Phases 2-6: detection, visual timeline, backend hand-off, voiceprints

Phase 2 (call detection): CallDetector using CoreAudio per-process mic attribution (anarlog technique) — robust start+stop for Zoom/Teams/Signal/Meet, ignoring our own recording; auto-record toggle. Built; pending live multi-app confirmation by the user. Phase 3 (visual timeline foundation): AppAdapter protocol + SpeakerObservation, TimelineBuilder (hysteresis/overlap/self-merge/aliases), VisualTimeline (schema 1.1), TextRecognizer (Vision OCR), FrameSampler + GridCallAnalyzer (name OCR + saturated-highlight active-speaker attribution), SignalAdapter, VisualObserver (window capture; frames released, never saved; minimized->visual_gap, idle != gap). Synthetic-frame tested; adapter geometry pending real Signal fixtures + live VisualObserver validation. Phase 5 (backend hand-off): SparkControlClient (multipart label-merge, sequential, TLS-skip, 503 Retry-After/413), SessionPackager (chunk plan + WAV slice + timeline slice/rebase), TranscriptAssembler + SpeakersFile, TranscriptPipeline. Validated END-TO-END against the live backend (chunk -> label-merge -> speakers.json). Phase 6 (voiceprints): VoiceprintStore (known_voiceprints, persist named fingerprints, skip Unknown). Wired: 'Send to backend' button + transcript status, auto-send toggle (default off) + self-name setting. All adversarial-review findings fixed. App + XCTest suite build; tests pass.
2026-06-06 00:15:49 -05:00
parent fd7e1a5907
commit 863136aeec
27 changed files with 2108 additions and 22 deletions
@@ -0,0 +1,62 @@
+import XCTest
+import CoreGraphics
+import CoreText
+@testable import Ten31Transcripts
+
+/// Validates the visual adapter against synthetic call frames (no real
+/// screenshots needed): OCR anchors the tiles and the highlight is attributed to
+/// the correct speaker, tracking it as it moves.
+final class GridCallAnalyzerTests: XCTestCase {
+
+    private func drawText(_ s: String, _ ctx: CGContext, center: CGPoint, size: CGFloat) {
+        let font = CTFontCreateWithName("Helvetica-Bold" as CFString, size, nil)
+        let attrs = [kCTFontAttributeName: font,
+                     kCTForegroundColorAttributeName: CGColor(red: 1, green: 1, blue: 1, alpha: 1)] as CFDictionary
+        let line = CTLineCreateWithAttributedString(CFAttributedStringCreate(nil, s as CFString, attrs)!)
+        let b = CTLineGetBoundsWithOptions(line, [])
+        ctx.textPosition = CGPoint(x: center.x - b.width / 2, y: center.y - b.height / 2)
+        CTLineDraw(line, ctx)
+    }
+
+    private func frame(speakingIndex: Int) -> CGImage {
+        let W = 800, H = 600
+        let ctx = CGContext(data: nil, width: W, height: H, bitsPerComponent: 8, bytesPerRow: 0,
+                            space: CGColorSpaceCreateDeviceRGB(),
+                            bitmapInfo: CGImageAlphaInfo.premultipliedLast.rawValue)!
+        ctx.setFillColor(CGColor(red: 0.1, green: 0.1, blue: 0.12, alpha: 1))
+        ctx.fill(CGRect(x: 0, y: 0, width: W, height: H))
+        let rects: [(String, CGRect)] = [
+            ("GRANT", CGRect(x: 40, y: 320, width: 340, height: 230)),
+            ("SARAH", CGRect(x: 420, y: 320, width: 340, height: 230)),
+            ("DMITRI", CGRect(x: 40, y: 50, width: 340, height: 230)),
+            ("ALEX", CGRect(x: 420, y: 50, width: 340, height: 230)),
+        ]
+        for (i, (name, rect)) in rects.enumerated() {
+            ctx.setFillColor(CGColor(red: 0.18, green: 0.18, blue: 0.2, alpha: 1)); ctx.fill(rect)
+            if i == speakingIndex {
+                ctx.setStrokeColor(CGColor(red: 0.1, green: 0.85, blue: 0.2, alpha: 1)); ctx.setLineWidth(14)
+                ctx.stroke(rect.insetBy(dx: 7, dy: 7))
+            }
+            drawText(name, ctx, center: CGPoint(x: rect.midX, y: rect.midY), size: 54)
+        }
+        return ctx.makeImage()!
+    }
+
+    func testReadsNamesAndPicksHighlightedSpeaker() {
+        let obs = SignalAdapter().analyze(cgImage: frame(speakingIndex: 1), at: 0)   // SARAH
+        XCTAssertGreaterThanOrEqual(obs.count, 2)
+        let speaking = obs.filter { $0.speaking }
+        XCTAssertEqual(speaking.count, 1)
+        // SARAH tile center in top-left pixels ≈ (590, 165)
+        XCTAssertEqual(speaking.first?.bbox.midX ?? 0, 590, accuracy: 160)
+        XCTAssertEqual(speaking.first?.bbox.midY ?? 0, 165, accuracy: 160)
+    }
+
+    func testHighlightTracksToAnotherTile() {
+        let obs = SignalAdapter().analyze(cgImage: frame(speakingIndex: 2), at: 1)   // DMITRI
+        let speaking = obs.filter { $0.speaking }
+        XCTAssertEqual(speaking.count, 1)
+        XCTAssertEqual(speaking.first?.bbox.midX ?? 0, 210, accuracy: 160)
+        XCTAssertEqual(speaking.first?.bbox.midY ?? 0, 435, accuracy: 160)
+    }
+}