open-agents-ai 0.77.0 → 0.78.0

package/dist/index.js

@@ -9550,7 +9550,18 @@ function installCronJob(task, workingDir) {
   const oaBin = findOaBinary();
   const logDir = resolve21(workingDir, ".oa", "scheduled", "logs");
   const logFile = join23(logDir, `${task.id}.log`);
-  const cronLine = `${task.schedule} cd ${JSON.stringify(workingDir)} && ${oaBin} ${JSON.stringify(task.task)} >> ${JSON.stringify(logFile)} 2>&1 ${CRON_MARKER}${task.id}`;
+  const storeFile = resolve21(workingDir, ".oa", "scheduled", "tasks.json");
+  const taskEscaped = task.task.replace(/'/g, "'\\''");
+  const taskId = task.id;
+  const wrapperCmd = [
+    `cd ${JSON.stringify(workingDir)}`,
+    `mkdir -p ${JSON.stringify(logDir)}`,
+    // Run the task and capture exit code
+    `${oaBin} '${taskEscaped}' >> ${JSON.stringify(logFile)} 2>&1; _oa_exit=$?`,
+    // Update store: increment runCount, set lastRun, handle oneShot/maxRuns
+    `node -e "const fs=require('fs'),p=${JSON.stringify(storeFile)};try{const s=JSON.parse(fs.readFileSync(p,'utf8'));const t=s.tasks.find(x=>x.id==='${taskId}');if(t){t.runCount=(t.runCount||0)+1;t.lastRun=new Date().toISOString();if(t.oneShot)t.enabled=false;if(t.maxRuns&&t.runCount>=t.maxRuns)t.enabled=false;s.updatedAt=new Date().toISOString();fs.writeFileSync(p,JSON.stringify(s,null,2));}}catch(e){}" 2>/dev/null`
+  ].join("; ");
+  const cronLine = `${task.schedule} ${wrapperCmd} ${CRON_MARKER}${task.id}`;
   const filtered = lines.filter((l) => !l.includes(`${CRON_MARKER}${task.id}`));
   filtered.push(cronLine);
   writeCrontab(filtered);
@@ -10815,8 +10826,21 @@ function installCronAgentJob(job, workingDir) {
   const oaBin = findOaBinary2();
   const logDir = resolve25(workingDir, ".oa", "cron-agents", "logs");
   const logFile = join27(logDir, `${job.id}.log`);
+  const storeFile = resolve25(workingDir, ".oa", "cron-agents", "store.json");
   const taskEscaped = job.task.replace(/'/g, "'\\''");
-  const cronLine = `${job.schedule} cd ${JSON.stringify(workingDir)} && ${oaBin} '${taskEscaped}' >> ${JSON.stringify(logFile)} 2>&1 ${CRON_AGENT_MARKER}${job.id}`;
+  const jobId = job.id;
+  const wrapperCmd = [
+    `cd ${JSON.stringify(workingDir)}`,
+    `mkdir -p ${JSON.stringify(logDir)}`,
+    `_oa_start=$(date +%s%3N 2>/dev/null || date +%s)`,
+    // Run the task and capture exit code
+    `${oaBin} '${taskEscaped}' >> ${JSON.stringify(logFile)} 2>&1; _oa_exit=$?`,
+    `_oa_end=$(date +%s%3N 2>/dev/null || date +%s)`,
+    // Update store: increment runCount, set lastRunAt, append to history
+    // Pass shell vars as node CLI args: node -e "..." exitCode startMs endMs
+    `node -e "const fs=require('fs'),p=${JSON.stringify(storeFile)};const[,,ex,st,en]=process.argv;try{const s=JSON.parse(fs.readFileSync(p,'utf8'));const j=s.jobs.find(x=>x.id==='${jobId}');if(j){j.runCount=(j.runCount||0)+1;j.lastRunAt=new Date().toISOString();const dur=Math.max(0,(+en||0)-(+st||0));j.history=(j.history||[]).concat({runAt:j.lastRunAt,exitCode:+(ex||0),outputSummary:'cron execution',durationMs:dur});if(j.history.length>20)j.history=j.history.slice(-20);if(j.maxRuns>0&&j.runCount>=j.maxRuns)j.status='completed';s.updatedAt=new Date().toISOString();fs.writeFileSync(p,JSON.stringify(s,null,2));}}catch(e){}" $_oa_exit $_oa_start $_oa_end 2>/dev/null`
+  ].join("; ");
+  const cronLine = `${job.schedule} ${wrapperCmd} ${CRON_AGENT_MARKER}${job.id}`;
   const filtered = lines.filter((l) => !l.includes(`${CRON_AGENT_MARKER}${job.id}`));
   filtered.push(cronLine);
   writeCrontab2(filtered);
@@ -30568,18 +30592,31 @@ var init_voice = __esm({
         return `Voice model set to ${VOICE_MODELS[key].label}. Enable with /voice.`;
       }
       /**
-       * Speak text asynchronously (non-blocking).
