Merge branch 'phone-interface'

This commit is contained in:
2025-12-17 11:36:58 +08:00
5 changed files with 104 additions and 65 deletions

View File

@@ -103,7 +103,7 @@ DEFAULT_INSTRUCTIONS = """# 角色
- 一次询问一个问题 - 一次询问一个问题
- 不要在你的回复中使用 emojis, asterisks, markdown, 或其他特殊字符 - 不要在你的回复中使用 emojis, asterisks, markdown, 或其他特殊字符
- 不同阶段直接的过渡语句自然 - 不同阶段直接的过渡语句自然
- 你已经说过下面的开场白所以不需要重复说:“您好,这里是无锡交警,我将为您远程处理交通事故。请将人员撤离至路侧安全区域,开启危险报警双闪灯、放置三角警告牌、做好安全防护,谨防二次事故伤害。若您已经准备好了,请点击继续办理,如需人工服务,请说转人工。” - 你已经说过下面的开场白,用户点击继续办理说明已经认可,所以不需要重复说:“您好,这里是无锡交警,我将为您远程处理交通事故。请将人员撤离至路侧安全区域,开启危险报警双闪灯、放置三角警告牌、做好安全防护,谨防二次事故伤害。若您已经准备好了,请点击继续办理,如需人工服务,请说转人工。”
""" """
DEFAULT_TALKING_MODE = 'push_to_talk' DEFAULT_TALKING_MODE = 'push_to_talk'
@@ -1030,6 +1030,7 @@ async def entrypoint(ctx: JobContext, avatar_dispatcher_url: str = None, vision_
logger.info("Using default DeepSeek backend") logger.info("Using default DeepSeek backend")
llm = openai.LLM.with_deepseek( llm = openai.LLM.with_deepseek(
model='deepseek-chat', model='deepseek-chat',
temperature=0.1
) )
session = AgentSession( session = AgentSession(

View File

@@ -715,7 +715,7 @@ export function PhoneSimulator({
> >
<PhoneIcon className="w-8 h-8" /> <PhoneIcon className="w-8 h-8" />
</div> </div>
<span className="font-medium text-white">Call Agent</span> <span className="font-medium text-white"></span>
</button> </button>
<div className="relative"> <div className="relative">
@@ -725,7 +725,7 @@ export function PhoneSimulator({
> >
<VoiceIcon className="w-3 h-3" /> <VoiceIcon className="w-3 h-3" />
<span> <span>
{currentVoiceId === "BV001_streaming" ? "Female Voice" : "Male Voice"} {currentVoiceId === "BV001_streaming" ? "女性声音" : "男性声音"}
</span> </span>
</button> </button>
{showVoiceMenu && ( {showVoiceMenu && (
@@ -745,7 +745,7 @@ export function PhoneSimulator({
: "text-white" : "text-white"
}`} }`}
> >
<span>Female Voice</span> <span></span>
{currentVoiceId === "BV001_streaming" && <CheckIcon />} {currentVoiceId === "BV001_streaming" && <CheckIcon />}
</button> </button>
<button <button
@@ -760,7 +760,7 @@ export function PhoneSimulator({
: "text-white" : "text-white"
}`} }`}
> >
<span>Male Voice</span> <span></span>
{currentVoiceId === "BV002_streaming" && ( {currentVoiceId === "BV002_streaming" && (
<CheckIcon /> <CheckIcon />
)} )}
@@ -1198,6 +1198,20 @@ export function PhoneSimulator({
</button> </button>
</div> </div>
)} )}
{/* Fallback: Show End Call Button when in push-to-talk mode but no agent/audio */}
{phoneMode === "normal" &&
isPushToTalkMode &&
!voiceAssistant.agent && (
<div className="w-full flex items-center justify-center">
<button
className="p-4 rounded-full bg-red-500 text-white hover:bg-red-600 transition-colors"
onClick={handleDisconnect}
>
<PhoneOffIcon className="w-6 h-6" />
</button>
</div>
)}
</div> </div>
</div> </div>
) )

