Add voice message transcription support (#54)
* Add voice message transcription support (all channels) Adds OpenAI Whisper transcription for voice messages across all channels: - Telegram: ctx.message.voice - WhatsApp: audioMessage via downloadMediaMessage - Signal: audio attachments from local files - Slack: audio files via url_private_download - Discord: audio attachments Voice messages sent to agent as "[Voice message]: <transcript>" Configuration (config takes priority over env): - lettabot.yaml: transcription.apiKey, transcription.model - Env: OPENAI_API_KEY, TRANSCRIPTION_MODEL Closes #47 Written by Cameron ◯ Letta Code "The best interface is no interface - just talk." * Add voice message documentation to README - Add Voice Messages to features list - Add configuration section for transcription - Document supported channels Written by Cameron ◯ Letta Code * Notify users when voice transcription is not configured Instead of silently ignoring voice messages, send a helpful message linking to the documentation. Written by Cameron ◯ Letta Code * feat: upgrade to letta-code-sdk main + fix Signal voice transcription - Switch from published SDK (v0.0.3) to local main branch (file:../letta-code-sdk) - Update bot.ts for new SDK API: createSession(agentId?, options) signature - Add conversationId tracking to store for proper conversation persistence - Fix Signal voice transcription: read attachments from ~/.local/share/signal-cli/attachments/ - Fix Telegram markdown ESM issue: make markdownToTelegramV2 async with dynamic import - Add transcription config to lettabot.yaml - Add extensive debug logging for queue and session processing Signal voice messages now properly transcribe and send to agent. 🐾 Generated with [Letta Code](https://letta.com) Co-Authored-By: Letta <noreply@letta.com> * fix: update Signal CLI message sender to use daemon JSON-RPC API - Switch from signal-cli-rest-api to signal-cli daemon (port 8090) - Use JSON-RPC send method instead of REST /v2/send - Support group IDs with group: prefix - Handle 201 responses and empty bodies correctly 🐾 Generated with [Letta Code](https://letta.com) Co-Authored-By: Letta <noreply@letta.com> * Add placeholder for untranscribed voice messages on Signal If a voice-only message arrives and transcription fails or is disabled, forward a placeholder so the user knows the message was received. Written by Cameron ◯ Letta Code --------- Co-authored-by: Letta <noreply@letta.com>
This commit is contained in:
Cameron
26
README.md
26
README.md
@@ -10,6 +10,7 @@ Your personal AI assistant that remembers everything across **Telegram, Slack, D
|
||||
- **Unified Memory** - Single agent remembers everything from all channels
|
||||
- **Persistent Memory** - Agent remembers conversations across sessions (days/weeks/months)
|
||||
- **Local Tool Execution** - Agent can read files, search code, run commands on your machine
|
||||
- **Voice Messages** - Automatic transcription via OpenAI Whisper
|
||||
- **Heartbeat** - Periodic check-ins where the agent reviews tasks
|
||||
- **Scheduling** - Agent can create one-off reminders and recurring tasks
|
||||
- **Streaming Responses** - Real-time message updates as the agent thinks
|
||||
@@ -97,6 +98,31 @@ That's it! Message your bot on Telegram.
|
||||
|
||||
> **Note:** For detailed environment variable reference and multi-channel setup, see [SKILL.md](./SKILL.md)
|
||||
|
||||
## Voice Messages
|
||||
|
||||
LettaBot can transcribe voice messages using OpenAI Whisper. Voice messages are automatically converted to text and sent to the agent with a `[Voice message]:` prefix.
|
||||
|
||||
**Supported channels:** Telegram, WhatsApp, Signal, Slack, Discord
|
||||
|
||||
### Configuration
|
||||
|
||||
Add your OpenAI API key to `lettabot.config.yaml`:
|
||||
|
||||
```yaml
|
||||
transcription:
|
||||
provider: openai
|
||||
apiKey: sk-...
|
||||
model: whisper-1 # optional, defaults to whisper-1
|
||||
```
|
||||
|
||||
Or set via environment variable:
|
||||
|
||||
```bash
|
||||
export OPENAI_API_KEY=sk-...
|
||||
```
|
||||
|
||||
If no API key is configured, voice messages are silently ignored.
|
||||
|
||||
## Skills
|
||||
LettaBot is compatible with [skills.sh](https://skills.sh) and [Clawdhub](https://clawdhub.com/).
