@iletai/nzb 1.3.3 → 1.3.5

package/dist/config.js

@@ -14,6 +14,7 @@ const configSchema = z.object({
     SHOW_REASONING: z.string().optional(),
     LOG_CHANNEL_ID: z.string().optional(),
     NODE_EXTRA_CA_CERTS: z.string().optional(),
+    OPENAI_API_KEY: z.string().optional(),
 });
 const raw = configSchema.parse(process.env);
 // Apply NODE_EXTRA_CA_CERTS from .env if not already set via environment.
@@ -43,6 +44,7 @@ export const config = {
     apiPort: parsedPort,
     logChannelId: parsedLogChannelId,
     workerTimeoutMs: parsedWorkerTimeout,
+    openaiApiKey: raw.OPENAI_API_KEY,
     get copilotModel() {
         return _copilotModel;
     },

package/dist/copilot/orchestrator.js

@@ -404,7 +404,8 @@ export async function sendToOrchestrator(prompt, source, callback, onToolEvent,
                 }
                 // Log both sides of the conversation after delivery
                 try {
-                    logConversation(logRole, prompt, sourceLabel);
+                    const telegramMsgId = source.type === "telegram" ? source.messageId : undefined;
+                    logConversation(logRole, prompt, sourceLabel, telegramMsgId);
                 }
                 catch {
                     /* best-effort */

package/dist/setup.js

@@ -215,6 +215,20 @@ ${BOLD}╔═══════════════════════
     else {
         console.log(`\n${DIM}  Skipping Google. You can always set it up later with: nzb setup${RESET}\n`);
     }
+    // ── Voice / Whisper Setup ────────────────────────────────
+    console.log(`${BOLD}━━━ Voice Message Setup (optional) ━━━${RESET}\n`);
+    console.log(`NZB can transcribe voice messages using OpenAI's Whisper API.`);
+    console.log(`You need an OpenAI API key from ${CYAN}https://platform.openai.com/api-keys${RESET}`);
+    console.log();
+    const existingOpenaiKey = existing.OPENAI_API_KEY;
+    const openaiKey = await ask(rl, `  OpenAI API Key ${existingOpenaiKey ? `${DIM}(Enter to keep existing)${RESET}` : `${DIM}(Enter to skip)${RESET}`}: `);
+    const finalOpenaiKey = openaiKey.trim() || existingOpenaiKey || "";
+    if (finalOpenaiKey) {
+        console.log(`\n${GREEN}  ✓ Whisper transcription enabled${RESET}\n`);
+    }
+    else {
+        console.log(`\n${DIM}  Skipping voice. Voice messages will be saved but not transcribed.${RESET}\n`);
+    }
     // ── Model picker ─────────────────────────────────────────
     console.log(`\n${BOLD}━━━ Default Model ━━━${RESET}\n`);
     console.log(`${DIM}Fetching available models from Copilot...${RESET}`);
@@ -241,6 +255,8 @@ ${BOLD}╔═══════════════════════
         lines.push(`AUTHORIZED_USER_ID=${userId}`);
     lines.push(`API_PORT=${apiPort}`);
     lines.push(`COPILOT_MODEL=${model}`);
+    if (finalOpenaiKey)
+        lines.push(`OPENAI_API_KEY=${finalOpenaiKey}`);
     writeFileSync(ENV_PATH, lines.join("\n") + "\n");
     // ── Done ─────────────────────────────────────────────────
     console.log(`

