services/detection_service.py

"""
Detection service — runs detection-only jobs through the pipeline +
orchestrator + normalization + confidence, returns a UnifiedFaceReport.
"""

from __future__ import annotations

import time
from typing import Optional

from confidence.engine import ConfidenceEngine
from confidence.conflicts import ConflictDetector
from metrics.collector import MetricsCollector
from models.jobs import JobRequest
from models.providers import ProviderCapability
from normalization.merger import ReportMerger
from orchestrator.runner import Orchestrator
from pipeline import (
    InputValidator,
    ImagePreprocessor,
    ImageHasher,
    FeatureExtractor,
)
from providers.registry import ProviderRegistry
from storage.cache import Cache
from utils.logging import execution_context, new_execution_id


class DetectionService:
    """Handles detection-only jobs."""

    def __init__(
        self,
        registry: ProviderRegistry,
        orchestrator: Orchestrator,
        cache: Cache,
        metrics: MetricsCollector,
        validator: InputValidator,
        preprocessor: ImagePreprocessor,
        hasher: ImageHasher,
        feature_extractor: FeatureExtractor,
        confidence_engine: ConfidenceEngine,
        conflict_detector: ConflictDetector,
    ) -> None:
        self._registry = registry
        self._orchestrator = orchestrator
        self._cache = cache
        self._metrics = metrics
        self._validator = validator
        self._preprocessor = preprocessor
        self._hasher = hasher
        self._feature_extractor = feature_extractor
        self._merger = ReportMerger(confidence_engine, conflict_detector)

    async def detect(self, request: JobRequest) -> dict:
        eid = new_execution_id()
        with execution_context(execution_id=eid, provider_id="detection_service"):
            t0 = time.perf_counter()

            # 1. Validate
            vr = self._validator.validate(
                image_url=request.image_url,
                image_base64=request.image_base64,
            )
            if not vr.valid:
                return {"success": False, "error": vr.error, "error_type": "ValidationError"}

            # 2. Preprocess
            if vr.source == "url":
                pre = self._preprocessor.from_url(request.image_url)
            else:
                pre = self._preprocessor.from_bytes(vr.image_bytes, vr.source)

            # 3. Hash
            img_hash = self._hasher.hash(pre.image)

            # 4. Feature extraction (uses default detector)
            pipeline_output = self._feature_extractor.extract(
                pre.image, img_hash, pre.width, pre.height, pre.source,
                original_bytes=pre.original_bytes,
                original_format=pre.original_format,
            )

            # 5. Orchestrate detection providers
            results = await self._orchestrator.run(
                pipeline_output=pipeline_output,
                capabilities=[ProviderCapability.DETECTION],
                provider_whitelist=request.providers or None,
                execution_id=eid,
            )

            # 6. Merge + confidence
            elapsed = (time.perf_counter() - t0) * 1000.0
            report = self._merger.merge(
                results=results,
                image_hash=img_hash,
                job_id=eid,
                total_elapsed_ms=elapsed,
                kind="detection",
            )

            self._metrics.timings.record("job.detection", elapsed)
            self._metrics.counters.inc("jobs.detection.completed")

            return {
                "success": True,
                "report": report.model_dump(),
                "elapsed_ms": round(elapsed, 3),
            }