kimaki 0.1.3 → 0.1.4

package/dist/cli.js

@@ -129,7 +129,7 @@ async function run({ restart, addChannels }) {
             '2. Click "Reset Token" to generate a new bot token (in case of errors try again)\n' +
             "3. Copy the token (you won't be able to see it again!)", 'Step 3: Get Bot Token');
         const tokenInput = await password({
-            message: 'Enter your Discord Bot Token (will be hidden):',
+            message: 'Enter your Discord Bot Token (from "Bot" section - click "Reset Token" if needed):',
             validate(value) {
                 if (!value)
                     return 'Bot token is required';
@@ -142,6 +142,24 @@ async function run({ restart, addChannels }) {
             process.exit(0);
         }
         token = tokenInput;
+        note(`You can get a Gemini api Key at https://aistudio.google.com/apikey`, `Gemini API Key`);
+        const geminiApiKey = await password({
+            message: 'Enter your Gemini API Key for voice channels and audio transcription (optional, press Enter to skip):',
+            validate(value) {
+                if (value && value.length < 10)
+                    return 'Invalid API key format';
+                return undefined;
+            },
+        });
+        if (isCancel(geminiApiKey)) {
+            cancel('Setup cancelled');
+            process.exit(0);
+        }
+        // Store API key in database
+        if (geminiApiKey) {
+            db.prepare('INSERT OR REPLACE INTO bot_api_keys (app_id, gemini_api_key) VALUES (?, ?)').run(appId, geminiApiKey || null);
+            note('API key saved successfully', 'API Key Stored');
+        }
         note(`Bot install URL:\n${generateBotInstallUrl({ clientId: appId })}\n\nYou MUST install the bot in your Discord server before continuing.`, 'Step 4: Install Bot to Server');
         const installed = await text({
             message: 'Press Enter AFTER you have installed the bot in your server:',

package/dist/discordBot.js

@@ -78,7 +78,7 @@ async function createUserAudioLogStream(guildId, channelId) {
     }
 }
 // Set up voice handling for a connection (called once per connection)
-async function setupVoiceHandling({ connection, guildId, channelId, }) {
+async function setupVoiceHandling({ connection, guildId, channelId, appId, }) {
     voiceLogger.log(`Setting up voice handling for guild ${guildId}, channel ${channelId}`);
     // Check if this voice channel has an associated directory
     const channelDirRow = getDatabase()
@@ -98,11 +98,17 @@ async function setupVoiceHandling({ connection, guildId, channelId, }) {
     }
     // Create user audio stream for debugging
     voiceData.userAudioStream = await createUserAudioLogStream(guildId, channelId);
+    // Get API keys from database
+    const apiKeys = getDatabase()
+        .prepare('SELECT gemini_api_key FROM bot_api_keys WHERE app_id = ?')
+        .get(appId);
     // Create GenAI worker
     const genAiWorker = await createGenAIWorker({
         directory,
         guildId,
         channelId,
+        appId,
+        geminiApiKey: apiKeys?.gemini_api_key,
         systemMessage: dedent `
     You are Kimaki, an AI similar to Jarvis: you help your user (an engineer) controlling his coding agent, just like Jarvis controls Ironman armor and machines. Speak fast.
 