View File

@@ -479,21 +479,34 @@ export default function Playground({
]); ]);
const instructionsContent = ( const instructionsContent = (
<ConfigurationPanelItem title="Instructions"> <>
<textarea <ConfigurationPanelItem title="Instructions">
className="w-full bg-gray-950 text-white text-sm p-3 rounded-md border border-gray-800 focus:border-gray-600 focus:outline-none transition-colors resize-none disabled:opacity-50 disabled:cursor-not-allowed" <textarea
style={{ minHeight: "80px" }} className="w-full bg-gray-950 text-white text-sm p-3 rounded-md border border-gray-800 focus:border-gray-600 focus:outline-none transition-colors resize-none disabled:opacity-50 disabled:cursor-not-allowed"
rows={3} style={{ minHeight: "80px" }}
placeholder="Enter system instructions for the agent..." rows={3}
value={config.settings.instructions} placeholder="Enter system instructions for the agent..."
onChange={(e) => { value={config.settings.instructions}
const newSettings = { ...config.settings }; onChange={(e) => {
newSettings.instructions = e.target.value; const newSettings = { ...config.settings };
setUserSettings(newSettings); newSettings.instructions = e.target.value;
}} setUserSettings(newSettings);
disabled={roomState !== ConnectionState.Disconnected} }}
/> disabled={roomState !== ConnectionState.Disconnected}
</ConfigurationPanelItem> />
</ConfigurationPanelItem>
<ConfigurationPanelItem title="Color">
<ColorPicker
colors={themeColors}
selectedColor={config.settings.theme_color}
onSelect={(color) => {
const userSettings = { ...config.settings };
userSettings.theme_color = color;
setUserSettings(userSettings);
}}
/>
</ConfigurationPanelItem>
</>
); );
const handleRpcCall = useCallback(async () => { const handleRpcCall = useCallback(async () => {
@@ -516,13 +529,13 @@ export default function Playground({
const settingsTileContent = useMemo(() => { const settingsTileContent = useMemo(() => {
return ( return (
<div className="flex flex-col h-full w-full items-start overflow-y-auto"> <div className="flex flex-col h-full w-full items-start overflow-y-auto">
{config.description && ( {/* {config.description && (
<ConfigurationPanelItem title="Description"> <ConfigurationPanelItem title="Description">
{config.description} {config.description}
</ConfigurationPanelItem> </ConfigurationPanelItem>
)} )} */}
<ConfigurationPanelItem title="Room"> {/* <ConfigurationPanelItem title="Room">
<div className="flex flex-col gap-2"> <div className="flex flex-col gap-2">
<EditableNameValueRow <EditableNameValueRow
name="Room name" name="Room name"
@@ -556,9 +569,9 @@ export default function Playground({
} }
/> />
</div> </div>
</ConfigurationPanelItem> </ConfigurationPanelItem> */}
<ConfigurationPanelItem title="Agent"> {/* <ConfigurationPanelItem title="Agent">
<div className="flex flex-col gap-2"> <div className="flex flex-col gap-2">
<EditableNameValueRow <EditableNameValueRow
name="Agent name" name="Agent name"
@@ -621,9 +634,9 @@ export default function Playground({
. .
</p> </p>
</div> </div>
</ConfigurationPanelItem> </ConfigurationPanelItem> */}
<ConfigurationPanelItem title="User"> {/* <ConfigurationPanelItem title="User">
<div className="flex flex-col gap-2"> <div className="flex flex-col gap-2">
<EditableNameValueRow <EditableNameValueRow
name="Name" name="Name"
@@ -675,7 +688,7 @@ export default function Playground({
connectionState={roomState} connectionState={roomState}
/> />
</div> </div>
</ConfigurationPanelItem> </ConfigurationPanelItem> */}
{roomState === ConnectionState.Connected && {roomState === ConnectionState.Connected &&
config.settings.inputs.screen && ( config.settings.inputs.screen && (
@@ -725,19 +738,6 @@ export default function Playground({
<AudioInputTile trackRef={localMicTrack} /> <AudioInputTile trackRef={localMicTrack} />
</ConfigurationPanelItem> </ConfigurationPanelItem>
)} )}
<div className="w-full">
<ConfigurationPanelItem title="Color">
<ColorPicker
colors={themeColors}
selectedColor={config.settings.theme_color}
onSelect={(color) => {
const userSettings = { ...config.settings };
userSettings.theme_color = color;
setUserSettings(userSettings);
}}
/>
</ConfigurationPanelItem>
</div>
{config.show_qr && ( {config.show_qr && (
<div className="w-full"> <div className="w-full">
<ConfigurationPanelItem title="QR Code"> <ConfigurationPanelItem title="QR Code">
@@ -748,7 +748,6 @@ export default function Playground({
</div> </div>
); );
}, [ }, [
config.description,
config.settings, config.settings,
config.show_qr, config.show_qr,
localParticipant, localParticipant,
@@ -842,7 +841,7 @@ export default function Playground({
return ( return (
<> <>
<PlaygroundHeader {/* <PlaygroundHeader
title={config.title} title={config.title}
logo={logo} logo={logo}
githubLink={config.github_link} githubLink={config.github_link}
@@ -856,10 +855,10 @@ export default function Playground({
handleDisconnect(); handleDisconnect();
} }
}} }}
/> /> */}
<div <div
className={`flex gap-4 py-4 grow w-full selection:bg-${config.settings.theme_color}-900`} className={`flex gap-4 py-4 grow w-full selection:bg-${config.settings.theme_color}-900`}
style={{ height: `calc(100% - ${headerHeight}px)` }} style={{ height: `100%` }}
> >
<div className="flex flex-col grow basis-1/2 gap-4 h-full lg:hidden"> <div className="flex flex-col grow basis-1/2 gap-4 h-full lg:hidden">
<PlaygroundTabbedTile <PlaygroundTabbedTile
@@ -926,14 +925,14 @@ export default function Playground({
</PlaygroundTile> </PlaygroundTile>
</div> </div>
)} )}
<PlaygroundTile {/* <PlaygroundTile
padding={false} padding={false}
backgroundColor="gray-950" backgroundColor="gray-950"
className="h-full w-full basis-1/4 items-start overflow-y-auto hidden max-w-[480px] lg:flex" className="h-full w-full basis-1/4 items-start overflow-y-auto hidden max-w-[480px] lg:flex"
childrenClassName="h-full grow items-start" childrenClassName="h-full grow items-start"
> >
{settingsTileContent} {settingsTileContent}
</PlaygroundTile> </PlaygroundTile> */}
</div> </div>
</> </>
); );

