deepgram compatibility for whisper-client

Cargo.toml

@@ -66,6 +66,7 @@ hypr-transcribe-aws = { path = "crates/transcribe-aws", package = "transcribe-aw
 hypr-transcribe-azure = { path = "crates/transcribe-azure", package = "transcribe-azure" }
 hypr-transcribe-deepgram = { path = "crates/transcribe-deepgram", package = "transcribe-deepgram" }
 hypr-transcribe-gcp = { path = "crates/transcribe-gcp", package = "transcribe-gcp" }
+hypr-transcribe-interface = { path = "crates/transcribe-interface", package = "transcribe-interface" }
 hypr-transcribe-whisper-local = { path = "crates/transcribe-whisper-local", package = "transcribe-whisper-local" }
 hypr-turso = { path = "crates/turso", package = "turso" }
 hypr-whisper = { path = "crates/whisper", package = "whisper" }
@@ -114,6 +115,7 @@ tauri-plugin-tray = { path = "plugins/tray" }
 tauri-plugin-windows = { path = "plugins/windows" }
 
 async-stream = "0.3.6"
+async-trait = "0.1.88"
 futures-channel = "0.3.31"
 futures-core = "0.3.31"
 futures-util = "0.3.31"
@@ -214,7 +216,8 @@ kalosm-model-types = { git = "https://github.com/floneum/floneum", rev = "52967a
 kalosm-sound = { git = "https://github.com/floneum/floneum", rev = "52967ae", default-features = false }
 kalosm-streams = { git = "https://github.com/floneum/floneum", rev = "52967ae" }
 
-deepgram = { version = "0.6.8", default-features = false }
+deepgram = { version = "0.7.0", default-features = false }
+
 libsql = "0.9.17"
 
 block2 = "0.6"

admin/server/tsconfig.json

@@ -6,9 +6,7 @@
     "strict": true,
     "verbatimModuleSyntax": true,
     "skipLibCheck": true,
-    "types": [
-      "node"
-    ],
+    "types": ["node"],
     "jsx": "react-jsx",
     "jsxImportSource": "hono/jsx",
     "outDir": "./dist"

apps/desktop/src-tauri/capabilities/default.json

@@ -52,32 +52,27 @@
     "keygen:default",
     {
       "identifier": "opener:allow-open-url",
-      "allow": [{ "url": "https://**" }, { "url": "mailto:*" }, { "url": "obsidian://**" }]
+      "allow": [
+        { "url": "https://**" },
+        { "url": "mailto:*" },
+        { "url": "obsidian://**" }
+      ]
     },
     {
       "identifier": "opener:allow-open-path",
-      "allow": [
-        { "path": "$APPDATA/*" },
-        { "path": "$APPDATA/**" }
-      ]
+      "allow": [{ "path": "$APPDATA/*" }, { "path": "$APPDATA/**" }]
     },
     {
       "identifier": "fs:allow-exists",
       "allow": [{ "path": "/Applications/*" }]
     },
     {
       "identifier": "fs:allow-watch",
-      "allow": [
-        { "path": "$APPDATA/*" },
-        { "path": "$APPDATA/**" }
-      ]
+      "allow": [{ "path": "$APPDATA/*" }, { "path": "$APPDATA/**" }]
     },
     {
       "identifier": "fs:allow-write-file",
-      "allow": [
-        { "path": "$APPDATA/*" },
-        { "path": "$APPDATA/**" }
-      ]
+      "allow": [{ "path": "$APPDATA/*" }, { "path": "$APPDATA/**" }]
     },
     {
       "identifier": "http:default",

apps/desktop/src-tauri/capabilities/macos.json

@@ -4,7 +4,5 @@
   "description": "macos specific capabilities",
   "windows": ["*"],
   "platforms": ["macOS"],
-  "permissions": [
-    "apple-calendar:default"
-  ]
+  "permissions": ["apple-calendar:default"]
 }

apps/desktop/src-tauri/tauri.conf.json

@@ -57,21 +57,15 @@
             "localePath": "./windows/locales/en-US.wxl"
           }
         },
-        "fragmentPaths": [
-          "./windows/fragments/registry.wxs"
-        ],
-        "componentRefs": [
-          "HyprnoteRegistryEntries"
-        ]
+        "fragmentPaths": ["./windows/fragments/registry.wxs"],
+        "componentRefs": ["HyprnoteRegistryEntries"]
       }
     }
   },
   "plugins": {
     "deep-link": {
       "desktop": {
-        "schemes": [
-          "hypr"
-        ]
+        "schemes": ["hypr"]
       }
     },
     "updater": {

crates/audio-utils/src/lib.rs

@@ -51,7 +51,8 @@ pub fn f32_to_i16_samples(samples: &[f32]) -> Vec<i16> {
 pub fn f32_to_i16_bytes(chunk: Vec<f32>) -> bytes::Bytes {
     let mut bytes = Vec::with_capacity(chunk.len() * 2);
     for sample in chunk {
-        let i16_sample = (sample * I16_SCALE) as i16;
+        let scaled = (sample * I16_SCALE).clamp(-I16_SCALE, I16_SCALE);
+        let i16_sample = scaled as i16;
         bytes.extend_from_slice(&i16_sample.to_le_bytes());
     }
     bytes::Bytes::from(bytes)
@@ -116,3 +117,14 @@ where
 