@@ -347,6 +353,13 @@ export function getDatabase() {
         channel_type TEXT NOT NULL,
         created_at DATETIME DEFAULT CURRENT_TIMESTAMP
       )
+    `);
+        db.exec(`
+      CREATE TABLE IF NOT EXISTS bot_api_keys (
+        app_id TEXT PRIMARY KEY,
+        gemini_api_key TEXT,
+        created_at DATETIME DEFAULT CURRENT_TIMESTAMP
+      )
     `);
     }
     return db;
@@ -458,7 +471,7 @@ async function waitForServer(port, maxAttempts = 30) {
     }
     throw new Error(`Server did not start on port ${port} after ${maxAttempts} seconds`);
 }
-async function processVoiceAttachment({ message, thread, projectDirectory, isNewThread = false, }) {
+async function processVoiceAttachment({ message, thread, projectDirectory, isNewThread = false, appId, }) {
     const audioAttachment = Array.from(message.attachments.values()).find((attachment) => attachment.contentType?.startsWith('audio/'));
     if (!audioAttachment)
         return null;
@@ -488,9 +501,20 @@ async function processVoiceAttachment({ message, thread, projectDirectory, isNew
             voiceLogger.log(`Could not get project tree:`, e);
         }
     }
+    // Get Gemini API key from database if appId is provided
+    let geminiApiKey;
+    if (appId) {
+        const apiKeys = getDatabase()
+            .prepare('SELECT gemini_api_key FROM bot_api_keys WHERE app_id = ?')
+            .get(appId);
+        if (apiKeys?.gemini_api_key) {
+            geminiApiKey = apiKeys.gemini_api_key;
+        }
+    }
     const transcription = await transcribeAudio({
         audio: audioBuffer,
         prompt: transcriptionPrompt,
+        geminiApiKey,
     });
     voiceLogger.log(`Transcription successful: "${transcription.slice(0, 50)}${transcription.length > 50 ? '...' : ''}"`);
     // Update thread name with transcribed content only for new threads
@@ -1233,6 +1257,7 @@ export async function startDiscordBot({ token, appId, discordClient, }) {
                     message,
                     thread,
                     projectDirectory,
+                    appId: currentAppId,
                 });
                 if (transcription) {
                     messageContent = transcription;
@@ -1291,6 +1316,7 @@ export async function startDiscordBot({ token, appId, discordClient, }) {
                     thread,
                     projectDirectory,
                     isNewThread: true,
+                    appId: currentAppId,
                 });
                 if (transcription) {
                     messageContent = transcription;
@@ -1651,6 +1677,7 @@ export async function startDiscordBot({ token, appId, discordClient, }) {
                     connection,
                     guildId: newState.guild.id,
                     channelId: voiceChannel.id,
+                    appId: currentAppId,
                 });
                 // Handle connection state changes
                 connection.on(VoiceConnectionStatus.Disconnected, async () => {

package/dist/genai-worker-wrapper.js

@@ -98,6 +98,8 @@ export function createGenAIWorker(options) {
             systemMessage: options.systemMessage,
             guildId: options.guildId,
             channelId: options.channelId,
+            appId: options.appId,
+            geminiApiKey: options.geminiApiKey,
         };
         worker.postMessage(initMessage);
     });

package/dist/genai-worker.js

@@ -210,6 +210,7 @@ parentPort.on('message', async (message) => {
                 session = await startGenAiSession({
                     tools,
                     systemMessage: message.systemMessage,
+                    geminiApiKey: message.geminiApiKey,
                     onAssistantAudioChunk({ data }) {
                         // Write to audio log if enabled
                         if (audioLogStream && !audioLogStream.destroyed) {

package/dist/genai.js

@@ -68,7 +68,7 @@ function defaultAudioChunkHandler({ data, mimeType, }) {
     const buffer = convertToWav(audioParts, mimeType);
     saveBinaryFile(fileName, buffer);
 }
-export async function startGenAiSession({ onAssistantAudioChunk, onAssistantStartSpeaking, onAssistantStopSpeaking, onAssistantInterruptSpeaking, systemMessage, tools, } = {}) {
+export async function startGenAiSession({ onAssistantAudioChunk, onAssistantStartSpeaking, onAssistantStopSpeaking, onAssistantInterruptSpeaking, systemMessage, tools, geminiApiKey, } = {}) {
     let session = undefined;
     const callableTools = [];
     let isAssistantSpeaking = false;
@@ -161,8 +161,13 @@ export async function startGenAiSession({ onAssistantAudioChunk, onAssistantStar
             }
         }
     }
+    const apiKey = geminiApiKey || process.env.GEMINI_API_KEY;
+    if (!apiKey) {
+        genaiLogger.error('No Gemini API key provided');
+        throw new Error('Gemini API key is required for voice interactions');
+    }
     const ai = new GoogleGenAI({
-        apiKey: process.env.GEMINI_API_KEY,
+        apiKey,
     });
     const model = 'models/gemini-2.5-flash-live-preview';
     session = await ai.live.connect({