|
||||
|
||||
|
||||
788
package-lock.json
generated
788
package-lock.json
generated
File diff suppressed because it is too large
Load Diff
@@ -39,7 +39,7 @@
|
||||
"@clack/prompts": "^0.11.0",
|
||||
"@hapi/boom": "^10.0.1",
|
||||
"@letta-ai/letta-client": "^1.7.6",
|
||||
"@letta-ai/letta-code-sdk": "^0.0.3",
|
||||
"@letta-ai/letta-code-sdk": "file:../letta-code-sdk",
|
||||
"@types/express": "^5.0.6",
|
||||
"@types/node": "^25.0.10",
|
||||
"@types/node-schedule": "^2.1.8",
|
||||
@@ -50,6 +50,7 @@
|
||||
"gray-matter": "^4.0.3",
|
||||
"node-schedule": "^2.1.1",
|
||||
"open": "^11.0.0",
|
||||
"openai": "^6.17.0",
|
||||
"qrcode-terminal": "^0.12.0",
|
||||
"telegram-markdown-v2": "^0.0.4",
|
||||
"tsx": "^4.21.0",
|
||||
|
||||
@@ -132,9 +132,34 @@ Ask the bot owner to approve with:
|
||||
this.client.on('messageCreate', async (message) => {
|
||||
if (message.author?.bot) return;
|
||||
|
||||
const content = (message.content || '').trim();
|
||||
let content = (message.content || '').trim();
|
||||
const userId = message.author?.id;
|
||||
if (!userId) return;
|
||||
|
||||
// Handle audio attachments
|
||||
const audioAttachment = message.attachments.find(a => a.contentType?.startsWith('audio/'));
|
||||
if (audioAttachment?.url) {
|
||||
try {
|
||||
const { loadConfig } = await import('../config/index.js');
|
||||
const config = loadConfig();
|
||||
if (!config.transcription?.apiKey && !process.env.OPENAI_API_KEY) {
|
||||
await message.reply('Voice messages require OpenAI API key for transcription. See: https://github.com/letta-ai/lettabot#voice-messages');
|
||||
} else {
|
||||
// Download audio
|
||||
const response = await fetch(audioAttachment.url);
|
||||
const buffer = Buffer.from(await response.arrayBuffer());
|
||||
|
||||
const { transcribeAudio } = await import('../transcription/index.js');
|
||||
const ext = audioAttachment.contentType?.split('/')[1] || 'mp3';
|
||||
const transcript = await transcribeAudio(buffer, audioAttachment.name || `audio.${ext}`);
|
||||
|
||||
console.log(`[Discord] Transcribed audio: "${transcript.slice(0, 50)}..."`);
|
||||
content = (content ? content + '\n' : '') + `[Voice message]: ${transcript}`;
|
||||
}
|
||||
} catch (error) {
|
||||
console.error('[Discord] Error transcribing audio:', error);
|
||||
}
|
||||
}
|
||||
|
||||
const access = await this.checkAccess(userId);
|
||||
if (access === 'blocked') {
|
||||
|
||||
@@ -46,6 +46,11 @@ type SignalSseEvent = {
|
||||
groupId?: string;
|
||||
groupName?: string;
|
||||
};
|
||||
attachments?: Array<{
|
||||
contentType?: string;
|
||||
filename?: string;
|
||||
id?: string;
|
||||
}>;
|
||||
};
|
||||
syncMessage?: {
|
||||
sentMessage?: {
|
||||
@@ -57,6 +62,11 @@ type SignalSseEvent = {
|
||||
groupId?: string;
|
||||
groupName?: string;
|
||||
};
|
||||
attachments?: Array<{
|
||||
contentType?: string;
|
||||
filename?: string;
|
||||
id?: string;
|
||||
}>;
|
||||
};
|
||||
};
|
||||
typingMessage?: {
|
||||
@@ -444,6 +454,11 @@ This code expires in 1 hour.`;
|
||||
|
||||
if (!envelope) return;
|
||||
|
||||
// Debug: log when we receive any message
|
||||
if (envelope.dataMessage || envelope.syncMessage) {
|
||||
console.log('[Signal] Received envelope:', JSON.stringify(envelope, null, 2));
|
||||
}
|
||||
|
||||
// Handle incoming data messages (from others)
|
||||
const dataMessage = envelope.dataMessage;
|
||||
|
||||
@@ -455,23 +470,26 @@ This code expires in 1 hour.`;
|
||||
let source: string | undefined;
|
||||
let chatId: string | undefined;
|
||||
let groupInfo: { groupId?: string; groupName?: string } | undefined;
|
||||
let attachments: Array<{ contentType?: string; filename?: string; id?: string }> | undefined;
|
||||
|
||||
if (dataMessage?.message) {
|
||||
if (dataMessage?.message || dataMessage?.attachments?.length) {
|
||||
// Regular incoming message
|
||||
messageText = dataMessage.message;
|
||||
source = envelope.source || envelope.sourceUuid;
|
||||
groupInfo = dataMessage.groupInfo;
|
||||
attachments = dataMessage.attachments;
|
||||
|
||||
if (groupInfo?.groupId) {
|
||||
chatId = `group:${groupInfo.groupId}`;
|
||||
} else {
|
||||
chatId = source;
|
||||
}
|
||||
} else if (syncMessage?.message) {
|
||||
} else if (syncMessage?.message || syncMessage?.attachments?.length) {
|
||||
// Sync message (Note to Self or sent from another device)
|
||||
messageText = syncMessage.message;
|
||||
source = syncMessage.destination || syncMessage.destinationUuid;
|
||||
groupInfo = syncMessage.groupInfo;
|
||||
attachments = syncMessage.attachments;
|
||||
|
||||
// For Note to Self, destination is our own number
|
||||
const isNoteToSelf = source === this.config.phoneNumber ||
|
||||
@@ -487,20 +505,73 @@ This code expires in 1 hour.`;
|
||||
}
|
||||
}
|
||||
|
||||
if (!messageText || !source || !chatId) {
|
||||
// Check if we have a valid message before attachment processing
|
||||
if (!source || !chatId) {
|
||||
return;
|
||||
}
|
||||
|
||||
// Handle voice message attachments
|
||||
const voiceAttachment = attachments?.find(a => a.contentType?.startsWith('audio/'));
|
||||
if (voiceAttachment?.id) {
|
||||
console.log(`[Signal] Voice attachment detected: ${voiceAttachment.contentType}, id: ${voiceAttachment.id}`);
