open-agents-ai 0.187.547 → 0.187.548

package/dist/index.js

@@ -539822,6 +539822,91 @@ function findMicCaptureCommand() {
   }
   return null;
 }
+function findTranscribeFileScript() {
+  const thisDir = dirname24(fileURLToPath10(import.meta.url));
+  const candidates = [
+    join87(thisDir, "../../../../packages/execution/scripts/transcribe-file.py"),
+    join87(thisDir, "../../../packages/execution/scripts/transcribe-file.py"),
+    join87(thisDir, "../../execution/scripts/transcribe-file.py"),
+    join87(thisDir, "../scripts/transcribe-file.py"),
+    join87(thisDir, "../../scripts/transcribe-file.py")
+  ];
+  for (const p2 of candidates) {
+    if (existsSync71(p2)) return p2;
+  }
+  try {
+    const globalRoot = execSync46("npm root -g", {
+      encoding: "utf-8",
+      timeout: 5e3,
+      stdio: ["pipe", "pipe", "pipe"]
+    }).trim();
+    const candidates2 = [
+      join87(globalRoot, "open-agents-ai", "dist", "scripts", "transcribe-file.py"),
+      join87(globalRoot, "open-agents-ai", "scripts", "transcribe-file.py")
+    ];
+    for (const p2 of candidates2) {
+      if (existsSync71(p2)) return p2;
+    }
+  } catch {
+  }
+  return null;
+}
+async function transcribeFileViaWhisper(filePath, model) {
+  const script = findTranscribeFileScript();
+  if (!script) return null;
+  const bin = process.platform === "win32" ? "Scripts" : "bin";
+  const exe = process.platform === "win32" ? "python.exe" : "python3";
+  const venvPython2 = join87(homedir25(), ".open-agents", "venv", bin, exe);
+  if (!existsSync71(venvPython2)) return null;
+  return new Promise((resolve43) => {
+    const child = spawn18(venvPython2, [script], {
+      stdio: ["pipe", "pipe", "pipe"],
+      env: process.env
+    });
+    let out = "";
+    let err = "";
+    let timer = null;
+    const stop2 = (val) => {
+      if (timer) clearTimeout(timer);
+      try {
+        child.kill("SIGTERM");
+      } catch {
+      }
+      resolve43(val ? parse3(val) : null);
+    };
+    function parse3(raw) {
+      try {
+        const j = JSON.parse(raw);
+        if (j.error) return null;
+        if (typeof j.text !== "string") return null;
+        return { text: j.text.trim(), duration: null, speakers: [], segments: [] };
+      } catch {
+        return null;
+      }
+    }
+    child.stdout?.on("data", (d2) => {
+      out += d2.toString("utf-8");
+    });
+    child.stderr?.on("data", (d2) => {
+      err += d2.toString("utf-8");
+    });
+    child.on?.("error", () => stop2(null));
+    child.on?.("close", () => {
+      if (out.trim()) resolve43(parse3(out));
+      else {
+        void err;
+        resolve43(null);
+      }
+    });
+    timer = setTimeout(() => stop2(null), 12e4);
+    try {
+      child.stdin?.write(JSON.stringify({ path: filePath, model }) + "\n");
+      child.stdin?.end();
+    } catch {
+      stop2(null);
+    }
+  });
+}
 function findLiveWhisperScript() {
   const thisDir = dirname24(fileURLToPath10(import.meta.url));
   const candidates = [
@@ -540523,32 +540608,48 @@ transcribe-cli error: ${transcribeCliError}` : "";
           } catch {
           }
         }
-        if (!tc) return null;
         ensureVenvForTranscribeCli();
-        try {
-          const result = await tc.transcribe(filePath, {
-            model: this.config.model,
-            format: "json",
-            diarize: false,
-            wordTimestamps: false
-          });
+        let lastErr = null;
+        if (tc) {
+          try {
+            const result = await tc.transcribe(filePath, {
+              model: this.config.model,
+              format: "json",
+              diarize: false,
+              wordTimestamps: false
+            });
+            if (outputDir) {
+              const { basename: basename21 } = await import("node:path");
+              const transcriptDir = join87(outputDir, ".oa", "transcripts");
+              mkdirSync38(transcriptDir, { recursive: true });
+              const outFile = join87(transcriptDir, `${basename21(filePath)}.txt`);
+              writeFileSync34(outFile, result.text, "utf-8");
+            }
+            return {
+              text: result.text,
+              duration: result.duration,
+              speakers: result.speakers,
+              segments: result.segments
+            };
+          } catch (err) {
+            lastErr = err;
+          }
+        }
+        const fb = await transcribeFileViaWhisper(filePath, this.config.model);
+        if (fb) {
           if (outputDir) {
             const { basename: basename21 } = await import("node:path");