package/dist/voice.js

@@ -1,25 +1,60 @@
-import { openai } from '@ai-sdk/openai';
-import { experimental_transcribe as transcribe } from 'ai';
+import { GoogleGenAI } from '@google/genai';
 import { createLogger } from './logger.js';
 const voiceLogger = createLogger('VOICE');
-export async function transcribeAudio({ audio, prompt, language, temperature, }) {
+export async function transcribeAudio({ audio, prompt, language, temperature, geminiApiKey, }) {
     try {
-        const result = await transcribe({
-            model: openai.transcription('whisper-1'),
-            audio,
-            ...(prompt || language || temperature !== undefined
-                ? {
-                    providerOptions: {
-                        openai: {
-                            ...(prompt && { prompt }),
-                            ...(language && { language }),
-                            ...(temperature !== undefined && { temperature }),
+        // Use provided API key or fall back to environment variable
+        const apiKey = geminiApiKey || process.env.GEMINI_API_KEY;
+        if (!apiKey) {
+            throw new Error('Gemini API key is required for audio transcription');
+        }
+        // Initialize Google Generative AI
+        const genAI = new GoogleGenAI({ apiKey });
+        // Convert audio to base64 string if it's not already
+        let audioBase64;
+        if (typeof audio === 'string') {
+            audioBase64 = audio;
+        }
+        else if (audio instanceof Buffer) {
+            audioBase64 = audio.toString('base64');
+        }
+        else if (audio instanceof Uint8Array) {
+            audioBase64 = Buffer.from(audio).toString('base64');
+        }
+        else if (audio instanceof ArrayBuffer) {
+            audioBase64 = Buffer.from(audio).toString('base64');
+        }
+        else {
+            throw new Error('Invalid audio format');
+        }
+        // Build the transcription prompt
+        let transcriptionPrompt = `Please transcribe this audio file accurately. Here is some relevant information and filenames that may be present in the audio:\n<context>\n${prompt}\n</context>\n`;
+        if (language) {
+            transcriptionPrompt += `\nThe audio is in ${language}.`;
+        }
+        // Create the content with audio using the inline data format
+        const response = await genAI.models.generateContent({
+            model: 'gemini-2.5-flash',
+            contents: [
+                {
+                    parts: [
+                        { text: transcriptionPrompt },
+                        {
+                            inlineData: {
+                                data: audioBase64,
+                                mimeType: 'audio/mpeg',
+                            },
                         },
-                    },
+                    ],
+                },
+            ],
+            config: temperature !== undefined
+                ? {
+                    temperature,
                 }
-                : {}),
+                : undefined,
         });
-        return result.text;
+        return response.text || '';
     }
     catch (error) {
         voiceLogger.error('Failed to transcribe audio:', error);

package/package.json

@@ -2,7 +2,7 @@
   "name": "kimaki",
   "module": "index.ts",
   "type": "module",
-  "version": "0.1.3",
+  "version": "0.1.4",
   "repository": "https://github.com/remorses/kimaki",
   "bin": "bin.js",
   "files": [
@@ -19,7 +19,7 @@
     "tsx": "^4.20.5"
   },
   "dependencies": {
-    "@ai-sdk/openai": "^2.0.23",
+    "@ai-sdk/google": "^2.0.16",
     "@clack/prompts": "^0.11.0",
     "@discordjs/opus": "^0.10.0",
     "@discordjs/voice": "^0.19.0",

package/src/cli.ts

@@ -223,9 +223,9 @@ async function run({ restart, addChannels }: CliOptions) {
         "3. Copy the token (you won't be able to see it again!)",
       'Step 3: Get Bot Token',
     )
-
     const tokenInput = await password({
-      message: 'Enter your Discord Bot Token (will be hidden):',
+      message:
+        'Enter your Discord Bot Token (from "Bot" section - click "Reset Token" if needed):',
       validate(value) {
         if (!value) return 'Bot token is required'
         if (value.length < 50) return 'Invalid token format (too short)'
@@ -238,6 +238,33 @@ async function run({ restart, addChannels }: CliOptions) {
     }
     token = tokenInput
 
+    note(
+      `You can get a Gemini api Key at https://aistudio.google.com/apikey`,
+      `Gemini API Key`,
+    )
+
+    const geminiApiKey = await password({
+      message:
+        'Enter your Gemini API Key for voice channels and audio transcription (optional, press Enter to skip):',
+      validate(value) {
+        if (value && value.length < 10) return 'Invalid API key format'
+        return undefined
+      },
+    })
+
+    if (isCancel(geminiApiKey)) {
+      cancel('Setup cancelled')
+      process.exit(0)
+    }
+
+    // Store API key in database
+    if (geminiApiKey) {
+      db.prepare(
+        'INSERT OR REPLACE INTO bot_api_keys (app_id, gemini_api_key) VALUES (?, ?)',
+      ).run(appId, geminiApiKey || null)
+      note('API key saved successfully', 'API Key Stored')
+    }
+
     note(
       `Bot install URL:\n${generateBotInstallUrl({ clientId: appId })}\n\nYou MUST install the bot in your Discord server before continuing.`,
       'Step 4: Install Bot to Server',

package/src/discordBot.ts

@@ -152,10 +152,12 @@ async function setupVoiceHandling({
   connection,
   guildId,
   channelId,
+  appId,
 }: {
   connection: VoiceConnection
   guildId: string
   channelId: string
+  appId: string
 }) {
   voiceLogger.log(
     `Setting up voice handling for guild ${guildId}, channel ${channelId}`,
@@ -188,11 +190,18 @@ async function setupVoiceHandling({
   // Create user audio stream for debugging
   voiceData.userAudioStream = await createUserAudioLogStream(guildId, channelId)
 