package/dist/store/db.js

@@ -33,9 +33,19 @@ export function getDb() {
         role TEXT NOT NULL CHECK(role IN ('user', 'assistant', 'system')),
         content TEXT NOT NULL,
         source TEXT NOT NULL DEFAULT 'unknown',
+        telegram_msg_id INTEGER,
         ts DATETIME DEFAULT CURRENT_TIMESTAMP
       )
     `);
+        // Migrate: add telegram_msg_id column if missing
+        try {
+            db.prepare(`SELECT telegram_msg_id FROM conversation_log LIMIT 1`).get();
+        }
+        catch {
+            db.exec(`ALTER TABLE conversation_log ADD COLUMN telegram_msg_id INTEGER`);
+        }
+        // Index for fast reply-to lookups
+        db.exec(`CREATE INDEX IF NOT EXISTS idx_conv_telegram_msg ON conversation_log (telegram_msg_id)`);
         db.exec(`
       CREATE TABLE IF NOT EXISTS memories (
         id INTEGER PRIMARY KEY AUTOINCREMENT,
@@ -60,6 +70,7 @@ export function getDb() {
           role TEXT NOT NULL CHECK(role IN ('user', 'assistant', 'system')),
           content TEXT NOT NULL,
           source TEXT NOT NULL DEFAULT 'unknown',
+          telegram_msg_id INTEGER,
           ts DATETIME DEFAULT CURRENT_TIMESTAMP
         )
       `);
@@ -83,11 +94,13 @@ export function getDb() {
             getState: db.prepare(`SELECT value FROM nzb_state WHERE key = ?`),
             setState: db.prepare(`INSERT OR REPLACE INTO nzb_state (key, value) VALUES (?, ?)`),
             deleteState: db.prepare(`DELETE FROM nzb_state WHERE key = ?`),
-            logConversation: db.prepare(`INSERT INTO conversation_log (role, content, source) VALUES (?, ?, ?)`),
+            logConversation: db.prepare(`INSERT INTO conversation_log (role, content, source, telegram_msg_id) VALUES (?, ?, ?, ?)`),
             pruneConversation: db.prepare(`DELETE FROM conversation_log WHERE id NOT IN (SELECT id FROM conversation_log ORDER BY id DESC LIMIT 200)`),
             addMemory: db.prepare(`INSERT INTO memories (category, content, source) VALUES (?, ?, ?)`),
             removeMemory: db.prepare(`DELETE FROM memories WHERE id = ?`),
             memorySummary: db.prepare(`SELECT id, category, content FROM memories ORDER BY category, last_accessed DESC`),
+            getConversationByMsgId: db.prepare(`SELECT id FROM conversation_log WHERE telegram_msg_id = ? LIMIT 1`),
+            getConversationAround: db.prepare(`SELECT role, content, source, ts FROM conversation_log WHERE id BETWEEN ? - 4 AND ? + 4 ORDER BY id ASC`),
         };
     }
     return db;
@@ -106,16 +119,38 @@ export function deleteState(key) {
     getDb(); // ensure init
     stmtCache.deleteState.run(key);
 }
-/** Log a conversation turn (user, assistant, or system). */
-export function logConversation(role, content, source) {
+/** Log a conversation turn (user, assistant, or system) with optional Telegram message ID. */
+export function logConversation(role, content, source, telegramMsgId) {
     getDb(); // ensure init
-    stmtCache.logConversation.run(role, content, source);
+    stmtCache.logConversation.run(role, content, source, telegramMsgId ?? null);
     // Keep last 200 entries to support context recovery after session loss
     logInsertCount++;
     if (logInsertCount % 50 === 0) {
         stmtCache.pruneConversation.run();
     }
 }
+/** Get conversation context around a Telegram message ID (±2 turns). */
+export function getConversationContext(telegramMsgId) {
+    getDb(); // ensure init
+    const row = stmtCache.getConversationByMsgId.get(telegramMsgId);
+    if (!row)
+        return undefined;
+    const rows = stmtCache.getConversationAround.all(row.id, row.id);
+    if (rows.length === 0)
+        return undefined;
+    return rows
+        .map((r) => {
+        const tag = r.role === "user" ? "You" : r.role === "assistant" ? "NZB" : "System";
+        const content = r.content.length > 400 ? r.content.slice(0, 400) + "…" : r.content;
+        return `${tag}: ${content}`;
+    })
+        .join("\n");
+}
+/** Update the most recent assistant log entry with its Telegram message ID (for reply-to lookups). */
+export function updateLastAssistantTelegramMsgId(telegramMsgId) {
+    const db = getDb();
+    db.prepare(`UPDATE conversation_log SET telegram_msg_id = ? WHERE id = (SELECT id FROM conversation_log WHERE role = 'assistant' ORDER BY id DESC LIMIT 1)`).run(telegramMsgId);
+}
 /** Get recent conversation history formatted for injection into system message. */
 export function getRecentConversation(limit = 20) {
     const db = getDb();

