gemcode 0.3.101__tar.gz → 0.3.103__tar.gz

{gemcode-0.3.101/src/gemcode.egg-info → gemcode-0.3.103}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: gemcode
-Version: 0.3.101
+Version: 0.3.103
 Summary: Local-first coding agent on Google Gemini + ADK
 Author: GemCode Contributors
 License:                                  Apache License

{gemcode-0.3.101 → gemcode-0.3.103}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "gemcode"
-version = "0.3.101"
+version = "0.3.103"
 description = "Local-first coding agent on Google Gemini + ADK"
 readme = "README.md"
 requires-python = ">=3.11"

{gemcode-0.3.101 → gemcode-0.3.103}/src/gemcode/cli.py

@@ -710,6 +710,11 @@ def main() -> None:
       action="store_true",
       help="Do not play model audio to speakers (still prints text if any)",
     )
+    audio_parser.add_argument(
+      "--list-devices",
+      action="store_true",
+      help="List available audio devices and exit (for mic troubleshooting)",
+    )
 
     args = audio_parser.parse_args(sys.argv[2:])
     load_cli_environment()
@@ -730,6 +735,36 @@ def main() -> None:
     session_id = args.session or str(uuid.uuid4())
     from gemcode.live_audio_engine import run_live_audio
 
+    if args.list_devices:
+      try:
+        import sounddevice as sd  # type: ignore
+      except Exception:
+        print(
+          "\n[gemcode live-audio] Audio deps missing. Install:\n"
+          "  python3 -m pip install -U \"gemcode[live]\"\n",
+          file=sys.stderr,
+        )
+        raise SystemExit(2)
+      try:
+        devs = sd.query_devices()
+        default_in, default_out = sd.default.device
+        print("Audio devices:")
+        for i, d in enumerate(devs):
+          name = str(d.get("name") or "")
+          mi = int(d.get("max_input_channels") or 0)
+          mo = int(d.get("max_output_channels") or 0)
+          mark = ""
+          if i == default_in:
+            mark += " [default-in]"
+          if i == default_out:
+            mark += " [default-out]"
+          print(f"  {i:>2}: in={mi} out={mo}  {name}{mark}")
+        print("\nTip: set GEMCODE_LIVE_AUDIO_INPUT_DEVICE to a device index or name.")
+      except Exception as e:
+        print(f"[gemcode live-audio] Could not list devices: {e}", file=sys.stderr)
+        raise SystemExit(2)
+      raise SystemExit(0)
+
     # One-time explicit permission prompt (HITL) for mic/speaker use.
     if hasattr(sys.stdin, "isatty") and sys.stdin.isatty():
       try:
@@ -787,7 +822,13 @@ def main() -> None:
             )
           )
         captured = buf.getvalue()
-        if captured and "An unexpected error occurred in live flow: 1000" not in captured:
+        noisy = (
+          "An unexpected error occurred in live flow: 1000" in captured
+          or "APIError: 1000" in captured
+          or "APIError: 1011" in captured
+          or "An unexpected error occurred in live flow: 1011" in captured
+        )
+        if captured and not noisy:
           # Re-emit unexpected stderr.
           print(captured, file=sys.stderr, end="")
       else:
@@ -808,6 +849,16 @@ def main() -> None:
         if isinstance(e, APIError) and (getattr(e, "status_code", None) == 1000 or "1000" in str(e)):
           print("\n[gemcode live-audio] Session ended.", file=sys.stderr)
           raise SystemExit(0)
+        if isinstance(e, APIError) and getattr(e, "status_code", None) == 1011:
+          print(
+            "\n[gemcode live-audio] Gemini Live internal error (1011).\n"
+            "This is usually transient. Try again, or try:\n"
+            "  - set a different live model:  gemcode live-audio --model <id>\n"
+            "  - disable playback:           gemcode live-audio --no-playback\n"
+            "  - shorten the session:        gemcode live-audio --seconds 10\n",
+            file=sys.stderr,
+          )
+          raise SystemExit(2)
       except Exception:
         pass
       # websockets can also surface a close directly.

{gemcode-0.3.101 → gemcode-0.3.103}/src/gemcode/live_audio_engine.py

@@ -8,6 +8,7 @@ This wires GemCode's existing outer session + callbacks into ADK's
 from __future__ import annotations
 
 import asyncio
+import os
 import time
 import sys
 from dataclasses import dataclass
@@ -126,7 +127,8 @@ async def run_live_audio(
 
   run_config = RunConfig(
     # Prefer the enum value (avoids pydantic serializer warnings in some SDK versions).
-    response_modalities=[types.Modality.AUDIO],
+    # Request TEXT too so users see transcripts even when the model returns only audio.
+    response_modalities=[types.Modality.AUDIO, types.Modality.TEXT],
     speech_config=speech_config,
     # Keep SDK defaults for STT/TTS transcription configs.
   )
@@ -188,16 +190,20 @@ async def run_live_audio(
   # Mic capture → async queue (threaded producer).
   pcm_q: asyncio.Queue[bytes] = asyncio.Queue(maxsize=50)
   stop_at = time.time() + max(1, int(seconds))
+  chunks_sent = 0
+  non_silent_chunks = 0
 
   def _mic_thread() -> None:
     # ~20ms frames is a good latency/overhead balance.
     blocksize = max(120, int(input_rate // 50))
+    device = os.environ.get("GEMCODE_LIVE_AUDIO_INPUT_DEVICE")
     try:
       stream = sd.RawInputStream(  # type: ignore[attr-defined]
         samplerate=int(input_rate),
         channels=1,
         dtype="int16",
         blocksize=int(blocksize),
+        device=device if device else None,
       )
     except Exception:
       # Let the consumer surface this as empty audio.
@@ -227,6 +233,14 @@ async def run_live_audio(
         chunk = await asyncio.wait_for(pcm_q.get(), timeout=0.25)
       except asyncio.TimeoutError:
         continue
+      chunks_sent += 1
+      try:
+        arr = np.frombuffer(chunk, dtype="int16")  # type: ignore[attr-defined]
+        # Consider it non-silent if mean abs amplitude crosses a tiny threshold.
+        if arr.size and float(np.mean(np.abs(arr))) > 25.0:  # type: ignore[attr-defined]
+          non_silent_chunks += 1
+      except Exception:
+        pass
       live_queue.send_realtime(
         types.Blob(data=chunk, mime_type=_mime_type_for_rate(input_rate))
       )
@@ -247,6 +261,15 @@ async def run_live_audio(
 
   if not printed_any:
     print("\n[gemcode live-audio] No model text received (audio may have been silent).")
+  if chunks_sent <= 2 or non_silent_chunks == 0:
+    print(
+      "\n[gemcode live-audio] Mic input looks silent or unavailable.\n"
+      "Check:\n"
+      "- System Settings → Privacy & Security → Microphone (allow your terminal)\n"
+      "- Your input device selection\n"
+      "Tip: set GEMCODE_LIVE_AUDIO_INPUT_DEVICE to a device index/name from sounddevice.query_devices().\n",
+      file=sys.stderr,
+    )
 
   await runner.close()
 

{gemcode-0.3.101 → gemcode-0.3.103/src/gemcode.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: gemcode
-Version: 0.3.101
+Version: 0.3.103
 Summary: Local-first coding agent on Google Gemini + ADK
 Author: GemCode Contributors
 License:                                  Apache License