|
||||
try {
|
||||
const { loadConfig } = await import('../config/index.js');
|
||||
const config = loadConfig();
|
||||
if (!config.transcription?.apiKey && !process.env.OPENAI_API_KEY) {
|
||||
if (chatId) {
|
||||
await this.sendMessage({
|
||||
chatId,
|
||||
text: 'Voice messages require OpenAI API key for transcription. See: https://github.com/letta-ai/lettabot#voice-messages'
|
||||
});
|
||||
}
|
||||
} else {
|
||||
// Read attachment from signal-cli attachments directory
|
||||
const { readFileSync } = await import('node:fs');
|
||||
const { homedir } = await import('node:os');
|
||||
const { join } = await import('node:path');
|
||||
|
||||
const attachmentPath = join(homedir(), '.local/share/signal-cli/attachments', voiceAttachment.id);
|
||||
console.log(`[Signal] Reading attachment from: ${attachmentPath}`);
|
||||
const buffer = readFileSync(attachmentPath);
|
||||
console.log(`[Signal] Read ${buffer.length} bytes`);
|
||||
|
||||
const { transcribeAudio } = await import('../transcription/index.js');
|
||||
const ext = voiceAttachment.contentType?.split('/')[1] || 'ogg';
|
||||
const transcript = await transcribeAudio(buffer, `voice.${ext}`);
|
||||
|
||||
console.log(`[Signal] Transcribed voice message: "${transcript.slice(0, 50)}..."`);
|
||||
messageText = (messageText ? messageText + '\n' : '') + `[Voice message]: ${transcript}`;
|
||||
}
|
||||
} catch (error) {
|
||||
console.error('[Signal] Error transcribing voice message:', error);
|
||||
}
|
||||
}
|
||||
|
||||
// After processing attachments, check if we have any message content.
|
||||
// If this was a voice-only message and transcription failed/was disabled,
|
||||
// still forward a placeholder so the user knows we got it.
|
||||
if (!messageText && voiceAttachment?.id) {
|
||||
messageText = '[Voice message received]';
|
||||
}
|
||||
if (!messageText) {
|
||||
return;
|
||||
}
|
||||
|
||||
// Handle Note to Self - check selfChatMode
|
||||
console.log(`[Signal] Processing message: chatId=${chatId}, source=${source}, selfChatMode=${this.config.selfChatMode}`);
|
||||
if (chatId === 'note-to-self') {
|
||||
if (!this.config.selfChatMode) {
|
||||
// selfChatMode disabled - ignore Note to Self messages
|
||||
console.log('[Signal] Note to Self ignored (selfChatMode disabled)');
|
||||
return;
|
||||
}
|
||||
// selfChatMode enabled - allow the message through
|
||||
console.log('[Signal] Note to Self allowed (selfChatMode enabled)');
|
||||
} else {
|
||||
// External message - check access control
|
||||
console.log('[Signal] Checking access for external message');
|
||||
const access = await this.checkAccess(source);
|
||||
console.log(`[Signal] Access result: ${access}`);
|
||||
|
||||
if (access === 'blocked') {
|
||||
console.log(`[Signal] Blocked message from unauthorized user: ${source}`);
|
||||
|
||||
@@ -44,16 +44,44 @@ export class SlackAdapter implements ChannelAdapter {
|
||||
});
|
||||
|
||||
// Handle messages
|
||||
this.app.message(async ({ message, say }) => {
|
||||
this.app.message(async ({ message, say, client }) => {
|
||||
// Type guard for regular messages
|
||||
if (message.subtype !== undefined) return;
|
||||
if (!('user' in message) || !('text' in message)) return;
|
||||
|
||||
const userId = message.user;
|
||||
const text = message.text || '';
|
||||
let text = message.text || '';
|
||||
const channelId = message.channel;
|
||||
const threadTs = message.thread_ts || message.ts; // Reply in thread if applicable
|
||||
|
||||
// Handle audio file attachments
|
||||
const files = (message as any).files as Array<{ mimetype?: string; url_private_download?: string; name?: string }> | undefined;
|
||||
const audioFile = files?.find(f => f.mimetype?.startsWith('audio/'));
|
||||
if (audioFile?.url_private_download) {
|
||||
try {
|
||||
const { loadConfig } = await import('../config/index.js');
|
||||
const config = loadConfig();
|
||||
if (!config.transcription?.apiKey && !process.env.OPENAI_API_KEY) {
|
||||
await say('Voice messages require OpenAI API key for transcription. See: https://github.com/letta-ai/lettabot#voice-messages');
|
||||
} else {
|
||||
// Download file (requires bot token for auth)
|
||||
const response = await fetch(audioFile.url_private_download, {
|
||||
headers: { 'Authorization': `Bearer ${this.config.botToken}` }
|
||||
});
|
||||
const buffer = Buffer.from(await response.arrayBuffer());
|
||||
|
||||
const { transcribeAudio } = await import('../transcription/index.js');
|
||||
const ext = audioFile.mimetype?.split('/')[1] || 'mp3';
|
||||
const transcript = await transcribeAudio(buffer, audioFile.name || `audio.${ext}`);
|
||||
|
||||
console.log(`[Slack] Transcribed audio: "${transcript.slice(0, 50)}..."`);
|
||||
text = (text ? text + '\n' : '') + `[Voice message]: ${transcript}`;
|
||||
}
|
||||
} catch (error) {
|
||||
console.error('[Slack] Error transcribing audio:', error);
|
||||
}
|
||||
}
|
||||
|
||||
// Check allowed users
|
||||
if (this.config.allowedUsers && this.config.allowedUsers.length > 0) {
|
||||
if (!this.config.allowedUsers.includes(userId)) {
|
||||
|
||||
@@ -5,18 +5,18 @@
|
||||
* Supports: headers, bold, italic, code, links, blockquotes, lists, etc.
