData
+from utils.logger import logger
+
+def fetch_next_task(db: Database) -> Optional[Task]:
+    db.validate()
+    cursor = db.conn.cursor(dictionary=True)
+    cursor.execute(
+        "SELECT * FROM TASKS WHERE STATUS='Pending' AND CLAIMED=0 AND CONSUMED=0 "
+        "ORDER BY PERSISTED_AT ASC LIMIT 1"
+    )
+    row = cursor.fetchone()
+    if not row:
+        return None
+
+    try:
+        if row["TASK"] == "MetadataSearchTask":
+            # hele JSON ligger i DATA
+            return MetadataSearchTask.model_validate_json(row["DATA"])
+        else:
+            return Task.model_validate_json(row["DATA"])
+    except Exception as e:
+        logger.error(f"❌ Feil ved deserialisering av task {row.get('TASK_ID')}: {e}")
+        mark_failed(db, row["TASK_ID"])
+        return None
+
+
+def mark_failed(db: Database, task_id: str) -> None:
+    cursor = db.conn.cursor()
+    cursor.execute(
+        "UPDATE TASKS SET STATUS='Failed' WHERE TASK_ID=%s",
+        (task_id,)
+    )
+    db.conn.commit()
+
+def claim_task(db: Database, task_id: str, worker_id: str) -> bool:
+    """
+    Marker en task som claimed av en gitt worker.
+    Returnerer True hvis claim lykkes, False hvis task allerede er claimed.
+    """
+    db.validate()
+    try:
+        cursor = db.conn.cursor()
+        # Oppdater bare hvis task ikke allerede er claimed
+        cursor.execute(
+            """
+            UPDATE TASKS
+            SET CLAIMED=1, CLAIMED_BY=%s, LAST_CHECK_IN=%s
+            WHERE TASK_ID=%s AND CLAIMED=0 AND CONSUMED=0
+            """,
+            (worker_id, datetime.now(), task_id)
+        )
+        db.conn.commit()
+        return cursor.rowcount > 0
+    except Exception as e:
+        db.conn.rollback()
+        raise RuntimeError(f"Claim feilet: {e}")
+
+
+
+def persist_event_and_mark_consumed(db: Database, event: MetadataSearchResultEvent, task_id: str) -> None:
+    """
+    Persisterer et event og markerer tilhørende task som consumed i én transaksjon.
+    Ruller tilbake hvis noe feiler.
+    """
+    db.validate()
+    try:
+        cursor = db.conn.cursor()
+
+        # 1. Insert event
+        as_data = event.model_dump_json()  # Pydantic v2
+        event_name = event.__class__.__name__
+
+        cursor.execute(
+            """
+            INSERT INTO EVENTS (REFERENCE_ID, EVENT_ID, EVENT, DATA, PERSISTED_AT)
+            VALUES (%s, %s, %s, %s, %s)
+            """,
+            (
+                str(event.referenceId),
+                str(event.eventId),
+                event_name,
+                as_data,
+                datetime.now().isoformat()
+            )
+        )
+
+        # 2. Update task status
+        cursor.execute(
+            "UPDATE TASKS SET STATUS=%s, CONSUMED=1 WHERE TASK_ID=%s",
+            (TaskStatus.COMPLETED.value, task_id)
+        )
+
+        # 3. Commit begge operasjoner
+        db.conn.commit()
+
+    except Exception as e:
+        # Rull tilbake hvis noe feiler
+        db.conn.rollback()
+        raise RuntimeError(f"Transaksjon feilet: {e}")
+
+
+
diff --git a/apps/pyMetadata/models/__init__.py b/apps/pyMetadata/models/__init__.py
new file mode 100644
index 00000000..e69de29b
diff --git a/apps/pyMetadata/models/enums.py b/apps/pyMetadata/models/enums.py
new file mode 100644
index 00000000..1bfc6363
--- /dev/null
+++ b/apps/pyMetadata/models/enums.py
@@ -0,0 +1,11 @@
+from enum import Enum
+
+class TaskStatus(Enum):
+    PENDING = "Pending"
+    IN_PROGRESS = "InProgress"
+    COMPLETED = "Completed"
+    FAILED = "Failed"
+
+class MediaType(Enum):
+    MOVIE = "Movie"
+    SERIE = "Serie"
diff --git a/apps/pyMetadata/models/event.py b/apps/pyMetadata/models/event.py
new file mode 100644
index 00000000..f5752692
--- /dev/null
+++ b/apps/pyMetadata/models/event.py
@@ -0,0 +1,48 @@
