ten31-transcripts/Ten31TranscriptsTests/Phase5Tests.swift

import XCTest
@testable import Ten31Transcripts

final class Phase5Tests: XCTestCase {
    func testPlanChunksShort() {
        let c = SessionPackager.planChunks(durationSec: 70)
        XCTAssertEqual(c.count, 1)
        XCTAssertEqual(c[0].end, 70, accuracy: 0.001)
    }

    func testPlanChunksLongOverlapsWindowsWithContiguousBodies() {
        let c = SessionPackager.planChunks(durationSec: 400, chunkSeconds: 150, overlapSeconds: 15)
        XCTAssertEqual(c.count, 3)
        // Owned bodies tile the call with no gaps/overlap.
        XCTAssertEqual(c[0].bodyStart, 0);   XCTAssertEqual(c[0].bodyEnd, 150)
        XCTAssertEqual(c[1].bodyStart, 150); XCTAssertEqual(c[1].bodyEnd, 300)
        XCTAssertEqual(c[2].bodyEnd, 400)
        // Sliced windows overlap by the margin (and clamp at the ends).
        XCTAssertEqual(c[0].start, 0); XCTAssertEqual(c[0].end, 165)   // +15 trailing
        XCTAssertEqual(c[1].start, 135)                                // -15 leading
        XCTAssertLessThan(c[1].start, c[0].end)                        // windows overlap
        XCTAssertEqual(c[2].end, 400)                                  // clamped
    }

    func testAssembleDedupsOverlapByBody() {
        // A segment at global 152–156 sits in chunk1's body but also in chunk0's
        // trailing margin (overlap). It must be kept exactly once (by chunk1).
        let r0 = #"{"duration":165,"speakers":[{"cluster":"S0","name":"A","source":"visual","overlap_confidence":0.9}],"segments":[{"start_ms":152000,"end_ms":156000,"speaker":"A","text":"boundary"}],"fingerprints":{},"models":{}}"#
        let r1 = #"{"duration":180,"speakers":[{"cluster":"S0","name":"A","source":"visual","overlap_confidence":0.9}],"segments":[{"start_ms":17000,"end_ms":21000,"speaker":"A","text":"boundary"}],"fingerprints":{},"models":{}}"#
        let c0 = try! JSONDecoder().decode(LabelMergeResponse.self, from: Data(r0.utf8))
        let c1 = try! JSONDecoder().decode(LabelMergeResponse.self, from: Data(r1.utf8))
        let asm = TranscriptAssembler.assemble(sessionId: "s", app: "meet", chunks: [
            .init(chunkStart: 0, response: c0, bodyStart: 0, bodyEnd: 150),
            .init(chunkStart: 135, response: c1, bodyStart: 150, bodyEnd: 300),
        ])
        XCTAssertEqual(asm.speakersFile.segments.count, 1)                       // deduped
        XCTAssertEqual(asm.speakersFile.segments[0].start, 152, accuracy: 0.01)
    }

    func testDropStuckSpansRemovesWholeCallCue() {
        let segs = [
            VisualTimeline.Segment(start: 0, end: 1900, name: "Grant Gilliam", confidence: 1, source: "vision"), // stuck whole-call tile
            VisualTimeline.Segment(start: 100, end: 130, name: "Matt Odell", confidence: 0.9, source: "vision"),  // real
            VisualTimeline.Segment(start: 0, end: 1900, name: "Grant", confidence: 1, source: "mic_vad"),         // self span: keep
        ]
        let out = TranscriptPipeline.dropStuckSpans(segs, duration: 1976)
        XCTAssertFalse(out.contains { $0.name == "Grant Gilliam" })   // 96% of call in one span → dropped
        XCTAssertTrue(out.contains { $0.name == "Matt Odell" })       // short real span kept
        XCTAssertTrue(out.contains { $0.source == "mic_vad" })        // self never dropped
    }

    func testRebaseClipsAndRebases() throws {
        let segs = [
            VisualTimeline.Segment(start: 140, end: 160, name: "A", confidence: 0.9, source: "vision"),
            VisualTimeline.Segment(start: 200, end: 260, name: "B", confidence: 0.8, source: "vision"),
        ]
        let data = try SessionPackager.rebasedTimelineData(segs, start: 150, end: 300)
        let arr = try XCTUnwrap(JSONSerialization.jsonObject(with: data) as? [[String: Any]])
        XCTAssertEqual(arr.count, 2)
        XCTAssertEqual(arr[0]["start"] as? Double, 0)
        XCTAssertEqual(arr[0]["end"] as? Double, 10)
        XCTAssertEqual(arr[1]["start"] as? Double, 50)
        XCTAssertEqual(arr[1]["end"] as? Double, 110)
    }

