YC (S26) | Open Computer History | Continuously record your company computer work, map your workflows, help you find work worth automating, and power your agents' context
Rust agentsagiaiai-memoryaudio-recordingcomputer-visionhermeshermes-agentllmlocal-ailocal-firstmachine-learningmcpmultimodalopenclawprivacyrewindscreen-recordingspeech-to-textycombinator
fix: preserve distinct audio segments in search results
Ezra Elletteacf844b85625
diff --git a/crates/screenpipe-db/src/db/search.rs b/crates/screenpipe-db/src/db/search.rsindex bad4290c1..aad0c5c18 100644--- a/crates/screenpipe-db/src/db/search.rs+++ b/crates/screenpipe-db/src/db/search.rs@@ -1202,9 +1202,10 @@ impl DatabaseManager {format!("WHERE {}", conditions.join(" AND "))};- // complete sql with group, order, limit and offset+ // Group only tag-join duplicates. Distinct speech segments can share a+ // chunk and offset_index, so grouping by those loses recorded speech.let sql = format!(- "{} {} GROUP BY audio_transcriptions.audio_chunk_id, audio_transcriptions.offset_index ORDER BY audio_transcriptions.timestamp {order_dir}, audio_transcriptions.audio_chunk_id {order_dir}, audio_transcriptions.offset_index {order_dir} LIMIT ? OFFSET ?",+ "{} {} GROUP BY audio_transcriptions.id ORDER BY audio_transcriptions.timestamp {order_dir}, audio_transcriptions.audio_chunk_id {order_dir}, audio_transcriptions.offset_index {order_dir}, audio_transcriptions.id {order_dir} LIMIT ? OFFSET ?",base_sql,where_clause,order_dir = match order {diff --git a/crates/screenpipe-db/tests/audio_segment_search_test.rs b/crates/screenpipe-db/tests/audio_segment_search_test.rsnew file mode 100644index 000000000..e0e157064--- /dev/null+++ b/crates/screenpipe-db/tests/audio_segment_search_test.rs@@ -0,0 +1,98 @@+// screenpipe — AI that knows everything you've seen, said, or heard+// https://screenpipe.com++use screenpipe_db::{AudioDevice, DatabaseManager, DeviceType, Order, TagContentType};++#[tokio::test]+async fn preserves_segments_with_shared_chunk_offset_and_paginates_without_tag_duplicates() {+ let temp = tempfile::tempdir().unwrap();+ let db = DatabaseManager::new(+ temp.path().join("audio.sqlite").to_str().unwrap(),+ Default::default(),+ )+ .await+ .unwrap();+ let chunk = db+ .insert_audio_chunk("realistic-segmented-speech.mp4", None)+ .await+ .unwrap();+ let device = AudioDevice {+ name: "System Audio".into(),+ device_type: DeviceType::Output,+ };+ let captured = chrono::Utc::now();+ for (index, text) in [+ "Juniper artwork",+ "requires review",+ "before customer sharing",+ ]+ .iter()+ .enumerate()+ {+ db.insert_audio_transcription(+ chunk,+ text,+ 0,+ "test",+ &device,+ None,+ Some(index as f64 * 2.0),+ Some(index as f64 * 2.0 + 2.0),+ Some(captured),+ )+ .await+ .unwrap();+ }+ db.add_tags(+ chunk,+ TagContentType::Audio,+ vec!["juniper".into(), "review".into()],+ )+ .await+ .unwrap();+ for (order, expected) in [+ (+ Order::Ascending,+ vec![+ "Juniper artwork",+ "requires review",+ "before customer sharing",+ ],+ ),+ (+ Order::Descending,+ vec![+ "before customer sharing",+ "requires review",+ "Juniper artwork",+ ],+ ),+ ] {+ let mut observed = Vec::new();+ for offset in 0..4 {+ let page = db+ .search_audio_ordered(+ "",+ 1,+ offset,+ None,+ None,+ None,+ None,+ None,+ None,+ None,+ None,+ &[],+ order,+ )+ .await+ .unwrap();+ for segment in page {+ assert_eq!(segment.tags.len(), 2);+ observed.push(segment.transcription);+ }+ }+ assert_eq!(observed, expected);+ }+}