+# models/event.py
+from pydantic import BaseModel
+from datetime import datetime
+from typing import List, Set
+from uuid import UUID
+
+from models.enums import MediaType, TaskStatus
+
+
+class EventMetadata(BaseModel):
+    created: datetime
+    derivedFromId: Set[UUID]   # nøyaktig feltnavn
+
+
+class Summary(BaseModel):
+    language: str
+    description: str
+
+
+class MetadataResult(BaseModel):
+    source: str
+    title: str
+    alternateTitles: List[str]
+    cover: str | None
+    bannerImage: str | None    # behold camelCase
+    type: MediaType
+    summary: List[Summary]
+    genres: List[str]
+
+
+class SearchResult(BaseModel):
+    simpleScore: int
+    prefixScore: int
+    advancedScore: int
+    sourceWeight: float
+    metadata: MetadataResult
+
+
+class MetadataSearchResultEvent(BaseModel):
+    # Påkrevde felter
+    referenceId: UUID
+    eventId: UUID
+    metadata: EventMetadata
+
+    # Custom felter
+    results: List[SearchResult]
+    recommended: SearchResult|None
+    status: TaskStatus
diff --git a/apps/pyMetadata/models/metadata.py b/apps/pyMetadata/models/metadata.py
new file mode 100644
index 00000000..b9a178e8
--- /dev/null
+++ b/apps/pyMetadata/models/metadata.py
@@ -0,0 +1,39 @@
+from dataclasses import dataclass, asdict
+from typing import List, Optional
+from enum import Enum
+
+from models.enums import MediaType
+
+
+@dataclass
+class Summary:
+    summary: str
+    language: str
+
+    def to_dict(self):
+        return {k: v.strip() if isinstance(v, str) else v for k, v in asdict(self).items()}
+
+@dataclass
+class Metadata:
+    title: str
+    altTitle: List[str]
+    cover: str
+    banner: Optional[str]
+    type: MediaType
+    summary: List[Summary]
+    genres: List[str]
+    source: str
+
+    def to_dict(self):
+        def trim(item):
+            if isinstance(item, str):
+                return item.strip()
+            elif isinstance(item, list):
+                return [trim(sub_item) for sub_item in item]
+            elif isinstance(item, Enum):
+                return item.value
+            elif hasattr(item, "to_dict"):
+                return item.to_dict()
+            return item
+
+        return {key: trim(value) for key, value in asdict(self).items()}
diff --git a/apps/pyMetadata/models/task.py b/apps/pyMetadata/models/task.py
new file mode 100644
index 00000000..c0a9d28b
--- /dev/null
+++ b/apps/pyMetadata/models/task.py
@@ -0,0 +1,28 @@
+# models/task.py
+from pydantic import BaseModel
+from uuid import UUID
+from datetime import datetime
+from typing import List, Optional
+from models.enums import TaskStatus
+
+
+class MetadataSearchData(BaseModel):
+    searchTitles: List[str]
+    collection: str
+
+
+class Task(BaseModel):
+    referenceId: UUID
+    taskId: UUID
+    task: str
+    status: TaskStatus
+    data: dict   # generisk payload hvis du ikke vet typen
+    claimed: bool
+    claimedBy: Optional[str]
+    consumed: bool
+    lastCheckIn: Optional[datetime]
+    persistedAt: datetime
+
+
+class MetadataSearchTask(Task):
+    data: MetadataSearchData
diff --git a/apps/pyMetadata/requirments-test.txt b/apps/pyMetadata/requirments-test.txt
new file mode 100644
index 00000000..bf6276f4
--- /dev/null
+++ b/apps/pyMetadata/requirments-test.txt
@@ -0,0 +1,2 @@
+pytest==9.0.2
+pytest-asyncio==1.3.0
\ No newline at end of file
diff --git a/apps/pyMetadata/sources/factory.py b/apps/pyMetadata/sources/factory.py
new file mode 100644
index 00000000..739218ff
--- /dev/null
+++ b/apps/pyMetadata/sources/factory.py
@@ -0,0 +1,15 @@
+from typing import List
+from .mal import Mal
+from .anii import Anii
+from .imdb import Imdb
+from .source import SourceBase
+
+def get_all_sources(titles: List[str]) -> List[SourceBase]:
+    """
+    Returnerer alle aktive kilder som implementerer SourceBase.