    func testRebaseSelfVadClipsAndRebases() throws {
        let spans = [VADSpan(start: 140, end: 160, confidence: 0.9),
                     VADSpan(start: 200, end: 260, confidence: 0.8)]
        let data = try SessionPackager.rebasedSelfVadData(spans, start: 150, end: 300)
        let arr = try XCTUnwrap(JSONSerialization.jsonObject(with: data) as? [[String: Any]])
        XCTAssertEqual(arr.count, 2)
        XCTAssertEqual(arr[0]["start"] as? Double, 0)
        XCTAssertEqual(arr[0]["end"] as? Double, 10)     // 160 clipped at 150 base → 0–10
        XCTAssertEqual(arr[1]["start"] as? Double, 50)
        XCTAssertEqual(arr[1]["end"] as? Double, 110)
        XCTAssertNil(arr[0]["name"])                      // self_vad carries no name
    }

    func testAssembleRanksMicChannelOverVisual() throws {
        // Same person resolved by visual in one chunk and by the mic channel in
        // another → the mic-channel attribution (the user's own track) wins.
        let visual = #"{"duration":100,"speakers":[{"cluster":"Speaker_0","name":"Grant","source":"visual","overlap_confidence":0.99,"fingerprint":[0.1]}],"segments":[],"fingerprints":{},"models":{}}"#
        let mic = #"{"duration":100,"speakers":[{"cluster":"mic","name":"Grant","source":"mic_channel","fingerprint":[0.2]}],"segments":[],"fingerprints":{"Grant":[0.2]},"models":{}}"#
        let rv = try JSONDecoder().decode(LabelMergeResponse.self, from: Data(visual.utf8))
        let rm = try JSONDecoder().decode(LabelMergeResponse.self, from: Data(mic.utf8))
        let asm = TranscriptAssembler.assemble(sessionId: "s", app: "meet",
            chunks: [.init(chunkStart: 0, response: rv), .init(chunkStart: 100, response: rm)])
        XCTAssertEqual(asm.speakersFile.speakers.first { $0.name == "Grant" }?.source, "mic_channel")
    }

    func testAssembleOffsetsAndUnifies() throws {
        let resp0 = #"{"duration":150,"speakers":[{"cluster":"Speaker_0","name":"Grant","source":"visual","overlap_confidence":0.99,"fingerprint":[0.1,0.2]}],"segments":[{"start_ms":1000,"end_ms":2000,"speaker":"Grant","text":"hi"}],"fingerprints":{"Grant":[0.1,0.2]},"models":{"diarization":"x"}}"#
        let resp1 = #"{"duration":100,"speakers":[{"cluster":"Speaker_0","name":"Sarah","source":"voiceprint","match_similarity":0.7,"fingerprint":[0.3,0.4]},{"cluster":"Speaker_1","name":"Unknown_0","source":"unmatched"}],"segments":[{"start_ms":500,"end_ms":1500,"speaker":"Sarah","text":"hello"}],"fingerprints":{"Sarah":[0.3,0.4]},"models":{"diarization":"x"}}"#
        let r0 = try JSONDecoder().decode(LabelMergeResponse.self, from: Data(resp0.utf8))
        let r1 = try JSONDecoder().decode(LabelMergeResponse.self, from: Data(resp1.utf8))
        let asm = TranscriptAssembler.assemble(sessionId: "s", app: "meet",
            chunks: [.init(chunkStart: 0, response: r0), .init(chunkStart: 150, response: r1)])
        XCTAssertEqual(asm.speakersFile.segments.count, 2)
        XCTAssertEqual(asm.speakersFile.segments[0].start, 1, accuracy: 0.001)
        XCTAssertEqual(asm.speakersFile.segments[1].start, 150.5, accuracy: 0.001)
        XCTAssertEqual(asm.speakersFile.speakers.count, 3)
        XCTAssertNotNil(asm.fingerprints["Grant"])
        XCTAssertNotNil(asm.fingerprints["Sarah"])
        XCTAssertNil(asm.fingerprints["Unknown_0"])
    }
}