package/dist/telegram/bot.js

@@ -439,12 +439,20 @@ export function createBot() {
         const onUsage = (usage) => {
             usageInfo = usage;
         };
-        // If user replies to a message, include that message's text as context
+        // If user replies to a message, include surrounding conversation context
         let userPrompt = ctx.message.text;
         const replyMsg = ctx.message.reply_to_message;
         if (replyMsg && "text" in replyMsg && replyMsg.text) {
-            const quoted = replyMsg.text.length > 500 ? replyMsg.text.slice(0, 500) + "…" : replyMsg.text;
-            userPrompt = `[Replying to: "${quoted}"]\n\n${userPrompt}`;
+            // Try to find full conversation context around the replied message
+            const { getConversationContext } = await import("../store/db.js");
+            const context = getConversationContext(replyMsg.message_id);
+            if (context) {
+                userPrompt = `[Continuing from earlier conversation:]\n---\n${context}\n---\n\n[Your reply]: ${userPrompt}`;
+            }
+            else {
+                const quoted = replyMsg.text.length > 500 ? replyMsg.text.slice(0, 500) + "…" : replyMsg.text;
+                userPrompt = `[Replying to: "${quoted}"]\n\n${userPrompt}`;
+            }
         }
         sendToOrchestrator(userPrompt, { type: "telegram", chatId, messageId: userMessageId }, (text, done) => {
             if (done) {
@@ -505,6 +513,11 @@ export function createBot() {
                                 await bot.api.setMessageReaction(chatId, userMessageId, [{ type: "emoji", emoji: "👍" }]);
                             }
                             catch { }
+                            try {
+                                const { updateLastAssistantTelegramMsgId } = await import("../store/db.js");
+                                updateLastAssistantTelegramMsgId(placeholderMsgId);
+                            }
+                            catch { }
                             return;
                         }
                         catch {
@@ -514,6 +527,11 @@ export function createBot() {
                                     await bot.api.setMessageReaction(chatId, userMessageId, [{ type: "emoji", emoji: "👍" }]);
                                 }
                                 catch { }
+                                try {
+                                    const { updateLastAssistantTelegramMsgId } = await import("../store/db.js");
+                                    updateLastAssistantTelegramMsgId(placeholderMsgId);
+                                }
+                                catch { }
                                 return;
                             }
                             catch {
@@ -736,6 +754,110 @@ export function createBot() {
             });
         }
     });
+    // Handle voice messages — download, transcribe via Whisper, send to AI
+    bot.on("message:voice", async (ctx) => {
+        const chatId = ctx.chat.id;
+        const userMessageId = ctx.message.message_id;
+        const duration = ctx.message.voice.duration;
+        void logInfo(`🎤 Voice received: ${duration}s`);
+        try {
+            await ctx.react("👀");
+        }
+        catch { }
+        // Limit voice duration to 5 minutes
+        if (duration > 300) {
+            await ctx.reply("❌ Voice too long (max 5 min).", { reply_parameters: { message_id: userMessageId } });
+            return;
+        }
+        try {
+            const file = await ctx.api.getFile(ctx.message.voice.file_id);
+            const filePath = file.file_path;
+            if (!filePath) {
+                await ctx.reply("❌ Could not download voice.", { reply_parameters: { message_id: userMessageId } });
+                return;
+            }
+            const url = `https://api.telegram.org/file/bot${config.telegramBotToken}/${filePath}`;
+            const { mkdtempSync, writeFileSync } = await import("fs");
+            const { join } = await import("path");
+            const { tmpdir } = await import("os");
+            const tmpDir = mkdtempSync(join(tmpdir(), "nzb-voice-"));
+            const ext = filePath.split(".").pop() || "oga";
+            const localPath = join(tmpDir, `voice.${ext}`);
+            const response = await fetch(url);
+            const buffer = Buffer.from(await response.arrayBuffer());
