Supported File Types

Footnotes

1. Poly creates sharp thumbnails and visual-search representations for supported images, including .png, .tif, .tiff, .bmp, .gif, .webp, .heic, and .heif. Animated formats use a representative still for thumbnails and visual search. Text visible inside an ordinary image is not indexed as document text. ↩
2. Poly previews, visually searches, and reads available camera metadata for supported RAW photos: .cr2, .cr3, .arw, .arq, .nef, .nrw, .raf, .dng, .orf, .ori, .rw2, .3fr, .pef, and .x3f. Use a dedicated RAW editor for sensor-level adjustments. ↩
3. Poly recognizes .jp2, .jxr, .wdp, and .ora, but currently keeps these formats searchable by name rather than offering a rich preview or content-aware search. The registered .jpg2 alias is also stored and searchable by name, but Poly does not infer JPEG 2000 from that suffix alone. ↩
4. JPEG XL, QOI, Photoshop PSD, Windows icon, and SVG receive format-appropriate previews when supported. .svgz remains searchable by name because Poly does not currently decompress it for SVG processing. .dib receives BMP image features when identified as image/bmp, while .hif receives HEIF or HEIC image features when supplied with an accurate media type. Poly does not currently infer these formats from the suffixes alone. Complex, layered, animated, or high-dynamic-range content may be represented as a flattened or static image. ↩
5. Poly supports content-aware previews, search, and agent reading for .pdf, .docx, .doc, .docm, .odt, .pages, .xlsx, .xls, .ods, .numbers, .odp, .pptx, .ppt, and .key. A .xlt Excel template, .pps PowerPoint Show, .dot Word template, or .pot PowerPoint template receives these document features only when correctly classified as application/vnd.ms-excel, application/vnd.ms-powerpoint, application/msword, or application/vnd.ms-powerpoint, respectively; these suffixes are not inferred by themselves. Connected previews are best effort, and macros or external data connections do not run. ↩
6. Poly gives supported ebooks and comics a paginated preview and content-aware search, including .epub, .mobi, .azw3, .fb2, .fbz, .cbz, .cbr, and .cb7. The ambiguous .prc suffix is stored and searchable by name; a verified Mobipocket book receives document features only when supplied with the accurate MIME type. Encrypted, DRM-protected, damaged, or unsupported files may have limited processing. ↩
7. Rich audio support includes .mp3, .wav, .ogg, .opus, .m4a, .flac, .aac, and .aiff. .aif receives these features when identified as AIFF from its contents or media type; Poly does not currently infer AIFF from the .aif suffix alone. Files labeled audio/mp4 are supported, while an ordinary .mp4 filename is normally treated as video. ↩
8. Poly recognizes .amr, .dsf, .ape, and .midi, but these formats currently have limited playback, transcription, and content-aware search. The .mid suffix is stored and searchable by name, but Poly does not currently infer MIDI from it. The legacy .mac suffix can represent Monkey's Audio, a MacPaint image, or a macro, so Poly stores it and makes it searchable by name without inferring audio from the suffix. ↩
9. .mp1 and .mp2 files receive rich audio features when supplied as audio/mpeg. .oga and .spx receive them when supplied as audio/ogg. .adts, .loas, and .ass files receive them when supplied as audio/aac; .aifc receives them when supplied as audio/x-aiff. Poly does not currently infer those MIME types from the suffixes alone. A text-classified .ass subtitle instead receives text preview and indexing without styled video rendering. ↩
10. Poly supports common video files such as .mp4, .webm, .mov, .mkv, .avi, .flv, .m4v, .mpg, .mpeg, and .wmv. Playback depends on supported media, and transcripts require usable audio. Search by file can match visual content as well as spoken content. ↩
11. Directories use inode/directory and have no filename extension or magic number. ↩
12. Bookmarks use text/uri-list and accept .webloc and .url files for desktop interchange. Each Poly bookmark has one destination; preview and content search depend on access to the linked page. Standalone .uri and .uris files are stored and searchable by name, but Poly does not infer bookmarks or split a multi-entry list from either suffix. Google Docs URLs receive document previews, content-aware search, and agent reading when the connected Google account can access them. Symbolic links use inode/symlink, have no filename extension, and currently remain outside Poly's preview and content-indexing support. ↩
13. YouTube video links use application/vnd.youtube.video-url. Poly previews the destination, makes available video details searchable, and lets the agent use accessible videos. Playback and reading depend on YouTube availability and access. ↩
14. Rich text support includes plain text .txt, Python .py, JavaScript .js, ECMAScript module .mjs, JSX .jsx, TypeScript with JSX .tsx, Vue SFC .vue, Rust source .rs, R source .r, Ruby source .rb, Haskell .hs, Scala .scala, Swift .swift, Kotlin .kt, Julia .jl, Emacs Lisp .el, Elixir or Euphoria .ex, Perl or Prolog .pl, SQL .sql, CSS .css, shell scripts .sh, CSV .csv, YAML .yml, TSV .tsv, XML .xml, WebVTT .vtt, HTML .html, HTML .htm when supplied as text/html, and iCalendar .ics. Poly previews and indexes readable content without executing source code, loading imports, or calculating spreadsheet-style formulas. .wsgi receives Python source features when supplied as text/x-python; Poly does not currently infer that type from the suffix alone. Connected previews are best effort. PHP .php files are recognized but currently remain searchable by name. ↩
15. Poly lets you inspect recognized .vcf and .vcard source without importing contacts. Contact fields are not currently content-indexed or available to the agent. ↩
16. Poly recognizes binary and structured payload types including LLVM .bc, WebAssembly .wasm, Java .class, MessagePack, URL-encoded forms, multipart forms, generic binary data, Mach-O .mach, ELF .elf, and object .obj. These remain searchable by name but are not executed, decoded, or content-indexed. Poly currently classifies ELF filenames with the generic application/octet-stream fallback. ↩
17. .polychat opens a Poly conversation. Conversation text is searchable; saved searches execute their stored query when opened. ↩
18. Poly recognizes these archive and compression formats but does not decompress or otherwise process their contents: .7z, .ar, .bz2, .cab, .cpio, .crx, .deb, .gz, .lz, .msi, .rar, .rpm, .tar, .taZ, .xz, .zip, .Z, and .zst. The .a static-library convention can be stored and found by name, but Poly does not currently infer the ar type from that suffix. The .tbz, .tbz2, .tgz, and .udeb suffixes are stored and searchable by name without archive-type inference. These files remain stored, synced, shared, versioned, and searchable by name. ↩
19. Poly stores opaque formats such as DICOM .dcm, SQLite .sqlite, Flash .swf, NES .nes, PostScript .ps, EPS .eps, and Portable Executables .exe and .dll. Their contents are not previewed or content-indexed. The .db suffix is ambiguous and is not inferred as SQLite without a supplied MIME type. Adobe Illustrator .ai artwork is likewise stored and searchable by name, although Poly does not infer a PostScript or Illustrator format from the suffix. Fonts including .ttf, .otf, .woff, and .woff2 remain manageable and searchable by name. ↩