IceWhaleTech/filens

Before Instructions only → After Instructions only

2f130f2ab623 → 66e01abbd368

- 20 Check `capacity.operations` before using source additions such as `search` or `similar`. Check `capacity.formats`, `capacity.conversionPairs`, and `capacity.ocr.available` for document operations; reading support does not imply conversion support, and an advertised OCR route does not prove that OCR is installed. Reading, directory scans and conversion are local; initial npm installation and explicit OCR setup may download dependencies.
+ 20 Check `capacity.operations` before using operations such as `search` or `similar`. Check `capacity.formats`, `capacity.conversionPairs`, and `capacity.ocr.available` for document operations; reading support does not imply conversion support, and an advertised OCR route does not prove that OCR is installed. Reading, directory scans and conversion are local; initial npm installation and explicit OCR setup may download dependencies.
- 33 {"operation":"understand","input":"/path/report.pdf","format":"json","layout":false,"maxBytes":8388608}
+ 33 {"op":"understand","input":"/path/report.pdf","format":"json","layout":false,"maxBytes":8388608}
- 65 Check `capacity.operations` for `summarize`; older published builds do not provide it. Use `filens summarize report.pdf --query 'budget risk' --summary-chars 1000 --json`, or pipe `{"operation":"summarize","input":"report.pdf","query":"budget risk","summaryChars":1000,"summaryRatio":0.3}`. `prune` is an alias. Query is optional. Defaults retain at most 30% and 2000 Unicode characters, including separators; this is a budget, not a promise to retain all important facts.
+ 65 Check `capacity.operations` for `summarize`; older published builds do not provide it. Use `filens summarize report.pdf --query 'budget risk' --summary-chars 1000 --json`, or pipe `{"op":"summarize","input":"report.pdf","query":"budget risk","summaryChars":1000,"summaryRatio":0.3}`. `prune` is an alias. Query is optional. Defaults retain at most 30% and 2000 Unicode characters, including separators; this is a budget, not a promise to retain all important facts.
- 71 Check `capacity.operations` for `search`. Use `filens search '/path/documents' --query 'budget risk' --json`, or pipe `{"operation":"search","input":"/path/documents","query":"预算 风险","format":"json"}`. Chinese and English quoted directory/query intents are supported. This is lexical BM25 with exact matching boosts, not embedding-based semantic retrieval. `--search-mode exact` matches case-sensitive literal text; `files` searches filenames without reading contents.
+ 71 Check `capacity.operations` for `search`. Use `filens search '/path/documents' --query 'budget risk' --json`, or pipe `{"op":"search","input":"/path/documents","query":"预算 风险","format":"json"}`. Chinese and English quoted directory/query intents are supported. This is lexical BM25 with exact matching boosts, not embedding-based semantic retrieval. `--search-mode exact` matches case-sensitive literal text; `files` searches filenames without reading contents.
- 81 Check `capacity.operations` for `similar`; published npm `0.2.0` lacks it. Source builds include it in Lite without OCR or model setup.
+ 81 Check `capacity.operations` for `similar`; older builds may lack it. Current Full and Lite bundles include it without OCR or model setup.
- 88 The aliases return `operation: "similar"`. JSON requests use `{"operation":"similar","input":"/path/photos","similarityMode":"images","threshold":8}`; mode is `images` (default) or `duplicates`. Full SHA-256 identifies duplicate files after size/prefix filtering. Perceptual image hashes identify possible variants, including different dimensions and compression, for JPEG, PNG, WebP, GIF, BMP and TIFF. This does not compare near-duplicate document text or retrieve different photos by semantic subject.
+ 88 The aliases return `op: "similar"`. JSON requests use `{"op":"similar","input":"/path/photos","similarityMode":"images","threshold":8}`; mode is `images` (default) or `duplicates`. Full SHA-256 identifies duplicate files after size/prefix filtering. Perceptual image hashes identify possible variants, including different dimensions and compression, for JPEG, PNG, WebP, GIF, BMP and TIFF. This does not compare near-duplicate document text or retrieve different photos by semantic subject.
- 111 {"operation":"convert","input":"/path/report.doc","output":"/path/report.docx","format":"json","options":{"locale":"zh-CN"}}
+ 111 {"op":"convert","input":"/path/report.doc","output":"/path/report.docx","format":"json","options":{"locale":"zh-CN"}}

Compare on GitHub

Before Not checked yet → After Instructions only

First observation → 2f130f2ab623

No earlier version was observed for comparison.

Compare on GitHub