+            writeFileSync(localPath, buffer);
+            let prompt;
+            if (config.openaiApiKey) {
+                // Transcribe using OpenAI Whisper API
+                try {
+                    const formData = new FormData();
+                    formData.append("file", new Blob([buffer], { type: "audio/ogg" }), `voice.${ext}`);
+                    formData.append("model", "whisper-1");
+                    const whisperResp = await fetch("https://api.openai.com/v1/audio/transcriptions", {
+                        method: "POST",
+                        headers: { Authorization: `Bearer ${config.openaiApiKey}` },
+                        body: formData,
+                    });
+                    if (!whisperResp.ok) {
+                        const errText = await whisperResp.text();
+                        throw new Error(`Whisper API ${whisperResp.status}: ${errText.slice(0, 200)}`);
+                    }
+                    const result = (await whisperResp.json());
+                    const transcript = result.text?.trim();
+                    if (!transcript) {
+                        prompt = `[User sent a voice message (${duration}s) but transcription was empty. File saved at: ${localPath}]`;
+                    }
+                    else {
+                        prompt = `[Voice message transcribed (${duration}s)]: ${transcript}`;
+                    }
+                }
+                catch (whisperErr) {
+                    console.error("[nzb] Whisper transcription failed:", whisperErr instanceof Error ? whisperErr.message : whisperErr);
+                    prompt = `[User sent a voice message (${duration}s), saved at: ${localPath}. Transcription failed: ${whisperErr instanceof Error ? whisperErr.message : String(whisperErr)}]`;
+                }
+            }
+            else {
+                prompt = `[User sent a voice message (${duration}s), saved at: ${localPath}. No OPENAI_API_KEY configured for transcription. You can tell the user to set it up in ~/.nzb/.env]`;
+            }
+            sendToOrchestrator(prompt, { type: "telegram", chatId, messageId: userMessageId }, (text, done) => {
+                if (done) {
+                    const formatted = toTelegramMarkdown(text);
+                    const chunks = chunkMessage(formatted);
+                    const fallbackChunks = chunkMessage(text);
+                    void (async () => {
+                        for (let i = 0; i < chunks.length; i++) {
+                            if (i > 0)
+                                await new Promise((r) => setTimeout(r, 300));
+                            const pageTag = chunks.length > 1 ? `📄 ${i + 1}/${chunks.length}\n` : "";
+                            try {
+                                await ctx.api.sendMessage(chatId, pageTag + chunks[i], {
+                                    parse_mode: "MarkdownV2",
+                                    reply_parameters: i === 0 ? { message_id: userMessageId } : undefined,
+                                });
+                            }
+                            catch {
+                                try {
+                                    await ctx.api.sendMessage(chatId, pageTag + (fallbackChunks[i] ?? chunks[i]), {
+                                        reply_parameters: i === 0 ? { message_id: userMessageId } : undefined,
+                                    });
+                                }
+                                catch { }
+                            }
+                        }
+                        try {
+                            await ctx.api.setMessageReaction(chatId, userMessageId, [{ type: "emoji", emoji: "👍" }]);
+                        }
+                        catch { }
+                    })();
+                }
+            });
+        }
+        catch (err) {
+            await ctx.reply(`❌ Error processing voice: ${err instanceof Error ? err.message : String(err)}`, {
+                reply_parameters: { message_id: userMessageId },
+            });
+        }
+    });
     return bot;
 }
 export async function startBot() {

package/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@iletai/nzb",
-	"version": "1.3.3",
+	"version": "1.3.5",
 	"description": "NZB — a personal AI assistant for developers, built on the GitHub Copilot SDK",
 	"bin": {
 		"nzb": "dist/cli.js"