|
||||
*/
|
||||
|
||||
import { convert } from 'telegram-markdown-v2';
|
||||
|
||||
/**
|
||||
* Convert markdown to Telegram MarkdownV2 format.
|
||||
* Handles proper escaping of special characters.
|
||||
*/
|
||||
export function markdownToTelegramV2(markdown: string): string {
|
||||
export async function markdownToTelegramV2(markdown: string): Promise<string> {
|
||||
try {
|
||||
// Dynamic import to avoid ESM/CommonJS compatibility issues
|
||||
const { convert } = await import('telegram-markdown-v2');
|
||||
// Use 'keep' strategy to preserve blockquotes (>) and other elements
|
||||
return convert(markdown, 'keep');
|
||||
} catch (e) {
|
||||
console.error('[Telegram] Markdown conversion failed:', e);
|
||||
console.error('[Telegram] Markdown conversion failed, using fallback:', e);
|
||||
// Fallback: escape special characters manually
|
||||
return escapeMarkdownV2(markdown);
|
||||
}
|
||||
|
||||
@@ -167,6 +167,56 @@ export class TelegramAdapter implements ChannelAdapter {
|
||||
}
|
||||
});
|
||||
|
||||
// Handle voice messages
|
||||
this.bot.on('message:voice', async (ctx) => {
|
||||
const userId = ctx.from?.id;
|
||||
const chatId = ctx.chat.id;
|
||||
|
||||
if (!userId) return;
|
||||
|
||||
// Check if transcription is configured (config or env)
|
||||
const { loadConfig } = await import('../config/index.js');
|
||||
const config = loadConfig();
|
||||
if (!config.transcription?.apiKey && !process.env.OPENAI_API_KEY) {
|
||||
await ctx.reply('Voice messages require OpenAI API key for transcription. See: https://github.com/letta-ai/lettabot#voice-messages');
|
||||
return;
|
||||
}
|
||||
|
||||
try {
|
||||
// Get file link
|
||||
const voice = ctx.message.voice;
|
||||
const file = await ctx.api.getFile(voice.file_id);
|
||||
const fileUrl = `https://api.telegram.org/file/bot${this.config.token}/${file.file_path}`;
|
||||
|
||||
// Download audio
|
||||
const response = await fetch(fileUrl);
|
||||
const buffer = Buffer.from(await response.arrayBuffer());
|
||||
|
||||
// Transcribe
|
||||
const { transcribeAudio } = await import('../transcription/index.js');
|
||||
const transcript = await transcribeAudio(buffer, 'voice.ogg');
|
||||
|
||||
console.log(`[Telegram] Transcribed voice message: "${transcript.slice(0, 50)}..."`);
|
||||
|
||||
// Send to agent as text with prefix
|
||||
if (this.onMessage) {
|
||||
await this.onMessage({
|
||||
channel: 'telegram',
|
||||
chatId: String(chatId),
|
||||
userId: String(userId),
|
||||
userName: ctx.from.username || ctx.from.first_name,
|
||||
messageId: String(ctx.message.message_id),
|
||||
text: `[Voice message]: ${transcript}`,
|
||||
timestamp: new Date(),
|
||||
});
|
||||
}
|
||||
} catch (error) {
|
||||
console.error('[Telegram] Error processing voice message:', error);
|
||||
// Optionally notify user
|
||||
await ctx.reply('Sorry, I could not transcribe that voice message.');
|
||||
}
|
||||
});
|
||||
|
||||
// Error handler
|
||||
this.bot.catch((err) => {
|
||||
console.error('[Telegram] Bot error:', err);
|
||||
@@ -199,7 +249,7 @@ export class TelegramAdapter implements ChannelAdapter {
|
||||
const { markdownToTelegramV2 } = await import('./telegram-format.js');
|
||||
|
||||
// Convert markdown to Telegram MarkdownV2 format
|
||||
const formatted = markdownToTelegramV2(msg.text);
|
||||
const formatted = await markdownToTelegramV2(msg.text);
|
||||
|
||||
const result = await this.bot.api.sendMessage(msg.chatId, formatted, {
|
||||
parse_mode: 'MarkdownV2',
|
||||
@@ -210,7 +260,7 @@ export class TelegramAdapter implements ChannelAdapter {
|
||||
|
||||
async editMessage(chatId: string, messageId: string, text: string): Promise<void> {
|
||||
const { markdownToTelegramV2 } = await import('./telegram-format.js');
|
||||