+       * Speak text asynchronously (non-blocking) at full volume.
        * Long text is chunked on sentence/line boundaries for reliable TTS.
        * Chunks are queued FIFO and played back-to-back without cutoff.
        */
       speak(text) {
+        this.enqueueSpeech(text, 1, 1);
+      }
+      /**
+       * Speak text at reduced volume with slight pitch shift to indicate
+       * subordinate/secondary output (sub-agent activity, tool narration, etc.).
+       * Volume: 55% of primary. Pitch: shifted down slightly (0.92x).
+       */
+      speakSubordinate(text) {
+        this.enqueueSpeech(text, 0.55, 0.92);
+      }
+      enqueueSpeech(text, volume, pitchFactor) {
         if (!this.enabled || !this.ready)
           return;
         const chunks = this.chunkText(text);
         if (this.speakQueue.length >= 30) {
           this.speakQueue.length = 0;
         }
-        this.speakQueue.push(...chunks);
+        for (const chunk of chunks) {
+          this.speakQueue.push({ text: chunk, volume, pitchFactor });
+        }
         if (!this.speaking) {
           this.drainQueue().catch(() => {
           });
@@ -30704,33 +30741,98 @@ var init_voice = __esm({
       // -------------------------------------------------------------------------
       /**
        * Drain the speak queue FIFO — each item is synthesized and played to
-       * completion before the next starts.  Items play back-to-back without
-       * cutting off previous audio.
+       * completion before the next starts.  A brief silence gap (250ms) is
+       * inserted between queued items for natural pacing.
        */
       async drainQueue() {
         this.speaking = true;
+        let isFirst = true;
         while (this.speakQueue.length > 0) {
-          const text = this.speakQueue.shift();
+          const item = this.speakQueue.shift();
+          if (!isFirst) {
+            await this.sleep(250);
+          }
+          isFirst = false;
           try {
-            await this.synthesizeAndPlay(text);
+            await this.synthesizeAndPlay(item.text, item.volume, item.pitchFactor);
           } catch {
           }
         }
         this.speaking = false;
       }
+      sleep(ms) {
+        return new Promise((resolve31) => setTimeout(resolve31, ms));
+      }
       // -------------------------------------------------------------------------
       // Synthesis pipeline
       // -------------------------------------------------------------------------
-      async synthesizeAndPlay(text) {
+      async synthesizeAndPlay(text, volume = 1, pitchFactor = 1) {
         if (!this.session || !this.config || !this.ort)
           return;
         const cleaned = text.replace(/\*/g, "");
         if (!cleaned.trim())
           return;
-        const textPhonemes = await this.textToPhonemes(cleaned);
+        const sentences = cleaned.split(/(?<=[.!?])\s+/).filter((s) => s.trim());
+        const sampleRate = this.config.audio.sample_rate;
+        const sentenceSilenceSamples = Math.round(sampleRate * 0.18);
+        const sentenceSilence = new Float32Array(sentenceSilenceSamples);
+        const allSegments = [];
+        for (let si = 0; si < sentences.length; si++) {
+          const sentence = sentences[si].trim();
+          if (!sentence)
+            continue;
+          const audioData2 = await this.synthesizeRaw(sentence);
+          if (!audioData2 || audioData2.length === 0)
+            continue;
+          allSegments.push(audioData2);
+          if (si < sentences.length - 1) {
+            allSegments.push(sentenceSilence);
+          }
+        }
+        if (allSegments.length === 0)
+          return;
+        const totalLen = allSegments.reduce((acc, seg) => acc + seg.length, 0);
+        let audioData = new Float32Array(totalLen);
+        let offset = 0;
+        for (const seg of allSegments) {
+          audioData.set(seg, offset);
+          offset += seg.length;
+        }
+        if (volume !== 1) {
+          for (let i = 0; i < audioData.length; i++) {
+            audioData[i] = audioData[i] * volume;
+          }
+        }
+        if (pitchFactor !== 1) {
+          audioData = this.resamplePitch(audioData, pitchFactor);
+        }
+        if (this.onPCMOutput) {
+          const int16 = new Int16Array(audioData.length);
+          for (let i = 0; i < audioData.length; i++) {
+            const s = Math.max(-1, Math.min(1, audioData[i]));
+            int16[i] = s < 0 ? s * 32768 : s * 32767;
+          }
+          this.onPCMOutput(Buffer.from(int16.buffer), sampleRate);
+        }
+        const wavPath = join40(tmpdir6(), `oa-voice-${Date.now()}.wav`);
+        this.writeWav(audioData, sampleRate, wavPath);
+        await this.playWav(wavPath);
+        try {
+          unlinkSync4(wavPath);
+        } catch {
+        }
+      }
+      /**
+       * Synthesize text to raw Float32 audio samples via ONNX inference.
