reya/coop
1
0
Fork 1
Code Issues Pull Requests Releases 4 Activity

feat: add support for speech to text (#68)

Browse Source 
Reviewed-on: #68
This commit was merged in pull request #68.
This commit is contained in:
reya
2026-10-27 06:25:19 +00:00
parent b0121f8df2
commit 77b2a5f3cd
28 changed files with 2238 additions and 230 deletions
Download Patch File Download Diff File Expand all files Collapse all files
crates/chat_ui/Cargo.toml
+3 -1
View File
@@ -6,14 +6,16 @@ publish.workspace = true
[dependencies]
state = { path = "../state" }
speech = { path = "../speech" }
ui = { path = "../ui" }
gpui-kit.workspace = true
gpui-kit = { workspace = true, features = ["speech"] }
common = { path = "../common" }
person = { path = "../person" }
chat = { path = "../chat" }
settings = { path = "../settings" }
gpui.workspace = true
instant.workspace = true
nostr-sdk.workspace = true
anyhow.workspace = true
itertools.workspace = true
crates/chat_ui/src/actions.rs
+4 -1
View File
@@ -1,4 +1,4 @@
use gpui::Action;
use gpui::{Action, actions};
use nostr_sdk::prelude::*;
use serde::Deserialize;
use settings::SignerKind;
@@ -15,3 +15,6 @@ pub enum Command {
Njump(PublicKey),
Trace(EventId),
}
// Start or stop dictation in the focused composer
actions!(chat, [ToggleDictation]);
crates/chat_ui/src/lib.rs
+136 -30
View File
@@ -1,5 +1,6 @@
use std::collections::{BTreeMap, HashMap, HashSet};
use std::sync::{Arc, LazyLock, RwLock};
use std::time::Duration;
pub use actions::*;
use anyhow::Error;
@@ -20,6 +21,7 @@ use gpui_kit::component::input::{Input, InputEvent, InputState, Textarea, Textar
use gpui_kit::component::menu::DropdownMenu;
use gpui_kit::component::notification::Notification;
use gpui_kit::component::scroll::Scrollbar;
use gpui_kit::component::speech::{SpeechButton, SpeechEvent, SpeechState, SpeechWaveform};
use gpui_kit::component::tooltip::Tooltip;
use gpui_kit::component::{ActiveTheme, Disableable, Icon, Sizable, WindowExt, h_flex, v_flex};
use itertools::Itertools;
@@ -91,6 +93,12 @@ pub struct ChatPanel {
/// Chat input state
input: Entity<TextareaState>,
/// Dictation session for the chat input
speech: Entity<SpeechState>,
/// When the running dictation started, for the elapsed counter
speech_started: Option<instant::Instant>,
/// Subject input state
subject_input: Entity<InputState>,
@@ -116,7 +124,7 @@ pub struct ChatPanel {
tasks: Vec<Task<Result<(), Error>>>,
/// Event subscriptions
subscriptions: SmallVec<[Subscription; 3]>,
subscriptions: SmallVec<[Subscription; 4]>,
}
impl ChatPanel {
@@ -160,6 +168,19 @@ impl ChatPanel {
.clean_on_escape()
});
// Define dictation state
let speech = cx.new(|cx| {
let state = SpeechState::new(cx)
.recognizer(speech::recognizer())
// Batch transcription of a long recording can take a while.
.stop_timeout(Duration::from_secs(180));
#[cfg(target_arch = "wasm32")]
let state = state.input(speech::input());
state
});
// Define subject input state
let subject_input = cx.new(|cx| InputState::new(window, cx).placeholder("New subject..."));
let subject_bar = cx.new(|_cx| false);
@@ -178,15 +199,39 @@ impl ChatPanel {
subscriptions.push(
// Subscribe the subject input event
cx.subscribe_in(
&subject_input,
window,
move |this, _input, event, window, cx| {
if let InputEvent::PressEnter { .. } = event {
this.change_subject(window, cx);
};
},
),
cx.subscribe_in(&subject_input, window, move |this, _, event, window, cx| {
if let InputEvent::PressEnter { .. } = event {
this.change_subject(window, cx);
};
}),
);
subscriptions.push(
// Insert the dictated text into the chat input
cx.subscribe_in(&speech, window, move |this, _speech, event, window, cx| {
match event {
SpeechEvent::Started => {
this.speech_started = Some(instant::Instant::now());
cx.notify();
}
SpeechEvent::Final(text) => {
this.speech_started = None;
if !text.is_empty() {
this.input
.update(cx, |input, cx| input.insert(text.clone(), window, cx));
}
}
SpeechEvent::Error(error) => {
this.speech_started = None;
window.push_notification(Notification::error(error.to_string()), cx);
}
SpeechEvent::Cancelled => {
this.speech_started = None;
}
_ => {}
};
}),
);
// Define all functions that will run after the current cycle
@@ -206,6 +251,8 @@ impl ChatPanel {
directory,
list_state,
input,
speech,
speech_started: None,
subject_input,
subject_bar,
replies_to,
@@ -835,6 +882,17 @@ impl ChatPanel {