const formatted = markdownToTelegramV2(text);
|
||||
const formatted = await markdownToTelegramV2(text);
|
||||
await this.bot.api.editMessageText(chatId, Number(messageId), formatted, { parse_mode: 'MarkdownV2' });
|
||||
}
|
||||
|
||||
|
||||
@@ -142,6 +142,7 @@ Ask the bot owner to approve with:
|
||||
DisconnectReason,
|
||||
fetchLatestBaileysVersion,
|
||||
makeCacheableSignalKeyStore,
|
||||
downloadMediaMessage,
|
||||
} = await import('@whiskeysockets/baileys');
|
||||
|
||||
// Load auth state
|
||||
@@ -253,10 +254,38 @@ Ask the bot owner to approve with:
|
||||
this.lidToJid.set(remoteJid, (m.key as any).senderPn);
|
||||
}
|
||||
|
||||
// Get message text
|
||||
const text = m.message?.conversation ||
|
||||
m.message?.extendedTextMessage?.text ||
|
||||
'';
|
||||
// Get message text or audio
|
||||
let text = m.message?.conversation ||
|
||||
m.message?.extendedTextMessage?.text ||
|
||||
'';
|
||||
|
||||
// Handle audio/voice messages
|
||||
const audioMessage = m.message?.audioMessage;
|
||||
if (audioMessage) {
|
||||
try {
|
||||
const { loadConfig } = await import('../config/index.js');
|
||||
const config = loadConfig();
|
||||
if (!config.transcription?.apiKey && !process.env.OPENAI_API_KEY) {
|
||||
await this.sock!.sendMessage(remoteJid, {
|
||||
text: 'Voice messages require OpenAI API key for transcription. See: https://github.com/letta-ai/lettabot#voice-messages'
|
||||
});
|
||||
continue;
|
||||
}
|
||||
|
||||
// Download audio
|
||||
const buffer = await downloadMediaMessage(m, 'buffer', {});
|
||||
|
||||
// Transcribe
|
||||
const { transcribeAudio } = await import('../transcription/index.js');
|
||||
const transcript = await transcribeAudio(buffer as Buffer, 'voice.ogg');
|
||||
|
||||
console.log(`[WhatsApp] Transcribed voice message: "${transcript.slice(0, 50)}..."`);
|
||||
text = `[Voice message]: ${transcript}`;
|
||||
} catch (error) {
|
||||
console.error('[WhatsApp] Error transcribing voice message:', error);
|
||||
continue;
|
||||
}
|
||||
}
|
||||
|
||||
if (!text) continue;
|
||||
|
||||
|
||||
@@ -100,28 +100,61 @@ async function sendSlack(chatId: string, text: string): Promise<void> {
|
||||
}
|
||||
|
||||
async function sendSignal(chatId: string, text: string): Promise<void> {
|
||||
const apiUrl = process.env.SIGNAL_CLI_REST_API_URL || 'http://localhost:8080';
|
||||
// We talk to the signal-cli daemon JSON-RPC API (the same daemon the Signal adapter uses).
|
||||
// This is *not* the signal-cli-rest-api container.
|
||||
const apiUrl = process.env.SIGNAL_CLI_REST_API_URL || 'http://127.0.0.1:8090';
|
||||
const phoneNumber = process.env.SIGNAL_PHONE_NUMBER;
|
||||
|
||||
|
||||
if (!phoneNumber) {
|
||||
throw new Error('SIGNAL_PHONE_NUMBER not set');
|
||||
}
|
||||
|
||||
const response = await fetch(`${apiUrl}/v2/send`, {
|
||||
|
||||
// Support group IDs in the same format we use everywhere else.
|
||||
const params: Record<string, unknown> = {
|
||||
account: phoneNumber,
|
||||
message: text,
|
||||
};
|
||||
|
||||
if (chatId.startsWith('group:')) {
|
||||
params.groupId = chatId.slice('group:'.length);
|
||||
} else {
|
||||
params.recipient = [chatId];
|
||||
}
|
||||
|
||||
const body = JSON.stringify({
|
||||
jsonrpc: '2.0',
|
||||
method: 'send',
|
||||
params,
|
||||
id: Date.now(),
|
||||
});
|
||||
|
||||
const response = await fetch(`${apiUrl}/api/v1/rpc`, {
|
||||
method: 'POST',
|
||||
headers: { 'Content-Type': 'application/json' },
|
||||
body: JSON.stringify({
|
||||
message: text,
|
||||
number: phoneNumber,
|
||||
recipients: [chatId],
|
||||
}),
|
||||
body,
|
||||
});
|
||||
|
||||
if (!response.ok) {
|
||||
const error = await response.text();
|
||||
throw new Error(`Signal API error: ${error}`);
|
||||
|
||||
// signal-cli returns status 201 with empty body sometimes.