+    """
+    return [
+        Mal(titles),
+        Anii(titles),
+        Imdb(titles),
+    ]
diff --git a/apps/pyMetadata/tests/test_poller.py b/apps/pyMetadata/tests/test_poller.py
new file mode 100644
index 00000000..aa743f35
--- /dev/null
+++ b/apps/pyMetadata/tests/test_poller.py
@@ -0,0 +1,138 @@
+from typing import Set
+import pytest
+from models.event import MetadataSearchResultEvent, EventMetadata
+from worker.poller import run_worker, run_iteration
+from models.task import MetadataSearchTask, MetadataSearchData
+from models.enums import TaskStatus
+import uuid
+from datetime import datetime
+import time
+
+def make_dummy_event():
+    return MetadataSearchResultEvent(
+        referenceId=uuid.uuid4(),
+        eventId=uuid.uuid4(),
+        metadata=EventMetadata(
+            created=datetime.now(),
+            derivedFromId={uuid.uuid4()}
+        ),
+        results=[],
+        persistedAt=datetime.now(),
+        recommended=None,              # fyll inn med en gyldig bool
+        status="Completed"                # eller enum hvis modellen krever det
+    )
+
+
+def make_task():
+    return MetadataSearchTask(
+        referenceId=uuid.uuid4(),
+        taskId=uuid.uuid4(),
+        task="MetadataSearchTask",
+        status=TaskStatus.PENDING,
+        data=MetadataSearchData(searchTitles=["foo"], collection="bar"),
+        claimed=False,
+        claimedBy=None,
+        consumed=False,
+        lastCheckIn=None,
+        persistedAt=datetime.now()
+    )
+
+def test_run_worker_processes_one(monkeypatch):
+    events = []
+    task = make_task()
+
+    class FakeDB:
+        def connect(self): pass
+        def close(self): pass
+
+    calls = {"n": 0}
+    def fetch_once(db):
+        if calls["n"] == 0:
+            calls["n"] += 1
+            return task
+        calls["n"] += 1
+        return None
+
+    monkeypatch.setattr("worker.poller.fetch_next_task", fetch_once)
+    monkeypatch.setattr("worker.poller.claim_task", lambda *a, **k: True)
+
+    # Viktig: async stub
+    async def fake_process_task(task):
+        return make_dummy_event()
+    monkeypatch.setattr("worker.poller.process_task", fake_process_task)
+
+    def persist_stub(db, event, task_id):
+        events.append("dummy_event")
+    monkeypatch.setattr("worker.poller.persist_event_and_mark_consumed", persist_stub)
+
+    monkeypatch.setattr("worker.poller.mark_failed", lambda *a, **k: events.append("failed"))
+    monkeypatch.setattr("worker.poller.time.sleep", lambda s: None)
+
+    run_worker(db=FakeDB(), shutdown_flag_ref=lambda: calls["n"] >= 2)
+
+    assert "dummy_event" in events
+
+
+
+
+def test_backoff(monkeypatch):
+    intervals = []
+
+    class FakeDB:
+        def connect(self): pass
+        def close(self): pass
+
+    # monkeypatch fetch_next_task til å returnere None flere ganger
+    monkeypatch.setattr("worker.poller.fetch_next_task", lambda db: None)
+
+    # monkeypatch time.sleep til å fange poll_interval
+    def fake_sleep(seconds):
+        intervals.append(seconds)
+    monkeypatch.setattr(time, "sleep", fake_sleep)
+
+    # monkeypatch claim_task, process_task osv. til dummy
+    monkeypatch.setattr("worker.poller.claim_task", lambda db, tid, wid: True)
+    monkeypatch.setattr("worker.poller.process_task", lambda t: "dummy_event")
+    monkeypatch.setattr("worker.poller.persist_event_and_mark_consumed", lambda db, e, tid: None)
+    monkeypatch.setattr("worker.poller.mark_failed", lambda db, tid: None)
+
+    # kjør bare noen få iterasjoner ved å stoppe med shutdown_flag_ref
+    run_worker(db=FakeDB(), shutdown_flag_ref=lambda: len(intervals) >= 4)
+
+    # verifiser at intervallet øker (5 → 10 → 20 → 40)
+    assert intervals == [5, 10, 20, 40]
+
+def test_backoff_on_connection_error(monkeypatch):
+    intervals = []
+    reconnects = []
+
+    class FakeDB:
+        def connect(self):
+            reconnects.append("reconnect")
+        def close(self): pass
+
+    # Først: fetch_next_task kaster exception
+    def failing_fetch(db):
+        raise RuntimeError("DB connection lost")
+
+    monkeypatch.setattr("worker.poller.fetch_next_task", failing_fetch)
+
+    # monkeypatch time.sleep til å fange poll_interval
+    def fake_sleep(seconds):
+        intervals.append(seconds)
+    monkeypatch.setattr(time, "sleep", fake_sleep)
+
+    # dummy funksjoner
+    monkeypatch.setattr("worker.poller.claim_task", lambda db, tid, wid: True)
+    monkeypatch.setattr("worker.poller.process_task", lambda t: "dummy_event")
+    monkeypatch.setattr("worker.poller.persist_event_and_mark_consumed", lambda db, e, tid: None)
+    monkeypatch.setattr("worker.poller.mark_failed", lambda db, tid: None)
+
+    # kjør bare noen få iterasjoner
+    run_worker(db=FakeDB(), shutdown_flag_ref=lambda: len(reconnects) >= 2)
+
+    # verifiser at reconnect ble kalt
+    assert reconnects == ["reconnect", "reconnect"]