             const transcriptDir = join87(outputDir, ".oa", "transcripts");
             mkdirSync38(transcriptDir, { recursive: true });
             const outFile = join87(transcriptDir, `${basename21(filePath)}.txt`);
-            writeFileSync34(outFile, result.text, "utf-8");
+            writeFileSync34(outFile, fb.text, "utf-8");
           }
-          return {
-            text: result.text,
-            duration: result.duration,
-            speakers: result.speakers,
-            segments: result.segments
-          };
-        } catch (err) {
-          this.emit("error", err instanceof Error ? err : new Error(String(err)));
-          return null;
+          return fb;
+        }
+        if (lastErr) {
+          this.emit("error", lastErr instanceof Error ? lastErr : new Error(String(lastErr)));
         }
+        return null;
       }
       // -------------------------------------------------------------------------
       // Auto-mode silence detection
@@ -598677,6 +598778,26 @@ function formatMetrics() {
   lines.push(`oa_errors_total ${metrics.totalErrors}`);
   return lines.join("\n") + "\n";
 }
+function pcmToWav(pcm, sampleRate, channels, bitsPerSample) {
+  const byteRate = sampleRate * channels * (bitsPerSample / 8);
+  const blockAlign = channels * (bitsPerSample / 8);
+  const dataSize = pcm.length;
+  const header = Buffer.alloc(44);
+  header.write("RIFF", 0);
+  header.writeUInt32LE(36 + dataSize, 4);
+  header.write("WAVE", 8);
+  header.write("fmt ", 12);
+  header.writeUInt32LE(16, 16);
+  header.writeUInt16LE(1, 20);
+  header.writeUInt16LE(channels, 22);
+  header.writeUInt32LE(sampleRate, 24);
+  header.writeUInt32LE(byteRate, 28);
+  header.writeUInt16LE(blockAlign, 32);
+  header.writeUInt16LE(bitsPerSample, 34);
+  header.write("data", 36);
+  header.writeUInt32LE(dataSize, 40);
+  return Buffer.concat([header, pcm]);
+}
 function jsonResponse(res, status, body) {
   const payload = JSON.stringify(body);
   res.writeHead(status, { "Content-Type": "application/json" });
@@ -602566,9 +602687,11 @@ data: ${JSON.stringify(data)}
           } catch {
           }
         };
+        let receivedAnyTranscript = false;
         const onTranscript = (payload) => {
           const p2 = payload;
           if (!p2 || typeof p2.text !== "string") return;
+          receivedAnyTranscript = true;
           sse(p2.final ? "final" : "partial", { text: p2.text });
         };
         listen.on("transcript", onTranscript);
@@ -602584,19 +602707,51 @@ data: ${JSON.stringify(data)}
           cleanup();
           res.end();
         });
+        const pcmChunks = [];
         req2.on("data", (chunk) => {
+          pcmChunks.push(chunk);
           try {
             const t2 = listen.liveTranscriber;
             if (t2?.write) t2.write(chunk);
           } catch {
           }
         });
-        req2.on("end", () => {
-          setTimeout(() => {
+        req2.on("end", async () => {
+          await new Promise((r2) => setTimeout(r2, 1500));
+          if (receivedAnyTranscript) {
             sse("done", { ok: true });
             cleanup();
             res.end();
-          }, 1500);
+            return;
+          }
+          try {
+            const pcm = Buffer.concat(pcmChunks);
+            if (pcm.length < 320) {
+              sse("done", { ok: true, empty: true });
+              cleanup();
+              res.end();
+              return;
+            }
+            const wav = pcmToWav(pcm, 16e3, 1, 16);
+            const fs7 = await import("node:fs");
+            const os8 = await import("node:os");
+            const path8 = await import("node:path");
+            const tmpPath = path8.join(os8.tmpdir(), `oa-stream-${Date.now()}-${Math.random().toString(36).slice(2, 8)}.wav`);
+            fs7.writeFileSync(tmpPath, wav);
+            const result = await listen.transcribeFile(tmpPath).catch(() => null);
+            try {
+              fs7.unlinkSync(tmpPath);
+            } catch {
+            }
+            if (result && result.text) {
+              sse("final", { text: result.text });
+            }
+            sse("done", { ok: true, mode: "one-shot" });
+          } catch (err) {
+            sse("error", { message: err instanceof Error ? err.message : String(err) });
+          }
+          cleanup();
+          res.end();
         });
       } catch (err) {
         jsonResponse(res, 500, { error: "transcribe_stream_failed", message: err instanceof Error ? err.message : String(err) });

