Home / Universal File Inspector

Universal Zero-Upload File Inspector

Verify true binary signatures, diagnose spoofed downloads, inspect 70GB+ LLM tensor headers & Parquet footers, and slice arbitrary hex offsets—100% locally inside a sandboxed Web Worker. Turn off Wi-Fi to verify offline execution.

0xFF

Drop any local file (1 KB to 100 GB+) to inspect binary headers & schema

Reads tiny File.slice() header/footer pages in a sandboxed Web Worker · 0 Bytes Uploaded

How Two-Stage Adaptive Slicing Works

Standard browser tools call file.arrayBuffer(), crashing tabs on files larger than 2 GB. Instead, our PagedBlobReader reads only 256 KB virtual pages via Blob.slice(offset, end):

• SafeTensors: Reads 8-byte u64 length N, then slices exact JSON header (up to 32 MB).
• GGUF v2/v3: Streams KV pairs while fast-skipping 150K-token BPE vocab arrays to reach tensor_infos.
• Parquet: Reads trailing 8 bytes (footer_len + PAR1) and decodes Thrift Compact metadata.

Deep Container Disambiguation

Checking the first 4 bytes at 0x00 is not enough for modern container families. We inspect internal directory records without decompressing payloads:

• PK\x03\x04 (ZIP): Distinguishes .xlsx, .ipa, .aab, .npz, .usdz & PyTorch .pt.
• ftyp (ISOBMFF): Parses major/compatible 4CC brands for .heic vs .avif.
• RIFF: Verifies form FourCC at 0x08 (WEBP, WAVE, AVI ).