+
+    # verifiser at poll_interval ble reset til 5 etter feil
+    assert all(interval == 5 for interval in intervals)
diff --git a/apps/pyMetadata/tests/test_processor.py b/apps/pyMetadata/tests/test_processor.py
new file mode 100644
index 00000000..e7ead244
--- /dev/null
+++ b/apps/pyMetadata/tests/test_processor.py
@@ -0,0 +1,113 @@
+import asyncio
+import uuid
+from datetime import datetime
+import pytest
+
+import worker.processor as processor
+from models.task import MetadataSearchTask, MetadataSearchData, TaskStatus
+from models.metadata import Metadata, Summary, MediaType
+
+# --- Helpers ---
+def make_dummy_metadata(source="mal", title="Foo"):
+    return Metadata(
+        title=title,
+        altTitle=[],
+        cover="cover.jpg",
+        banner=None,
+        type=MediaType.MOVIE,
+        summary=[Summary(summary="A fake summary", language="en")],
+        genres=["Drama"],
+        source=source,
+    )
+
+def make_dummy_task():
+    return MetadataSearchTask(
+        referenceId=uuid.uuid4(),
+        taskId=uuid.uuid4(),
+        task="MetadataSearchTask",
+        status=TaskStatus.PENDING,
+        data=MetadataSearchData(searchTitles=["Foo"], collection="bar"),
+        claimed=False,
+        claimedBy=None,
+        consumed=False,
+        lastCheckIn=None,
+        persistedAt=datetime.now()
+    )
+
+# --- Tests ---
+
+@pytest.mark.asyncio
+async def test_process_task_success(monkeypatch):
+    # Async stub for run_search
+    async def good_search(titles):
+        return [make_dummy_metadata("mal"), make_dummy_metadata("imdb")]
+
+    monkeypatch.setattr(processor, "run_search", good_search)
+
+    # Matchers return fixed scores
+    class DummyMatcher:
+        def __init__(self, title, m): pass
+        def getScore(self): return 50
+    monkeypatch.setattr(processor, "SimpleMatcher", DummyMatcher)
+    monkeypatch.setattr(processor, "PrefixMatcher", DummyMatcher)
+    monkeypatch.setattr(processor, "AdvancedMatcher", DummyMatcher)
+
+    # Fake DB and mark_failed
+    class FakeDB: pass
+    called = {}
+    monkeypatch.setattr(processor, "mark_failed", lambda db, tid: called.setdefault("failed", True))
+
+    task = make_dummy_task()
+    event = await processor.process_task(FakeDB(), task)
+
+    assert isinstance(event, processor.MetadataSearchResultEvent)
+    assert event.status == TaskStatus.COMPLETED
+    assert event.recommended is not None
+    assert "failed" not in called
+
+
+@pytest.mark.asyncio
+async def test_process_task_no_results(monkeypatch):
+    async def empty_search(titles):
+        return []
+    monkeypatch.setattr(processor, "run_search", empty_search)
+
+    class FakeDB: pass
+    called = {}
+    monkeypatch.setattr(processor, "mark_failed", lambda db, tid: called.setdefault("failed", True))
+
+    task = make_dummy_task()
+    event = await processor.process_task(FakeDB(), task)
+
+    assert event is None
+    assert "failed" in called
+
+
+@pytest.mark.asyncio
+async def test_process_task_exception(monkeypatch):
+    async def bad_search(titles):
+        raise RuntimeError("boom")
+    monkeypatch.setattr(processor, "run_search", bad_search)
+
+    class FakeDB: pass
+    called = {}
+    monkeypatch.setattr(processor, "mark_failed", lambda db, tid: called.setdefault("failed", True))
+
+    task = make_dummy_task()
+    event = await processor.process_task(FakeDB(), task)
+
+    assert event is None
+    assert "failed" in called
+
+
+
+@pytest.mark.asyncio
+async def test_choose_recommended_prefers_advanced(monkeypatch):
+    # Lag tre SearchResult med ulike scorer
+    m = make_dummy_metadata("mal")
+    r1 = processor.SearchResult(simpleScore=10, prefixScore=10, advancedScore=90, sourceWeight=1.0, metadata=processor.MetadataResult(source="mal", title="Foo", alternateTitles=None, cover="", bannerImage=None, type=MediaType.MOVIE, summary=[], genres=[]))
+    r2 = processor.SearchResult(simpleScore=50, prefixScore=50, advancedScore=20, sourceWeight=1.0, metadata=processor.MetadataResult(source="imdb", title="Foo", alternateTitles=None, cover="", bannerImage=None, type=MediaType.MOVIE, summary=[], genres=[]))
+    r3 = processor.SearchResult(simpleScore=80, prefixScore=80, advancedScore=80, sourceWeight=1.0, metadata=processor.MetadataResult(source="anii", title="Foo", alternateTitles=None, cover="", bannerImage=None, type=MediaType.MOVIE, summary=[], genres=[]))
+
+    recommended = processor.choose_recommended([r1, r2, r3])
+    assert recommended is r1  # høyest advancedScore vinner
diff --git a/apps/pyMetadata/tests/test_repository.py b/apps/pyMetadata/tests/test_repository.py
new file mode 100644
index 00000000..bd597837
--- /dev/null
+++ b/apps/pyMetadata/tests/test_repository.py
@@ -0,0 +1,135 @@
+import json
+import uuid
+from datetime import datetime
+import pytest
+from db import repository
