DuoBolt is a storage cleanup tool for macOS and Windows that identifies duplicate files through full byte comparison and groups similar photographs through content analysis. It runs its photo ranking models locally, so libraries are never uploaded and no account is needed. Cleanup operations are gated behind a review step, and each action is recorded in history for later restoration.
DuoBolt targets photographers, developers, and home server operators who need to reduce storage consumption without risking data loss. It reads RAW files directly and scans NAS volumes, network drives, and the macOS Photos library. A per-operation cap applies to the free tier, while a one-time licence removes it for duplicates, similar photos, and AI Curation.
Exact matching and similarity ranking are handled as separate jobs with separate results, so proven copies are never mixed with visual look-alikes. Every pick made by the ranking system carries an explanation naming the measurements behind it. Nothing is deleted automatically, and every operation can be traced or reversed through the built-in history.
| Cryptographic content hashing eliminating false duplicate matches | Verifies identical files through byte-level comparison |
| Visual similarity clustering of photographs by content | Groups bursts and near-identical frames into adjustable bands |
| AI-driven selection of optimal shot within image groups | Ranks variants using face quality and exposure metrics |
| Dual-pane inspection interface with granular filtering | Displays grouped results alongside associated file entries |
| Multiple disposal strategies including reversible operations | Offers trash, archive, copy or permanent deletion modes |
| Auditable operation log enabling later restoration | Records scan history and touched file paths |
| Selection exclusion safeguards for critical directories | Prevents accidental removal of protected folder contents |
| Configurable inclusion criteria governing scan boundaries | Defines folder scope, size floors and extension filters |
| Symbolic link resolution avoiding phantom duplication | Collapses links to prevent false copy detection |
| Copy-on-write block sharing recognition on macOS volumes | Reports actual reclaimable space for APFS clones |
| Broad RAW and consumer image format decoding support | Reads proprietary camera formats and common web images |
| Direct library targeting bypassing export workflows | Scans native photo databases before asset duplication |
| On-device inference without remote data transmission | Runs bundled models locally excluding telemetry collection |
| Tiered licensing model with operation count restrictions | Limits free cleanup batches while scans remain unrestricted |
| Concurrent hashing pipeline overlapping I/O with computation | Utilizes multiple cores for throughput on large datasets |
| Metadata-independent matching immune to filename manipulation | Ignores timestamps and sizes when declaring equivalence |
| Uncertainty routing mechanism deferring ambiguous decisions | Escalates questionable groups for manual operator review |
| Path context visibility revealing file storage locations | Shows folder origins to expose damaged copy patterns |
| Traceability supporting forensic and archival requirements | Maintains reversibility in environments demanding accountability |
| Streaming chunked reads accommodating network storage latency | Overlaps disk access with hashing stages efficiently |