yonko-sid/indelible

Purpose and Entry Points

INDELIBLE is a cross-platform digital-rights forensics system: it watermark-protects creator media, verifies suspected copies, monitors for piracy, and drafts enforcement material. The user-facing client starts at lib/main.dart and composes the Flutter application in lib/src/app.dart. The service boundary is the FastAPI application in backend/main.py, deployed as a Dockerized Python service via backend/Dockerfile.

Client Application

Flutter targets Android, iOS, desktop, and web. Authentication and API access are isolated behind lib/src/services/auth_service.dart and lib/src/services/api_service.dart; Firebase configuration is generated in lib/firebase_options.dart. App screens expose the core workflows—protect, verify, archive/activity, dashboard, and settings—through lib/src/screens/protect_screen.dart, lib/src/screens/verify_screen.dart, and lib/src/screens/dashboard_screen.dart.

HTTP Workflow Boundary

backend/main.py owns multipart upload handling, user-scoped alert retrieval, suspect-URL scanning, and protected-file download endpoints. The Flutter client submits files and form data to this API rather than performing DSP locally. Processed media is retained and served from backend/outputs; metadata/registries are file-backed, including backend/creator_registry.json and backend/blockchain_registry.json.

Protection and Proof Pipeline

Protection orchestrates signed payload construction, robust watermark embedding, creator registration, and chain-of-custody recording. backend/core/payload.py creates creator-fingerprint/timestamp/HMAC payloads with Reed-Solomon recovery encoding. backend/core/watermark.py applies/extracts the image watermark through Y-channel Haar DWT, LL-band redundant QIM, and robust verification search. backend/core/core_engine.py is the likely coordination layer between media processing, registries, and verification results.

Media and Video Processing

The image watermark engine is the central forensic primitive in backend/core/watermark.py. Video support is separated in backend/core/video_processor.py, using FFmpeg to extract and reassemble frames, keeping video transport/encoding concerns outside the DSP algorithm. Output artifacts and .meta sidecars represent the protected-media delivery boundary under backend/outputs.

Detection, Monitoring, and Enforcement

Known protected works are indexed with perceptual hashes in a BK-tree (backend/core/bktree_index.py) to cheaply filter candidate matches before cryptographic extraction. The background watchdog in backend/core/monitoring_daemon.py invokes web acquisition in backend/core/scraper.py, then applies similarity and watermark proof stages. backend/core/ai_engine.py integrates Google Gemini 2.5 Flash for piracy analysis and DMCA draft generation; this requires a Gemini API key external to source control.

Identity, Storage, and Infrastructure

Firebase Auth supplies client identity; Firebase storage integration is represented by backend/storage/firebase.py, while backend/storage/file_manager.py handles local file operations. Creator fingerprints deterministically map Firebase UIDs to registry records. Blockchain anchoring is abstracted in backend/core/blockchain.py and persisted in backend/blockchain_registry.json; README identifies Polygon as the intended immutable anchor. Runtime dependencies include Python 3.10+, FastAPI/Uvicorn, OpenCV/PyWavelets/NumPy, FFmpeg, and Docker/Railway-compatible deployment.