+  // Get API keys from database
+  const apiKeys = getDatabase()
+    .prepare('SELECT gemini_api_key FROM bot_api_keys WHERE app_id = ?')
+    .get(appId) as { gemini_api_key: string | null } | undefined
+
   // Create GenAI worker
   const genAiWorker = await createGenAIWorker({
     directory,
     guildId,
     channelId,
+    appId,
+    geminiApiKey: apiKeys?.gemini_api_key,
     systemMessage: dedent`
     You are Kimaki, an AI similar to Jarvis: you help your user (an engineer) controlling his coding agent, just like Jarvis controls Ironman armor and machines. Speak fast.
 
@@ -480,6 +489,14 @@ export function getDatabase(): Database.Database {
         created_at DATETIME DEFAULT CURRENT_TIMESTAMP
       )
     `)
+
+    db.exec(`
+      CREATE TABLE IF NOT EXISTS bot_api_keys (
+        app_id TEXT PRIMARY KEY,
+        gemini_api_key TEXT,
+        created_at DATETIME DEFAULT CURRENT_TIMESTAMP
+      )
+    `)
   }
 
   return db
@@ -614,11 +631,13 @@ async function processVoiceAttachment({
   thread,
   projectDirectory,
   isNewThread = false,
+  appId,
 }: {
   message: Message
   thread: ThreadChannel
   projectDirectory?: string
   isNewThread?: boolean
+  appId?: string
 }): Promise<string | null> {
   const audioAttachment = Array.from(message.attachments.values()).find(
     (attachment) => attachment.contentType?.startsWith('audio/'),
@@ -660,9 +679,22 @@ async function processVoiceAttachment({
     }
   }
 
+  // Get Gemini API key from database if appId is provided
+  let geminiApiKey: string | undefined
+  if (appId) {
+    const apiKeys = getDatabase()
+      .prepare('SELECT gemini_api_key FROM bot_api_keys WHERE app_id = ?')
+      .get(appId) as { gemini_api_key: string | null } | undefined
+    
+    if (apiKeys?.gemini_api_key) {
+      geminiApiKey = apiKeys.gemini_api_key
+    }
+  }
+
   const transcription = await transcribeAudio({
     audio: audioBuffer,
     prompt: transcriptionPrompt,
+    geminiApiKey,
   })
 
   voiceLogger.log(
@@ -1635,6 +1667,7 @@ export async function startDiscordBot({
           message,
           thread,
           projectDirectory,
+          appId: currentAppId,
         })
         if (transcription) {
           messageContent = transcription
@@ -1727,6 +1760,7 @@ export async function startDiscordBot({
           thread,
           projectDirectory,
           isNewThread: true,
+          appId: currentAppId,
         })
         if (transcription) {
           messageContent = transcription
@@ -2224,6 +2258,7 @@ export async function startDiscordBot({
           connection,
           guildId: newState.guild.id,
           channelId: voiceChannel.id,
+          appId: currentAppId!,
         })
 
         // Handle connection state changes

package/src/genai-worker-wrapper.ts

@@ -11,6 +11,8 @@ export interface GenAIWorkerOptions {
   systemMessage?: string
   guildId: string
   channelId: string
+  appId: string
+  geminiApiKey?: string | null
   onAssistantOpusPacket: (packet: ArrayBuffer) => void
   onAssistantStartSpeaking?: () => void
   onAssistantStopSpeaking?: () => void
@@ -146,6 +148,8 @@ export function createGenAIWorker(
       systemMessage: options.systemMessage,
       guildId: options.guildId,
       channelId: options.channelId,
+      appId: options.appId,
+      geminiApiKey: options.geminiApiKey,
     }
     worker.postMessage(initMessage)
   })

package/src/genai-worker.ts

@@ -271,6 +271,7 @@ parentPort.on('message', async (message: WorkerInMessage) => {
         session = await startGenAiSession({
           tools,
           systemMessage: message.systemMessage,
+          geminiApiKey: message.geminiApiKey,
           onAssistantAudioChunk({ data }) {
             // Write to audio log if enabled
             if (audioLogStream && !audioLogStream.destroyed) {