+       * Returns null if synthesis fails or produces no audio.
+       */
+      async synthesizeRaw(text) {
+        if (!this.session || !this.config || !this.ort)
+          return null;
+        const textPhonemes = await this.textToPhonemes(text);
         const phonemeIds = this.phonemesToIds(textPhonemes);
         if (phonemeIds.length === 0)
-          return;
+          return null;
         const inputLength = phonemeIds.length;
         const inputTensor = new this.ort.Tensor("int64", BigInt64Array.from(phonemeIds.map((id) => BigInt(id))), [1, inputLength]);
         const lengthTensor = new this.ort.Tensor("int64", BigInt64Array.from([BigInt(inputLength)]), [1]);
@@ -30745,23 +30847,28 @@ var init_voice = __esm({
         }
         const result = await this.session.run(feeds);
         const audioData = result["output"].data;
-        if (audioData.length === 0)
-          return;
-        if (this.onPCMOutput) {
-          const int16 = new Int16Array(audioData.length);
-          for (let i = 0; i < audioData.length; i++) {
-            const s = Math.max(-1, Math.min(1, audioData[i]));
-            int16[i] = s < 0 ? s * 32768 : s * 32767;
-          }
-          this.onPCMOutput(Buffer.from(int16.buffer), this.config.audio.sample_rate);
-        }
-        const wavPath = join40(tmpdir6(), `oa-voice-${Date.now()}.wav`);
-        this.writeWav(audioData, this.config.audio.sample_rate, wavPath);
-        await this.playWav(wavPath);
-        try {
-          unlinkSync4(wavPath);
-        } catch {
-        }
+        return audioData.length > 0 ? audioData : null;
+      }
+      /**
+       * Pitch-shift audio via linear-interpolation resampling.
+       * Factor > 1.0 = higher pitch (shorter duration).
+       * Factor < 1.0 = lower pitch (longer duration).
+       * Duration is preserved by adjusting the output length accordingly —
+       * we resample to change pitch, then the WAV sample rate stays constant.
+       */
+      resamplePitch(samples, factor) {
+        const newLen = Math.round(samples.length / factor);
+        if (newLen <= 0)
+          return samples;
+        const out = new Float32Array(newLen);
+        for (let i = 0; i < newLen; i++) {
+          const srcIdx = i * factor;
+          const lo = Math.floor(srcIdx);
+          const hi = Math.min(lo + 1, samples.length - 1);
+          const frac = srcIdx - lo;
+          out[i] = samples[lo] * (1 - frac) + samples[hi] * frac;
+        }
+        return out;
       }
       // -------------------------------------------------------------------------
       // Phonemization
@@ -35745,8 +35852,12 @@ with summary "no_reply" to silently skip without responding.
         }
         const existing = this.subAgents.get(msg.chatId);
         if (existing && !existing.aborted) {
-          existing.runner.injectUserMessage(msg.text);
-          this.tuiWrite(() => renderTelegramSubAgentEvent(msg.username, "mid-conversation steering injected"));
+          if (existing.runner) {
+            existing.runner.injectUserMessage(msg.text);
+            this.tuiWrite(() => renderTelegramSubAgentEvent(msg.username, "mid-conversation steering injected"));
+          } else {
+            this.tuiWrite(() => renderTelegramSubAgentEvent(msg.username, "queued (sub-agent still starting)"));
+          }
           return;
         }
         const subAgent = {
@@ -38168,7 +38279,7 @@ ${entry.fullContent}`
             const emoCtx = emoState ? { valence: emoState.valence, arousal: emoState.arousal, label: emoState.label, emoji: emoState.emoji } : void 0;
             const desc = describeToolCall(event.toolName ?? "unknown", event.toolArgs ?? {}, vLevel, emoCtx);
             renderVoiceText(desc);
-            voice.speak(desc);
+            voice.speakSubordinate(desc);
           }
           renderToolCallStart(event.toolName ?? "unknown", event.toolArgs ?? {}, config.verbose);
         });
@@ -38206,7 +38317,7 @@ ${entry.fullContent}`
             const desc = describeToolResult(event.toolName ?? "unknown", event.success ?? false, vLevel, event.content ?? void 0, emoCtx2);
             if (desc) {
               renderVoiceText(desc);
-              voice.speak(desc);
+              voice.speakSubordinate(desc);
             }
           }
         });

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "open-agents-ai",
-  "version": "0.77.0",
+  "version": "0.78.0",
   "description": "AI coding agent powered by open-source models (Ollama/vLLM) — interactive TUI with agentic tool-calling loop",
   "type": "module",
   "main": "./dist/index.js",