|
||||
if (response.status === 201) {
|
||||
console.log(`✓ Sent to signal:${chatId}`);
|
||||
return;
|
||||
}
|
||||
|
||||
|
||||
const textBody = await response.text();
|
||||
if (!response.ok) {
|
||||
throw new Error(`Signal API error: ${textBody}`);
|
||||
}
|
||||
|
||||
if (!textBody.trim()) {
|
||||
console.log(`✓ Sent to signal:${chatId}`);
|
||||
return;
|
||||
}
|
||||
|
||||
const parsed = JSON.parse(textBody) as { result?: unknown; error?: { code?: number; message?: string } };
|
||||
if (parsed.error) {
|
||||
throw new Error(`Signal RPC ${parsed.error.code ?? 'unknown'}: ${parsed.error.message ?? 'unknown error'}`);
|
||||
}
|
||||
|
||||
console.log(`✓ Sent to signal:${chatId}`);
|
||||
}
|
||||
|
||||
@@ -257,7 +290,7 @@ Environment variables:
|
||||
SLACK_BOT_TOKEN Required for Slack
|
||||
DISCORD_BOT_TOKEN Required for Discord
|
||||
SIGNAL_PHONE_NUMBER Required for Signal
|
||||
SIGNAL_CLI_REST_API_URL Signal API URL (default: http://localhost:8080)
|
||||
SIGNAL_CLI_REST_API_URL Signal daemon URL (default: http://127.0.0.1:8090)
|
||||
`);
|
||||
}
|
||||
|
||||
|
||||
@@ -49,6 +49,15 @@ export interface LettaBotConfig {
|
||||
integrations?: {
|
||||
google?: GoogleConfig;
|
||||
};
|
||||
|
||||
// Transcription (voice messages)
|
||||
transcription?: TranscriptionConfig;
|
||||
}
|
||||
|
||||
export interface TranscriptionConfig {
|
||||
provider: 'openai'; // Only OpenAI supported currently
|
||||
apiKey?: string; // Falls back to OPENAI_API_KEY env var
|
||||
model?: string; // Defaults to 'whisper-1'
|
||||
}
|
||||
|
||||
export interface ProviderConfig {
|
||||
|
||||
@@ -121,10 +121,14 @@ export class LettaBot {
|
||||
|
||||
// Add to queue
|
||||
this.messageQueue.push({ msg, adapter });
|
||||
console.log(`[Queue] Added to queue, length: ${this.messageQueue.length}, processing: ${this.processing}`);
|
||||
|
||||
// Process queue if not already processing
|
||||
if (!this.processing) {
|
||||
this.processQueue();
|
||||
console.log('[Queue] Starting queue processing');
|
||||
this.processQueue().catch(err => console.error('[Queue] Fatal error in processQueue:', err));
|
||||
} else {
|
||||
console.log('[Queue] Already processing, will process when current message finishes');
|
||||
}
|
||||
}
|
||||
|
||||
@@ -132,12 +136,18 @@ export class LettaBot {
|
||||
* Process messages one at a time
|
||||
*/
|
||||
private async processQueue(): Promise<void> {
|
||||
if (this.processing || this.messageQueue.length === 0) return;
|
||||
console.log(`[Queue] processQueue called: processing=${this.processing}, queueLength=${this.messageQueue.length}`);
|
||||
if (this.processing || this.messageQueue.length === 0) {
|
||||
console.log('[Queue] Exiting early: already processing or empty queue');
|
||||
return;
|
||||
}
|
||||
|
||||
this.processing = true;
|
||||
console.log('[Queue] Started processing');
|
||||
|
||||
while (this.messageQueue.length > 0) {
|
||||
const { msg, adapter } = this.messageQueue.shift()!;
|
||||
console.log(`[Queue] Processing message from ${msg.userId} (${this.messageQueue.length} remaining)`);
|
||||
try {
|
||||
await this.processMessage(msg, adapter);
|
||||
} catch (error) {
|
||||
@@ -145,6 +155,7 @@ export class LettaBot {
|
||||
}
|
||||
}
|
||||
|
||||
console.log('[Queue] Finished processing all messages');
|
||||
this.processing = false;
|
||||
}
|
||||
|
||||
@@ -152,6 +163,7 @@ export class LettaBot {
|
||||
* Process a single message
|
||||
*/
|
||||
private async processMessage(msg: InboundMessage, adapter: ChannelAdapter): Promise<void> {
|
||||
console.log('[Bot] Starting processMessage');
|
||||
// Track when user last sent a message (for heartbeat skip logic)
|
||||
this.lastUserMessageTime = new Date();
|
||||
|
||||
@@ -163,32 +175,39 @@ export class LettaBot {
|
||||
updatedAt: new Date().toISOString(),
|
||||
};
|
||||
|
||||
console.log('[Bot] Sending typing indicator');
|
||||
// Start typing indicator
|
||||
await adapter.sendTypingIndicator(msg.chatId);
|
||||
console.log('[Bot] Typing indicator sent');
|
||||
|
||||
// Create or resume session
|
||||
let session: Session;
|
||||
// Base options for all sessions (model only included for new agents)
|
||||
// Note: canUseTool workaround for SDK v0.0.3 bug - can be removed after letta-ai/letta-code-sdk#10 is released
|
||||
const baseOptions = {
|
||||
permissionMode: 'bypassPermissions' as const,
|
||||
allowedTools: this.config.allowedTools,
|
||||
cwd: this.config.workingDir,
|
||||
systemPrompt: SYSTEM_PROMPT,
|
||||
canUseTool: () => ({ allow: true }),