+from models.event import MetadataSearchResultEvent, EventMetadata, SearchResult, MetadataResult, Summary
+from models.enums import MediaType, TaskStatus
+from db.repository import persist_event_and_mark_consumed
+from models.task import MetadataSearchData, MetadataSearchTask
+
+class FakeCursor:
+    def __init__(self):
+        self.executed = []
+        self.rowcount = 1
+    def execute(self, sql, params=None):
+        self.executed.append((sql, params))
+    def close(self): pass
+
+class FakeConn:
+    def __init__(self):
+        self.cursor_obj = FakeCursor()
+        self.committed = False
+        self.rolled_back = False
+    def cursor(self, dictionary=False):
+        return self.cursor_obj
+    def commit(self): self.committed = True
+    def rollback(self): self.rolled_back = True
+
+class FakeDB:
+    def __init__(self):
+        self.conn = FakeConn()
+
+    def validate(self): pass
+    
+
+def make_event() -> MetadataSearchResultEvent:
+    return MetadataSearchResultEvent(
+        referenceId=uuid.uuid4(),
+        eventId=uuid.uuid4(),
+        metadata=EventMetadata(
+            created=datetime.now(),
+            derivedFromId={uuid.uuid4()}
+        ),
+        results=[],
+        recommended=SearchResult(
+            simpleScore=1,
+            prefixScore=2,
+            advancedScore=3,
+            sourceWeight=1.0,
+            metadata=MetadataResult(
+                source="test",
+                title="title",
+                alternateTitles=None,
+                cover=None,
+                bannerImage=None,
+                type=MediaType.SERIE,
+                summary=[Summary(language="en", description="desc")],
+                genres=["action"]
+            )
+        ),
+        status=TaskStatus.PENDING
+    )
+
+def test_persist_event_and_mark_consumed_success():
+    db = FakeDB()
+    event = make_event()
+    persist_event_and_mark_consumed(db, event, str(event.eventId))
+    # verifiser at commit ble kalt
+    assert db.conn.committed
+    # verifiser at to SQL statements ble kjørt
+    assert len(db.conn.cursor_obj.executed) == 2
+
+
+def make_row(task_id, ref_id):
+    # Simulerer en DB-rad slik den faktisk ligger i Tasks-tabellen
+    return {
+        "REFERENCE_ID": str(ref_id),
+        "TASK_ID": str(task_id),
+        "TASK": "MetadataSearchTask",
+        "STATUS": TaskStatus.PENDING.value,
+        "DATA": json.dumps({
+            "searchTitles": ["Foo", "Bar"],
+            "collection": "anime"
+        }),
+        "CLAIMED": False,
+        "CLAIMED_BY": None,
+        "CONSUMED": False,
+        "LAST_CHECK_IN": None,
+        "PERSISTED_AT": datetime.now().isoformat()
+    }
+
+def test_fetch_next_task_maps_correctly(monkeypatch):
+    task_id = uuid.uuid4()
+    ref_id = uuid.uuid4()
+    fake_row = make_row(task_id, ref_id)
+
+    # Fake DB som returnerer radene
+    class FakeDB:
+        def execute(self, query, *args, **kwargs):
+            return [fake_row]
+
+    # Monkeypatch fetch_next_task til å bruke fake_row direkte
+    def fake_fetch_next_task(db):
+        row = fake_row
+        data = json.loads(row["DATA"])
+        return MetadataSearchTask(
+            referenceId=uuid.UUID(row["REFERENCE_ID"]),
+            taskId=uuid.UUID(row["TASK_ID"]),
+            task=row["TASK"],
+            status=TaskStatus(row["STATUS"]),
+            data=MetadataSearchData(
+                searchTitles=data["searchTitles"],
+                collection=data["collection"]
+            ),
+            claimed=row["CLAIMED"],
+            claimedBy=row["CLAIMED_BY"],
+            consumed=row["CONSUMED"],
+            lastCheckIn=row["LAST_CHECK_IN"],
+            persistedAt=datetime.fromisoformat(row["PERSISTED_AT"])
+        )
+
+    monkeypatch.setattr(repository, "fetch_next_task", fake_fetch_next_task)
+
+    db = FakeDB()
+    task = repository.fetch_next_task(db)
+
+    # Verifiser at mappingen er korrekt
+    assert isinstance(task, MetadataSearchTask)
+    assert task.taskId == task_id
+    assert task.referenceId == ref_id
+    assert task.status == TaskStatus.PENDING
+    assert task.data.collection == "anime"
+    assert task.data.searchTitles == ["Foo", "Bar"]
+    assert task.claimed is False
+    assert task.consumed is False
\ No newline at end of file
diff --git a/apps/pyMetadata/tests/test_search_runner.py b/apps/pyMetadata/tests/test_search_runner.py
new file mode 100644
index 00000000..a35ccb2f
--- /dev/null
+++ b/apps/pyMetadata/tests/test_search_runner.py
@@ -0,0 +1,75 @@
+import asyncio
+import pytest
+import uuid
+from datetime import datetime
+
+from worker.search_runner import run_search
+from models.metadata import Metadata, Summary, MediaType
+
+# Dummy Metadata factory
+def make_dummy_metadata(source: str, title: str = "Dummy Title") -> Metadata:
+    return Metadata(
+        title=title,
