feat: update classifier service and models

fix: harden ownership, model loading, and provenance handling

docs: add IMAGE_API_INTEGRATION.md

fix: treat missing C2PA claims as absent provenance

fix: support deployed c2patool JSON argument format

chore: update uv.lock

fix: escalate explicit generated filename with forensic risk

fix: prevent explicit AI filename signals from approving reports

feat: add browser test console for image verification

fix: remove deprecated email aliases from image verification API

chore: pin image verification dependencies to CPU

feat: track image reports by user email

fix: align verification response field order

fix: harden optional OCR and forensic edge cases

docs: update README and deployment files for multi-signal image verification

feat: update image verification response schema and route

feat: add dynamic multi-signal trust scoring

feat: add Grad-CAM explainability output

feat: integrate PyTorch classifier as one verification signal

feat: add lightweight image forensic analysis

feat: add watermark template matching

feat: add visible watermark OCR detection

feat: add optional C2PA provenance analyzer

feat: add image metadata and AI tool marker detection

feat: add file validation and filename analysis

chore: inspect and prepare image verification architecture

Use newer Debian base for c2patool

Install c2patool in Space image

chore: track .keras files with Git LFS for Hugging Face

feat: add Docker support, update dependencies and services

feat: expose image verification API

feat: score trust and build reports

feat: generate explainability overlays

feat: add lightweight forensic analysis

feat: add replaceable image detector

feat: add provenance and watermark checks

feat: extract image metadata signals

feat: define verification report schema

feat: add upload and image utilities

chore: add application package skeleton

chore: initialize backend project config