     Ok(output)
 }
+
+pub fn mix_audio_channels(mic: &[f32], speaker: &[f32]) -> Vec<f32> {
+    let max_len = mic.len().max(speaker.len());
+    (0..max_len)
+        .map(|i| {
+            let mic_sample = mic.get(i).copied().unwrap_or(0.0);
+            let speaker_sample = speaker.get(i).copied().unwrap_or(0.0);
+            (mic_sample + speaker_sample).clamp(-1.0, 1.0)
+        })
+        .collect()
+}

crates/data/build.rs

@@ -1,4 +1,4 @@
-use owhisper_interface::{SpeakerIdentity, Word};
+use owhisper_interface::{SpeakerIdentity, Word2};
 
 fn run(name: &str) {
     let raw_path = format!("src/{}/raw.json", name);
@@ -7,11 +7,11 @@ fn run(name: &str) {
     let raw: serde_json::Value = serde_json::from_str(&raw_content).unwrap();
     let raw_words = raw["results"]["channels"][0]["alternatives"][0]["words"].clone();
 
-    let words: Vec<Word> = raw_words
+    let words: Vec<Word2> = raw_words
         .as_array()
         .unwrap()
         .iter()
-        .map(|v| Word {
+        .map(|v| Word2 {
             text: v["word"].as_str().unwrap().trim().to_string(),
             speaker: Some(SpeakerIdentity::Unassigned {
                 index: v["speaker"].as_u64().unwrap() as u8,

crates/data/src/english_3/raw.json

@@ -6,9 +6,7 @@
     "created": "2025-03-29T00:23:01.963Z",
     "duration": 245.136,
     "channels": 1,
-    "models": [
-      "3b3aabe4-608a-46ac-9585-7960a25daf1a"
-    ],
+    "models": ["3b3aabe4-608a-46ac-9585-7960a25daf1a"],
     "model_info": {
       "3b3aabe4-608a-46ac-9585-7960a25daf1a": {
         "name": "general-nova-3",

crates/data/src/english_4/raw.json

@@ -6,9 +6,7 @@
     "created": "2025-04-09T11:06:10.609Z",
     "duration": 80.496,
     "channels": 1,
-    "models": [
-      "3b3aabe4-608a-46ac-9585-7960a25daf1a"
-    ],
+    "models": ["3b3aabe4-608a-46ac-9585-7960a25daf1a"],
     "model_info": {
       "3b3aabe4-608a-46ac-9585-7960a25daf1a": {
         "name": "general-nova-3",

crates/data/src/english_5/raw.json

@@ -6,9 +6,7 @@
     "created": "2025-04-12T14:21:05.804Z",
     "duration": 2562.888,
     "channels": 1,
-    "models": [
-      "3b3aabe4-608a-46ac-9585-7960a25daf1a"
-    ],
+    "models": ["3b3aabe4-608a-46ac-9585-7960a25daf1a"],
     "model_info": {
       "3b3aabe4-608a-46ac-9585-7960a25daf1a": {
         "name": "general-nova-3",

crates/data/src/english_7/raw.json

@@ -6,9 +6,7 @@
     "created": "2025-07-28T00:50:32.281Z",
     "duration": 58.344,
     "channels": 1,
-    "models": [
-      "3b3aabe4-608a-46ac-9585-7960a25daf1a"
-    ],
+    "models": ["3b3aabe4-608a-46ac-9585-7960a25daf1a"],
     "model_info": {
       "3b3aabe4-608a-46ac-9585-7960a25daf1a": {
         "name": "general-nova-3",

crates/db-script/src/lib.rs

@@ -2,7 +2,7 @@
 pub mod conversation_to_words {
     fn transform(
         conversation: Vec<owhisper_interface::ConversationChunk>,
-    ) -> Vec<owhisper_interface::Word> {
+    ) -> Vec<owhisper_interface::Word2> {
         conversation
             .into_iter()
             .flat_map(|chunk| chunk.transcripts)
@@ -11,7 +11,7 @@ pub mod conversation_to_words {
                     .text
                     .split_whitespace()
                     .filter(|s| !s.is_empty())
-                    .map(|word| owhisper_interface::Word {
+                    .map(|word| owhisper_interface::Word2 {
                         text: word.trim().to_string(),
                         speaker: None,
                         confidence: transcript.confidence,