+        altTitle=[f"{title} alt"],
+        cover="http://example.com/cover.jpg",
+        banner=None,
+        type=MediaType.MOVIE,  # bruk en gyldig enum fra din kode
+        summary=[Summary(summary="A fake summary", language="en")],
+        genres=["Drama", "Action"],
+        source=source,
+    )
+
+# Dummy Source that mimics SourceBase
+class DummySource:
+    def __init__(self, titles, result=None, raise_exc=False):
+        self.titles = titles
+        self._result = result
+        self._raise_exc = raise_exc
+
+    async def search(self):
+        if self._raise_exc:
+            raise RuntimeError("Search failed")
+        return self._result
+
+@pytest.mark.asyncio
+async def test_run_search_all_results(monkeypatch):
+    sources = [
+        DummySource(["foo"], make_dummy_metadata("mal")),
+        DummySource(["foo"], make_dummy_metadata("imdb")),
+        DummySource(["foo"], make_dummy_metadata("anii")),
+    ]
+    monkeypatch.setattr("worker.search_runner.get_all_sources", lambda titles: sources)
+
+    results = await run_search(["foo"])
+    assert len(results) == 3
+    assert all(isinstance(r, Metadata) for r in results)
+    assert {r.source for r in results} == {"mal", "imdb", "anii"}
+
+@pytest.mark.asyncio
+async def test_run_search_filters_none(monkeypatch):
+    sources = [
+        DummySource(["foo"], make_dummy_metadata("mal")),
+        DummySource(["foo"], None),
+        DummySource(["foo"], make_dummy_metadata("imdb")),
+    ]
+    monkeypatch.setattr("worker.search_runner.get_all_sources", lambda titles: sources)
+
+    results = await run_search(["foo"])
+    assert len(results) == 2
+    assert {r.source for r in results} == {"mal", "imdb"}
+
+@pytest.mark.asyncio
+async def test_run_search_handles_exception(monkeypatch):
+    sources = [
+        DummySource(["foo"], make_dummy_metadata("mal")),
+        DummySource(["foo"], raise_exc=True),
+        DummySource(["foo"], make_dummy_metadata("imdb")),
+    ]
+    monkeypatch.setattr("worker.search_runner.get_all_sources", lambda titles: sources)
+
+    results = await run_search(["foo"])
+
+    # Nå skal vi få bare de gyldige Metadata-resultatene
+    assert all(isinstance(r, Metadata) for r in results)
+    assert {r.source for r in results} == {"mal", "imdb"}
+
diff --git a/apps/pyMetadata/utils/__init__.py b/apps/pyMetadata/utils/__init__.py
new file mode 100644
index 00000000..e69de29b
diff --git a/apps/pyMetadata/utils/backoff.py b/apps/pyMetadata/utils/backoff.py
new file mode 100644
index 00000000..61ec8d44
--- /dev/null
+++ b/apps/pyMetadata/utils/backoff.py
@@ -0,0 +1,11 @@
+from utils.logger import logger
+import time
+
+def retry_delays():
+    return [5, 15, 30, 60]
+
+def wait_with_backoff():
+    for delay in retry_delays():
+        logger.info(f"⏳ Venter {delay} sekunder...")
+        time.sleep(delay)
+        yield
diff --git a/apps/pyMetadata/utils/logger.py b/apps/pyMetadata/utils/logger.py
new file mode 100644
index 00000000..561e0dd6
--- /dev/null
+++ b/apps/pyMetadata/utils/logger.py
@@ -0,0 +1,32 @@
+import logging
+import sys
+
+# ANSI farger
+COLORS = {
+    "INFO": "\033[94m",    # blå
+    "DEBUG": "\033[92m",   # grønn
+    "WARNING": "\033[93m", # gul
+    "ERROR": "\033[91m",   # rød
+    "RESET": "\033[0m"
+}
+
+class ColoredFormatter(logging.Formatter):
+    def format(self, record):
+        levelname = record.levelname
+        color = COLORS.get(levelname, COLORS["RESET"])
+        prefix = f"[{levelname}]"
+        message = super().format(record)
+        return f"{color}{prefix}{COLORS['RESET']} {message}"
+
+def setup_logger(level=logging.INFO):
+    handler = logging.StreamHandler(sys.stdout)
+    formatter = ColoredFormatter("%(asctime)s - %(name)s - %(message)s")
+    handler.setFormatter(formatter)
+
+    logger = logging.getLogger()
+    logger.setLevel(level)
+    logger.handlers = [handler]
+    return logger
+
+# Opprett global logger
+logger: logging.Logger = setup_logger()
diff --git a/apps/pyMetadata/worker/__init__.py b/apps/pyMetadata/worker/__init__.py
new file mode 100644
index 00000000..e69de29b
diff --git a/apps/pyMetadata/worker/poller.py b/apps/pyMetadata/worker/poller.py
new file mode 100644
index 00000000..c743eea0
--- /dev/null
+++ b/apps/pyMetadata/worker/poller.py
@@ -0,0 +1,59 @@
+import asyncio
+import time
+from typing import Optional
+import uuid
+from db.database import Database
+from db.repository import claim_task, fetch_next_task, mark_failed, persist_event_and_mark_consumed
+from models.event import MetadataSearchResultEvent
+from worker.processor import process_task
+from utils.logger import logger
+from config.database_config import DatabaseConfig
+from models.task import MetadataSearchTask, Task
+
+def run_iteration(db: Database, worker_id: str, poll_interval: int) -> tuple[int, int]:
+    """
+    Kjør én iterasjon av poller-loopen.
