detect() result is now a required transform argument (0.4.0)

README.md

@@ -42,8 +42,8 @@ brew install tesseract
 
 **`detect()` is the core primitive** — it finds every PII entity with its
 location, category, and confidence. The transforms (redact / anonymize /
-de-identify) consume its output: run detection once, feed the same result
-into whichever transform you need.
+de-identify) take its result as a required argument: detection always runs
+first, once, and every transform consumes its output.
 
 ### Image — text + visual PII in document images
 
@@ -61,9 +61,9 @@ for e in entities:
     print(e.kind, e.category, e.bbox, e.text, e.score)
 
 # One detection, three possible outputs:
-redactor.redact("page.jpg", entities=entities).save("redacted.png")
-redactor.anonymize("page.jpg", entities=entities).save("anonymized.png")
-deid = redactor.deidentify("page.jpg", entities=entities)
+redactor.redact("page.jpg", entities).save("redacted.png")
+redactor.anonymize("page.jpg", entities).save("anonymized.png")
+deid = redactor.deidentify("page.jpg", entities)
 deid.image.save("deidentified.png"); deid.mapping.to_dict()
 ```
 
@@ -79,9 +79,9 @@ spans = r.detect(text)                   # char-offset spans — the core call
 for s in spans:
     print(s.category, s.start, s.end, s.text, s.score)
 
-r.redact(text, mask="[REDACTED]", spans=spans)
-r.anonymize(text, spans=spans)
-r.deidentify(text, spans=spans)          # .text + .mapping
+r.redact(text, spans, mask="[REDACTED]")
+r.anonymize(text, spans)
+r.deidentify(text, spans)                # .text + .mapping
 ```
 
 ### De-identify and anonymize (v0.2.0+, both modalities)