|
||||
};
|
||||
|
||||
console.log('[Bot] Creating/resuming session');
|
||||
try {
|
||||
if (this.store.agentId) {
|
||||
if (this.store.conversationId) {
|
||||
// Resume the specific conversation we've been using
|
||||
console.log(`[Bot] Resuming conversation: ${this.store.conversationId}`);
|
||||
process.env.LETTA_AGENT_ID = this.store.agentId || undefined;
|
||||
session = resumeSession(this.store.conversationId, baseOptions);
|
||||
} else if (this.store.agentId) {
|
||||
// Agent exists but no conversation - try default conversation
|
||||
console.log(`[Bot] Resuming agent default conversation: ${this.store.agentId}`);
|
||||
process.env.LETTA_AGENT_ID = this.store.agentId;
|
||||
|
||||
// Don't pass model when resuming - agent already has its model configured
|
||||
session = resumeSession(this.store.agentId, baseOptions);
|
||||
} else {
|
||||
|
||||
// Only pass model when creating a new agent
|
||||
session = createSession({ ...baseOptions, model: this.config.model, memory: loadMemoryBlocks(this.config.agentName) });
|
||||
// Create new agent with default conversation
|
||||
console.log('[Bot] Creating new agent');
|
||||
session = createSession(undefined, { ...baseOptions, model: this.config.model, memory: loadMemoryBlocks(this.config.agentName) });
|
||||
}
|
||||
console.log('[Bot] Session created/resumed');
|
||||
|
||||
const initTimeoutMs = 30000; // 30s timeout
|
||||
const withTimeout = async <T>(promise: Promise<T>, label: string): Promise<T> => {
|
||||
@@ -301,21 +320,26 @@ export class LettaBot {
|
||||
}
|
||||
|
||||
if (streamMsg.type === 'result') {
|
||||
// Save agent ID and attach ignore tool (only on first message)
|
||||
// Save agent ID and conversation ID
|
||||
if (session.agentId && session.agentId !== this.store.agentId) {
|
||||
const isNewAgent = !this.store.agentId;
|
||||
// Save agent ID along with the current server URL
|
||||
const currentBaseUrl = process.env.LETTA_BASE_URL || 'https://api.letta.com';
|
||||
this.store.setAgent(session.agentId, currentBaseUrl);
|
||||
console.log('Saved agent ID:', session.agentId, 'on server:', currentBaseUrl);
|
||||
this.store.setAgent(session.agentId, currentBaseUrl, session.conversationId || undefined);
|
||||
console.log('Saved agent ID:', session.agentId, 'conversation ID:', session.conversationId, 'on server:', currentBaseUrl);
|
||||
|
||||
// Setup new agents: set name, install skills
|
||||
if (isNewAgent) {
|
||||
if (this.config.agentName) {
|
||||
if (this.config.agentName && session.agentId) {
|
||||
updateAgentName(session.agentId, this.config.agentName).catch(() => {});
|
||||
}
|
||||
installSkillsToAgent(session.agentId);
|
||||
if (session.agentId) {
|
||||
installSkillsToAgent(session.agentId);
|
||||
}
|
||||
}
|
||||
} else if (session.conversationId && session.conversationId !== this.store.conversationId) {
|
||||
// Update conversation ID if it changed
|
||||
this.store.conversationId = session.conversationId;
|
||||
}
|
||||
break;
|
||||
}
|
||||
@@ -376,22 +400,23 @@ export class LettaBot {
|
||||
_context?: TriggerContext
|
||||
): Promise<string> {
|
||||
// Base options (model only for new agents)
|
||||
// Note: canUseTool workaround for SDK v0.0.3 bug - can be removed after letta-ai/letta-code-sdk#10 is released
|
||||
const baseOptions = {
|
||||
permissionMode: 'bypassPermissions' as const,
|
||||
allowedTools: this.config.allowedTools,
|
||||
cwd: this.config.workingDir,
|
||||
systemPrompt: SYSTEM_PROMPT,
|
||||
canUseTool: () => ({ allow: true }),
|
||||
};
|
||||
|
||||
let session: Session;
|
||||
if (this.store.agentId) {
|
||||
// Don't pass model when resuming - agent already has its model configured
|
||||
if (this.store.conversationId) {
|
||||
// Resume the specific conversation we've been using
|
||||
session = resumeSession(this.store.conversationId, baseOptions);
|
||||
} else if (this.store.agentId) {
|
||||
// Agent exists but no conversation - try default conversation
|
||||
session = resumeSession(this.store.agentId, baseOptions);
|
||||
} else {
|
||||
// Only pass model when creating a new agent
|
||||
session = createSession({ ...baseOptions, model: this.config.model, memory: loadMemoryBlocks(this.config.agentName) });
|
||||
// Create new agent with default conversation
|
||||
session = createSession(undefined, { ...baseOptions, model: this.config.model, memory: loadMemoryBlocks(this.config.agentName) });
|
||||
}
|
||||
|
||||
try {
|
||||
@@ -406,7 +431,9 @@ export class LettaBot {
|
||||
if (msg.type === 'result') {
|
||||