package/src/genai.ts

@@ -113,6 +113,7 @@ export async function startGenAiSession({
   onAssistantInterruptSpeaking,
   systemMessage,
   tools,
+  geminiApiKey,
 }: {
   onAssistantAudioChunk?: (args: { data: Buffer; mimeType: string }) => void
   onAssistantStartSpeaking?: () => void
@@ -120,6 +121,7 @@ export async function startGenAiSession({
   onAssistantInterruptSpeaking?: () => void
   systemMessage?: string
   tools?: Record<string, AITool<any, any>>
+  geminiApiKey?: string | null
 } = {}) {
   let session: Session | undefined = undefined
   const callableTools: Array<CallableTool & { name: string }> = []
@@ -242,8 +244,15 @@ export async function startGenAiSession({
     }
   }
 
+  const apiKey = geminiApiKey || process.env.GEMINI_API_KEY
+  
+  if (!apiKey) {
+    genaiLogger.error('No Gemini API key provided')
+    throw new Error('Gemini API key is required for voice interactions')
+  }
+
   const ai = new GoogleGenAI({
-    apiKey: process.env.GEMINI_API_KEY,
+    apiKey,
   })
 
   const model = 'models/gemini-2.5-flash-live-preview'

package/src/voice.ts

@@ -1,5 +1,4 @@
-import { openai } from '@ai-sdk/openai'
-import { experimental_transcribe as transcribe } from 'ai'
+import { GoogleGenAI } from '@google/genai'
 import { createLogger } from './logger.js'
 
 const voiceLogger = createLogger('VOICE')
@@ -9,30 +8,70 @@ export async function transcribeAudio({
   prompt,
   language,
   temperature,
+  geminiApiKey,
 }: {
   audio: Buffer | Uint8Array | ArrayBuffer | string
   prompt?: string
   language?: string
   temperature?: number
+  geminiApiKey?: string
 }): Promise<string> {
   try {
-    const result = await transcribe({
-      model: openai.transcription('whisper-1'),
-      audio,
-      ...(prompt || language || temperature !== undefined
-        ? {
-            providerOptions: {
-              openai: {
-                ...(prompt && { prompt }),
-                ...(language && { language }),
-                ...(temperature !== undefined && { temperature }),
+    // Use provided API key or fall back to environment variable
+    const apiKey = geminiApiKey || process.env.GEMINI_API_KEY
+
+    if (!apiKey) {
+      throw new Error('Gemini API key is required for audio transcription')
+    }
+
+    // Initialize Google Generative AI
+    const genAI = new GoogleGenAI({ apiKey })
+
+    // Convert audio to base64 string if it's not already
+    let audioBase64: string
+    if (typeof audio === 'string') {
+      audioBase64 = audio
+    } else if (audio instanceof Buffer) {
+      audioBase64 = audio.toString('base64')
+    } else if (audio instanceof Uint8Array) {
+      audioBase64 = Buffer.from(audio).toString('base64')
+    } else if (audio instanceof ArrayBuffer) {
+      audioBase64 = Buffer.from(audio).toString('base64')
+    } else {
+      throw new Error('Invalid audio format')
+    }
+
+    // Build the transcription prompt
+    let transcriptionPrompt = `Please transcribe this audio file accurately. Here is some relevant information and filenames that may be present in the audio:\n<context>\n${prompt}\n</context>\n`
+    if (language) {
+      transcriptionPrompt += `\nThe audio is in ${language}.`
+    }
+
+    // Create the content with audio using the inline data format
+    const response = await genAI.models.generateContent({
+      model: 'gemini-2.5-flash',
+      contents: [
+        {
+          parts: [
+            { text: transcriptionPrompt },
+            {
+              inlineData: {
+                data: audioBase64,
+                mimeType: 'audio/mpeg',
               },
             },
-          }
-        : {}),
+          ],
+        },
+      ],
+      config:
+        temperature !== undefined
+          ? {
+              temperature,
+            }
+          : undefined,
     })
 
-    return result.text
+    return response.text || ''
   } catch (error) {
     voiceLogger.error('Failed to transcribe audio:', error)
     throw new Error(

package/src/worker-types.ts

@@ -8,6 +8,8 @@ export type WorkerInMessage =
       systemMessage?: string
       guildId: string
       channelId: string
+      appId: string
+      geminiApiKey?: string | null
     }
   | {
       type: 'sendRealtimeInput'