View File

@@ -100,7 +100,18 @@ export const PlaygroundTabbedTile: React.FC<PlaygroundTabbedTileProps> = ({
padding: `${contentPadding * 4}px`, padding: `${contentPadding * 4}px`,
}} }}
> >
{tabs[activeTab].content} {tabs.map((tab, index) => (
<div
key={index}
style={{
display: index === activeTab ? 'block' : 'none',
height: '100%',
width: '100%',
}}
>
{tab.content}
</div>
))}
</div> </div>
</div> </div>
); );

View File

@@ -11,7 +11,7 @@ import {
Track, Track,
TranscriptionSegment, TranscriptionSegment,
} from "livekit-client"; } from "livekit-client";
import { useEffect, useState } from "react"; import { useEffect, useState, useRef } from "react";
export function TranscriptionTile({ export function TranscriptionTile({
agentAudioTrack, agentAudioTrack,
@@ -30,39 +30,51 @@ export function TranscriptionTile({
participant: localParticipant.localParticipant, participant: localParticipant.localParticipant,
}); });
const [transcripts, setTranscripts] = useState<Map<string, ChatMessageType>>(
new Map(),
);
const [messages, setMessages] = useState<ChatMessageType[]>([]); const [messages, setMessages] = useState<ChatMessageType[]>([]);
const { chatMessages, send: sendChat } = useChat(); const { chatMessages, send: sendChat } = useChat();
const transcriptMapRef = useRef<Map<string, ChatMessageType>>(new Map());
// store transcripts // Build messages from segments and chat - always rebuild from current state
useEffect(() => { useEffect(() => {
const transcriptMap = transcriptMapRef.current;
// Process agent segments - update existing or add new
if (agentAudioTrack) { if (agentAudioTrack) {
agentMessages.segments.forEach((s) => agentMessages.segments.forEach((s) => {
transcripts.set( const existing = transcriptMap.get(s.id);
transcriptMap.set(
s.id, s.id,
segmentToChatMessage( segmentToChatMessage(
s, s,
transcripts.get(s.id), existing,
agentAudioTrack.participant, agentAudioTrack.participant,
), ),
), );
); });
} }
localMessages.segments.forEach((s) => // Process local segments - update existing or add new
transcripts.set( localMessages.segments.forEach((s) => {
const existing = transcriptMap.get(s.id);
transcriptMap.set(
s.id, s.id,
segmentToChatMessage( segmentToChatMessage(
s, s,
transcripts.get(s.id), existing,
localParticipant.localParticipant, localParticipant.localParticipant,
), ),
), );
); });
const allMessages = Array.from(transcripts.values()); // Build all messages
const allMessages: ChatMessageType[] = [];
// Add all transcript messages
transcriptMap.forEach((msg) => {
allMessages.push(msg);
});
// Add chat messages
for (const msg of chatMessages) { for (const msg of chatMessages) {
const isAgent = agentAudioTrack const isAgent = agentAudioTrack
? msg.from?.identity === agentAudioTrack.participant?.identity ? msg.from?.identity === agentAudioTrack.participant?.identity
@@ -79,6 +91,7 @@ export function TranscriptionTile({
name = "Unknown"; name = "Unknown";
} }
} }
allMessages.push({ allMessages.push({
name, name,
message: msg.message, message: msg.message,
@@ -86,10 +99,11 @@ export function TranscriptionTile({
isSelf: isSelf, isSelf: isSelf,
}); });
} }
// Sort by timestamp
allMessages.sort((a, b) => a.timestamp - b.timestamp); allMessages.sort((a, b) => a.timestamp - b.timestamp);
setMessages(allMessages); setMessages(allMessages);
}, [ }, [
transcripts,
chatMessages, chatMessages,
localParticipant.localParticipant, localParticipant.localParticipant,
agentAudioTrack?.participant, agentAudioTrack?.participant,