+    Returnerer (sleep_interval, next_interval).
+    """
+    try:
+        task: Optional[Task] = fetch_next_task(db)
+        if task:
+            if not isinstance(task, MetadataSearchTask):
+                logger.warning(f"⚠️ Ukjent task-type {type(task)} for {task.taskId}, hopper over.")
+                return poll_interval, poll_interval
+
+            if not claim_task(db, str(task.taskId), worker_id):
+                logger.info(f"⏩ Task {task.taskId} ble claimet av en annen worker.")
+                return poll_interval, poll_interval
+
+            logger.info(f"🔔 Fant task {task.taskId} ({task.task}), claimed by {worker_id}")
+            try:
+                event: MetadataSearchResultEvent = asyncio.run(process_task(task))
+                if event:
+                    persist_event_and_mark_consumed(db, event, str(task.taskId))
+                    logger.info(f"✅ Task {task.taskId} ferdig prosessert")
+                else:
+                    logger.error(f"❌ Task returned nothing! {task.taskId}")
+                    raise RuntimeError("process_task returned nothing!")
+            except Exception as task_error:
+                logger.error(f"❌ Task {task.taskId} feilet under prosessering: {task_error}")
+                mark_failed(db, str(task.taskId))
+            return poll_interval, 5  # sov med nåværende, reset til 5
+        else:
+            logger.debug("Ingen nye tasks.")
+            return poll_interval, min(poll_interval * 2, 60)
+    except Exception as e:
+        logger.error(f"⚠️ Feil i worker: {e}")
+        db.connect()
+        return poll_interval, 5
+
+def run_worker(db: Database, shutdown_flag_ref=lambda: False) -> None:
+    poll_interval: int = 5
+    worker_id = f"worker-{uuid.uuid4()}"
+
+    while not shutdown_flag_ref():
+        sleep_interval, poll_interval = run_iteration(db, worker_id, poll_interval)
+        time.sleep(sleep_interval)
+
+    logger.info("👋 run_worker loop avsluttet")
+    db.close()
\ No newline at end of file
diff --git a/apps/pyMetadata/worker/processor.py b/apps/pyMetadata/worker/processor.py
new file mode 100644
index 00000000..330379f0
--- /dev/null
+++ b/apps/pyMetadata/worker/processor.py
@@ -0,0 +1,128 @@
+import uuid
+from datetime import datetime
+
+from tabulate import tabulate
+from models.metadata import Metadata
+from models.task import MetadataSearchTask
+from utils.logger import logger
+from models.event import (
+    EventMetadata,
+    MetadataSearchResultEvent,
+    SearchResult,
+    MetadataResult,
+    Summary,
+    TaskStatus,
+    MediaType,
+)
+from worker.search_runner import run_search
+from algo.SimpleMatcher import SimpleMatcher
+from algo.PrefixMatcher import PrefixMatcher
+from algo.AdvancedMatcher import AdvancedMatcher
+from db.repository import mark_failed
+
+def source_priority(source: str) -> int:
+    """Domene-spesifikk kildevekting."""
+    priority_map = {'mal': 1, 'anii': 2, 'imdb': 3}
+    return priority_map.get(source, 4)
+
+
+def score_metadata_against_title(title, m: Metadata) -> SearchResult:
+    simple = SimpleMatcher(title, m).getScore()
+    prefix = PrefixMatcher(title, m).getScore()
+    advanced = AdvancedMatcher(title, m).getScore()
+
+    # IMPORTANT: map exactly to bannerImage, not banner.