self.directory.read(cx).person(public_key, cx)
}
fn toggle_dictation(
&mut self,
_: &ToggleDictation,
_window: &mut Window,
cx: &mut Context<Self>,
) {
self.speech.update(cx, |speech, cx| {
speech.toggle(cx);
});
}
fn on_command(&mut self, command: &Command, window: &mut Window, cx: &mut Context<Self>) {
match command {
Command::Insert(content) => {
@@ -1390,21 +1448,26 @@ impl ChatPanel {
}
let reply_id = *id;
let copy_id = *id;
let dropdown_id = *id;
let public_key = *public_key;
actions
.reply(cx.listener(move |this, _event, _window, cx| {
this.reply_to(&reply_id, cx);
}))
.copy(cx.listener(move |this, _event, _window, cx| {
this.copy_message(&copy_id, cx);
}))
.dropdown(move |menu, _window, _cx| {
menu.menu("Copy author", Box::new(Command::Copy(public_key)))
.menu("Seen on", Box::new(Command::Trace(dropdown_id)))
})
let actions = actions.reply(cx.listener(move |this, _event, _window, cx| {
this.reply_to(&reply_id, cx);
}));
let actions = if let Some(message) = self.message(id) {
actions.copy(
SharedString::from(format!("message-copy-{id}")),
message.preview(),
)
} else {
actions
};
actions.dropdown(move |menu, _window, _cx| {
menu.menu("Copy author", Box::new(Command::Copy(public_key)))
.menu("Seen on", Box::new(Command::Trace(dropdown_id)))
})
}
fn render_attachment(&self, url: &Url, cx: &Context<Self>) -> impl IntoElement {
@@ -1669,7 +1732,12 @@ impl ChatPanel {
items
}
fn render_config_menu(&self, _window: &mut Window, cx: &Context<Self>) -> impl IntoElement {
fn render_config_menu(
&self,
disabled: bool,
_window: &mut Window,
cx: &Context<Self>,
) -> impl IntoElement {
let (backup, signer_kind) = self
.room
.read_with(cx, |this, _cx| {
@@ -1682,6 +1750,7 @@ impl ChatPanel {
.icon(IconName::Settings2)
.tooltip("Configuration")
.ghost()
.disabled(disabled)
.dropdown_menu(move |this, _window, _cx| {
let auto = matches!(signer_kind, SignerKind::Auto);
let encryption = matches!(signer_kind, SignerKind::Encryption);
@@ -1712,7 +1781,12 @@ impl ChatPanel {
})
}
fn render_emoji_menu(&self, _window: &Window, cx: &Context<Self>) -> impl IntoElement {
fn render_emoji_menu(
&self,
disabled: bool,
_window: &Window,
cx: &Context<Self>,
) -> impl IntoElement {
let nostr = NostrRegistry::global(cx);
let current_user = nostr.read(cx).current_user();
@@ -1729,6 +1803,7 @@ impl ChatPanel {
Button::new("emoji")
.icon(IconName::Emoji)
.ghost()
.disabled(disabled)
.dropdown_menu_with_anchor(gpui::Anchor::BottomLeft, move |this, _window, _cx| {
let menu = REACTION_EMOJIS.iter().fold(this, |this, emoji| {
this.menu(*emoji, Box::new(Command::Insert(emoji.to_string())))
@@ -1802,10 +1877,12 @@ impl Render for ChatPanel {
let is_typing = !self.input.read(cx).value().trim().is_empty();
let pending_attachments = !self.encrypted_attachments.read(cx).is_empty();
let dictating = self.speech.read(cx).status().is_active();
v_flex()
.image_cache(retain_all(self.id.clone()))
.on_action(cx.listener(Self::on_command))
.on_action(cx.listener(Self::toggle_dictation))
.size_full()
.when(*self.subject_bar.read(cx), |this| {
this.child(
@@ -1879,26 +1956,55 @@ impl Render for ChatPanel {
.items_end()
.child(
Button::new("upload")
.icon(IconName::Plus)
.icon(IconName::Paperclip)
.tooltip("Upload media")
.loading(self.uploading)
.disabled(self.uploading)
.disabled(dictating || self.uploading)
.ghost()
.on_click(cx.listener(move |this, _ev, window, cx| {
this.upload(window, cx);
})),
)
.child(Textarea::new(&self.input).appearance(false).flex_1())
.when(dictating, |this| {
this.child(
h_flex()
.h_8()
.flex_1()
.gap_2()
.pl_1()
.child(SpeechWaveform::new(&self.speech).small().flex_1())
.when_some(self.speech_started, |this, started| {
this.child(speech::elapsed(started, cx))
}),
)
})
.when(!dictating, |this| {
this.child(
Textarea::new(&self.input)
.appearance(false)
.small()
.min_h_8()
.py_1()
.flex_1(),
)
})
.child(
h_flex()
.items_end()
.gap_1()
.pl_1()
.child(SpeechButton::new(&self.speech)),
)
.child(
h_flex()
.pl_1()
.gap_1()
.child(self.render_emoji_menu(window, cx))
.child(self.render_config_menu(window, cx))
.child(self.render_emoji_menu(dictating, window, cx))
.child(self.render_config_menu(dictating, window, cx))
.child(
Button::new("send")
.icon(IconName::PaperPlaneFill)
.disabled(self.uploading)
.disabled(dictating || self.uploading)
.ghost()
.on_click(cx.listener(move |this, _ev, window, cx| {
this.send_text_message(window, cx);