package/dist/scripts/transcribe-file.py

@@ -0,0 +1,63 @@
+#!/usr/bin/env python3
+"""
+Transcribe an audio file using openai-whisper (CPU-only).
+
+stdin:  JSON {"path": "...", "model": "tiny|base|small|medium|large-v3"}
+stdout: JSON {"text": "...", "duration": null, "segments": []}
+        or   {"error": "<kind>", "message": "..."}
+
+Force-loads on CPU because the venv's torch may not match the local GPU
+(e.g. a GT 1030 with sm_61 against a sm_75+ torch build) — those mismatches
+print noisy CUDA warnings and then crash. CPU is slower but always works.
+"""
+import sys, os, json, warnings
+
+# Quiet CUDA warnings from torch even though we won't use the GPU.
+warnings.filterwarnings("ignore")
+os.environ.setdefault("CUDA_VISIBLE_DEVICES", "")  # hide all GPUs
+os.environ.setdefault("PYTHONWARNINGS", "ignore")
+
+try:
+    import whisper
+except Exception as e:
+    print(json.dumps({"error": "deps_missing", "message": str(e)}))
+    sys.exit(0)
+
+
+def main() -> None:
+    raw = sys.stdin.read()
+    try:
+        data = json.loads(raw) if raw.strip() else {}
+    except Exception as e:
+        print(json.dumps({"error": "bad_input", "message": str(e)}))
+        return
+
+    path = data.get("path")
+    if not path or not os.path.exists(path):
+        print(json.dumps({"error": "file_not_found", "path": path or ""}))
+        return
+
+    model_name = data.get("model") or "tiny"
+    try:
+        model = whisper.load_model(model_name, device="cpu")
+        result = model.transcribe(path, fp16=False)
+    except Exception as e:
+        print(json.dumps({"error": "whisper_failed", "message": str(e)}))
+        return
+
+    text = (result.get("text") or "").strip()
+    raw_segments = result.get("segments") or []
+    segments = [
+        {
+            "start": float(s.get("start", 0.0)),
+            "end": float(s.get("end", 0.0)),
+            "text": (s.get("text") or "").strip(),
+        }
+        for s in raw_segments
+    ]
+    duration = float(segments[-1]["end"]) if segments else None
+    print(json.dumps({"text": text, "duration": duration, "segments": segments}))
+
+
+if __name__ == "__main__":
+    main()

package/npm-shrinkwrap.json

@@ -1,12 +1,12 @@
 {
   "name": "open-agents-ai",
-  "version": "0.187.547",
+  "version": "0.187.548",
   "lockfileVersion": 3,
   "requires": true,
   "packages": {
     "": {
       "name": "open-agents-ai",
-      "version": "0.187.547",
+      "version": "0.187.548",
       "hasInstallScript": true,
       "license": "CC-BY-NC-4.0",
       "dependencies": {
@@ -3132,12 +3132,12 @@
       }
     },
     "node_modules/express-rate-limit": {
-      "version": "8.5.0",
-      "resolved": "https://registry.npmjs.org/express-rate-limit/-/express-rate-limit-8.5.0.tgz",
-      "integrity": "sha512-XKhFohWaSBdVJNTi5TaHziqnPkv04I9UQV6q1Wy7Ui6GGQZVW12ojDFwqer14EvCXxjvPG0CyWXx7cAXpALB4Q==",
+      "version": "8.5.1",
+      "resolved": "https://registry.npmjs.org/express-rate-limit/-/express-rate-limit-8.5.1.tgz",
+      "integrity": "sha512-5O6KYmyJEpuPJV5hNTXKbAHWRqrzyu+OI3vUnSd2kXFubIVpG7ezpgxQy76Zo5GQZtrQBg86hF+CM/NX+cioiQ==",
       "license": "MIT",
       "dependencies": {
-        "ip-address": "10.1.0"
+        "ip-address": "^10.2.0"
       },
       "engines": {
         "node": ">= 16"
@@ -3750,9 +3750,9 @@
       "license": "Apache-2.0 OR MIT"
     },
     "node_modules/ip-address": {
-      "version": "10.1.0",
-      "resolved": "https://registry.npmjs.org/ip-address/-/ip-address-10.1.0.tgz",
-      "integrity": "sha512-XXADHxXmvT9+CRxhXg56LJovE+bmWnEWB78LB83VZTprKTmaC5QfruXocxzTZ2Kl0DNwKuBdlIhjL8LeY8Sf8Q==",
+      "version": "10.2.0",
+      "resolved": "https://registry.npmjs.org/ip-address/-/ip-address-10.2.0.tgz",
+      "integrity": "sha512-/+S6j4E9AHvW9SWMSEY9Xfy66O5PWvVEJ08O0y5JGyEKQpojb0K0GKpz/v5HJ/G0vi3D2sjGK78119oXZeE0qA==",
       "license": "MIT",
       "engines": {
         "node": ">= 12"

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "open-agents-ai",
-  "version": "0.187.547",
+  "version": "0.187.548",
   "description": "AI coding agent powered by open-source models (Ollama/vLLM) — interactive TUI with agentic tool-calling loop",
   "type": "module",
   "main": "./dist/index.js",