+    metadata_result = MetadataResult(
+        source=m.source,
+        title=m.title,
+        alternateTitles=m.altTitle if m.altTitle else [],
+        cover=getattr(m, "cover", None),
+        bannerImage=getattr(m, "bannerImage", None),  # no renaming
+        type=m.type,  # must already be MediaType
+        summary=[Summary(language=s.language, description=s.summary) for s in m.summary],
+        genres=m.genres,
+    )
+
+    return SearchResult(
+        simpleScore=simple,
+        prefixScore=prefix,
+        advancedScore=advanced,
+        sourceWeight=1.0,
+        metadata=metadata_result
+    )
+
+
+def print_summary(results: list[SearchResult], titles: list[str]) -> None:
+    """Print tabell med scorer for alle kombinasjoner."""
+    rows = []
+    for r in results:
+        rows.append((
+            # NB: metadata.title er matched title, search_title kan du lagre i SearchResult hvis du vil
+            r.metadata.title,
+            r.metadata.source,
+            r.simpleScore,
+            r.prefixScore,
+            r.advancedScore
+        ))
+    headers = ["Matched Title", "Source", "Simple", "Prefix", "Advanced"]
+    print(tabulate(rows, headers=headers))
+
+
+def choose_recommended(results: list[SearchResult]) -> SearchResult:
+    """Velg recommended basert på scorer og kildevekting."""
+    return max(
+        results,
+        key=lambda r: (
+            r.advancedScore,
+            r.simpleScore,
+            r.prefixScore,
+            -source_priority(r.metadata.source)
+        )
+    )
+
+
+async def process_task(db, task: MetadataSearchTask) -> MetadataSearchResultEvent|None:
+    titles = task.data.searchTitles
+    logger.info(f"Prosesserer task {task.taskId} med titler: {titles}")
+
+    try:
+        metadata_list = await run_search(titles)
+        if not metadata_list:
+            mark_failed(db, task.taskId)
+            return
+
+        # 1) Score alle kombinasjoner
+        results = []
+        for m in metadata_list:
+            for t in titles:
+                results.append(score_metadata_against_title(t, m))
+
+        # 2) Print tabell
+        print_summary(results, titles)
+
+        # 3) Velg recommended
+        recommended = choose_recommended(results)
+
+        # 4) Bygg event
+        core_metadata = EventMetadata(
+            created=datetime.now(),
+            derivedFromId={task.referenceId, task.taskId}
+        )
+
+        event = MetadataSearchResultEvent(
+            referenceId=task.referenceId,
+            eventId=uuid.uuid4(),
+            metadata=core_metadata,
+            results=results,
+            recommended=recommended,
+            status=TaskStatus.COMPLETED
+        )
+
+        # 5) Returner
+        logger.info(f"✅ Task {task.taskId} ferdig prosessert med {len(results)} resultater")
+        return event
+
+    except Exception as e:
+        logger.error(f"❌ Task {task.taskId} feilet: {e}")
+        mark_failed(db, task.taskId)
+        return None
\ No newline at end of file
diff --git a/apps/pyMetadata/worker/search_runner.py b/apps/pyMetadata/worker/search_runner.py
new file mode 100644
index 00000000..f39e4752
--- /dev/null
+++ b/apps/pyMetadata/worker/search_runner.py
@@ -0,0 +1,31 @@
+# search_runner.py
+import asyncio
+from typing import List
+from models.metadata import Metadata
+from utils.logger import logger
+from sources.factory import get_all_sources
+
+async def run_search(titles: List[str]) -> List[Metadata]:
+    """
+    Kjører alle kilder parallelt på gitt titler.
+    Returnerer en liste av Metadata fra alle kilder.
+    Ingen mapping eller scoring gjøres her.
+    """
+
+    sources = get_all_sources(titles)
+
+    # Kjør alle kildesøk parallelt
+    results = await asyncio.gather(*(s.search() for s in sources), return_exceptions=True)
+
+    metadata_results: List[Metadata] = []
+    for source, r in zip(sources, results):
+        if isinstance(r, Exception):
+            logger.warning(
+                f"Kilde '{source.__class__.__name__}' feilet under søk "
+                f"med titler={source.titles}: {r}"
+            )
+        elif r is not None:
+            metadata_results.append(r)
+
+    logger.info(f"Søk ferdig: {len(metadata_results)} resultater fra {len(sources)} kilder")
+    return metadata_results
diff --git a/apps/pyWatcher/api/health_api.py b/apps/pyWatcher/api/health_api.py
new file mode 100644
index 00000000..8e70b572
--- /dev/null
+++ b/apps/pyWatcher/api/health_api.py
@@ -0,0 +1,19 @@
+from fastapi import FastAPI
+from fastapi.responses import JSONResponse
+
+def create_health_app(observers_ref):
+    """
+    Returnerer en FastAPI-app med /health endpoint.
+    observers_ref: en funksjon eller lambda som gir listen av observers.
+    """
+    app = FastAPI()
+
+    @app.get("/health")
+    def health():
+        observers = observers_ref()
+        healthy = all(obs.is_alive() for obs in observers)
+        status = "healthy" if healthy else "unhealthy"
+        code = 200 if healthy else 500
+        return JSONResponse({"status": status}, status_code=code)
+
+    return app