if (session.agentId && session.agentId !== this.store.agentId) {
|
||||
const currentBaseUrl = process.env.LETTA_BASE_URL || 'https://api.letta.com';
|
||||
this.store.setAgent(session.agentId, currentBaseUrl);
|
||||
this.store.setAgent(session.agentId, currentBaseUrl, session.conversationId || undefined);
|
||||
} else if (session.conversationId && session.conversationId !== this.store.conversationId) {
|
||||
this.store.conversationId = session.conversationId;
|
||||
}
|
||||
break;
|
||||
}
|
||||
|
||||
@@ -53,6 +53,15 @@ export class Store {
|
||||
this.save();
|
||||
}
|
||||
|
||||
get conversationId(): string | null {
|
||||
return this.data.conversationId || null;
|
||||
}
|
||||
|
||||
set conversationId(id: string | null) {
|
||||
this.data.conversationId = id;
|
||||
this.save();
|
||||
}
|
||||
|
||||
get baseUrl(): string | undefined {
|
||||
return this.data.baseUrl;
|
||||
}
|
||||
@@ -65,9 +74,10 @@ export class Store {
|
||||
/**
|
||||
* Set agent ID and associated server URL together
|
||||
*/
|
||||
setAgent(id: string | null, baseUrl?: string): void {
|
||||
setAgent(id: string | null, baseUrl?: string, conversationId?: string): void {
|
||||
this.data.agentId = id;
|
||||
this.data.baseUrl = baseUrl;
|
||||
this.data.conversationId = conversationId || this.data.conversationId;
|
||||
this.data.lastUsedAt = new Date().toISOString();
|
||||
if (id && !this.data.createdAt) {
|
||||
this.data.createdAt = new Date().toISOString();
|
||||
|
||||
@@ -101,6 +101,7 @@ export interface LastMessageTarget {
|
||||
*/
|
||||
export interface AgentStore {
|
||||
agentId: string | null;
|
||||
conversationId?: string | null; // Current conversation ID
|
||||
baseUrl?: string; // Server URL this agent belongs to
|
||||
createdAt?: string;
|
||||
lastUsedAt?: string;
|
||||
|
||||
7
src/transcription/index.ts
Normal file
7
src/transcription/index.ts
Normal file
@@ -0,0 +1,7 @@
|
||||
/**
|
||||
* Transcription service
|
||||
*
|
||||
* Currently supports OpenAI Whisper. Future providers can be added here.
|
||||
*/
|
||||
|
||||
export { transcribeAudio } from './openai.js';
|
||||
69
src/transcription/openai.ts
Normal file
69
src/transcription/openai.ts
Normal file
@@ -0,0 +1,69 @@
|
||||
/**
|
||||
* OpenAI Whisper transcription service
|
||||
*/
|
||||
|
||||
import OpenAI from 'openai';
|
||||
import { loadConfig } from '../config/index.js';
|
||||
|
||||
let openaiClient: OpenAI | null = null;
|
||||
|
||||
function getClient(): OpenAI {
|
||||
if (!openaiClient) {
|
||||
const config = loadConfig();
|
||||
// Config takes priority, then env var
|
||||
const apiKey = config.transcription?.apiKey || process.env.OPENAI_API_KEY;
|
||||
if (!apiKey) {
|
||||
throw new Error('OpenAI API key required for transcription. Set in config (transcription.apiKey) or OPENAI_API_KEY env var.');
|
||||
}
|
||||
openaiClient = new OpenAI({ apiKey });
|
||||
}
|
||||
return openaiClient;
|
||||
}
|
||||
|
||||
function getModel(): string {
|
||||
const config = loadConfig();
|
||||
return config.transcription?.model || process.env.TRANSCRIPTION_MODEL || 'whisper-1';
|
||||
}
|
||||
|
||||
/**
|
||||
* Transcribe audio using OpenAI Whisper API
|
||||
*
|
||||
* @param audioBuffer - The audio data as a Buffer
|
||||
* @param filename - Filename with extension (e.g., 'voice.ogg')
|
||||
* @returns The transcribed text
|
||||
*/
|
||||
export async function transcribeAudio(audioBuffer: Buffer, filename: string = 'audio.ogg'): Promise<string> {
|
||||
const client = getClient();
|
||||
|
||||
// Create a File object from the buffer
|
||||
// OpenAI SDK expects a File-like object
|
||||
// Convert Buffer to Uint8Array to satisfy BlobPart type
|
||||
const file = new File([new Uint8Array(audioBuffer)], filename, {
|
||||
type: getMimeType(filename)
|
||||
});
|
||||
|
||||
const response = await client.audio.transcriptions.create({
|
||||
file,
|
||||
model: getModel(),
|
||||
});
|
||||
|
||||
return response.text;
|
||||
}
|
||||
|
||||
/**
|
||||
* Get MIME type from filename extension
|
||||
*/
|
||||
function getMimeType(filename: string): string {
|
||||
const ext = filename.split('.').pop()?.toLowerCase();
|
||||
const mimeTypes: Record<string, string> = {
|
||||
'ogg': 'audio/ogg',
|
||||
'mp3': 'audio/mpeg',
|
||||
'mp4': 'audio/mp4',
|
||||
'm4a': 'audio/mp4',
|
||||
'wav': 'audio/wav',
|
||||
'webm': 'audio/webm',
|
||||
'mpeg': 'audio/mpeg',
|
||||
'mpga': 'audio/mpeg',
|
||||
};
|
||||
return mimeTypes[ext || ''] || 'audio/ogg';
|
||||
}
|
||||
Reference in New Issue
Block a user