@nicfox77/parakeet-stt 0.2.6 → 0.2.7

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@nicfox77/parakeet-stt",
-  "version": "0.2.6",
+  "version": "0.2.7",
   "description": "Parakeet TDT INT8 speech-to-text plugin for OpenClaw. Supports V2 (English) and V3 (Multilingual) models.",
   "type": "module",
   "repository": {
@@ -22,6 +22,8 @@
     "@sinclair/typebox": "0.34.48"
   },
   "openclaw": {
-    "extensions": ["./index.ts"]
+    "extensions": [
+      "./index.ts"
+    ]
   }
 }

package/scripts/install.sh

@@ -147,42 +147,45 @@ done
 configure_openclaw() {
     echo "Configuring OpenClaw to use Parakeet for audio transcription..."
     
-    # Try config.patch first (partial update)
+    # Use config.patch RPC for partial update
     if command -v openclaw &> /dev/null; then
-        openclaw gateway call config.patch --params '{
-            "patch": {
-                "tools": {
-                    "media": {
-                        "audio": {
-                            "models": [{
-                                "type": "cli",
-                                "command": "'$PARAKEET_DIR'/parakeet-audio-client.py",
-                                "args": ["{{MediaPath}}", "{{OutputDir}}"]
-                            }]
-                        }
-                    }
-                }
+        # Get current config hash (required for config.patch)
+        CONFIG_HASH=$(openclaw gateway call config.get --params '{}' --json 2>/dev/null | jq -r '.payload.hash' 2>/dev/null)
+        
+        if [ -n "$CONFIG_HASH" ] && [ "$CONFIG_HASH" != "null" ]; then
+            openclaw gateway call config.patch --params '{
+                "raw": "{ tools: { media: { audio: { models: [{ \"type\": \"cli\", \"command\": \"'$PARAKEET_DIR'/parakeet-audio-client.py\", \"args\": [\"{{MediaPath}}\"] }] } } } } }",
+                "baseHash": "'$CONFIG_HASH'"
+            }' 2>/dev/null && {
+                echo "Applied config.patch - Parakeet configured and gateway reloaded"
+                return 0
+            } || {
+                echo "Warning: config.patch failed"
             }
-        }' 2>/dev/null && {
-            echo "Config updated via config.patch"
-            echo ""
-            echo "IMPORTANT: Restart the gateway to activate transcription:"
-            echo "  openclaw gateway restart"
-            return 0
-        }
+        else
+            echo "Warning: Could not get config hash for config.patch"
+        fi
+    else
+        echo "Warning: openclaw CLI not found"
     fi
     
     # Fallback: manual instructions
     echo ""
-    echo "Please add to your openclaw.json under tools.media.audio:"
+    echo "Please manually add to your openclaw.json:"
     echo ""
-    echo '  "models": [{'
-    echo '    "type": "cli",'
-    echo '    "command": "'$PARAKEET_DIR'/parakeet-audio-client.py",'
-    echo '    "args": ["{{MediaPath}}", "{{OutputDir}}"]'
-    echo '  }]'
+    echo '  "tools": {'
+    echo '    "media": {'
+    echo '      "audio": {'
+    echo '        "models": [{'
+    echo '          "type": "cli",'
+    echo '          "command": "'$PARAKEET_DIR'/parakeet-audio-client.py",'
+    echo '          "args": ["{{MediaPath}}"]'
+    echo '        }]'
+    echo '      }'
+    echo '    }'
+    echo '  }'
     echo ""
-    echo "Then restart: openclaw gateway restart"
+    echo "Then run: openclaw gateway restart"
 }
 
 configure_openclaw || true
@@ -199,7 +202,4 @@ echo "To switch models, run:"
 echo "  $0 v2  # English optimized"
 echo "  $0 v3  # Multilingual"
 echo ""
-echo "IMPORTANT: Apply config to activate transcription:"
-echo "  openclaw gateway call config.apply --params '{\"note\": \"Parakeet STT\"}'"
-echo ""
-echo "Or restart the gateway service completely."
+echo "Audio transcription is now configured and ready."

package/scripts/parakeet-audio-client.py

@@ -1,4 +1,9 @@
 #!/usr/bin/env python3
+"""
+Parakeet Audio Client for OpenClaw
+Transcribes audio files using the Parakeet lazy daemon.
+Outputs transcript to stdout (OpenClaw CLI model requirement).
+"""
 import json
 import os
 import socket
@@ -10,7 +15,7 @@ SOCKET_PATH = "/tmp/parakeet-lazy.sock"
 DAEMON_PATH = os.path.expanduser("~/.openclaw/tools/parakeet/parakeet-lazy-daemon.py")
 
 def ensure_daemon():
-    # Check if daemon socket exists and responsive
+    """Check if daemon is running, start it if not."""
     try:
         with socket.socket(socket.AF_UNIX, socket.SOCK_STREAM) as s:
             s.settimeout(0.5)
@@ -32,6 +37,7 @@ def ensure_daemon():
         sys.exit(1)
 
 def query_daemon(audio_path):
+    """Query the daemon for transcription."""
     for attempt in range(3):
         try:
             with socket.socket(socket.AF_UNIX, socket.SOCK_STREAM) as s:
@@ -49,25 +55,36 @@ def query_daemon(audio_path):
             if response_data:
                 response = json.loads(response_data.strip())
                 if "text" in response:
-                    print(response["text"])
-                    return 0
+                    return response["text"]
                 else:
                     print(response.get("error", "Unknown error"), file=sys.stderr)
-                    return 1
+                    return None
             else:
                 time.sleep(0.5)
         except Exception as e:
             if attempt == 2:
                 print(f"Daemon communication failed: {e}", file=sys.stderr)
-                return 1
+                return None
             time.sleep(0.5)
-    return 1
+    return None
 
 if __name__ == "__main__":
     if len(sys.argv) < 2:
-        print("Usage: parakeet-audio-client.py <audio_path> [output_dir]", file=sys.stderr)
+        print("Usage: parakeet-audio-client.py <audio_path>", file=sys.stderr)
         sys.exit(1)
+    
     audio_path = sys.argv[1]
-    # output_dir = sys.argv[2] if len(sys.argv) > 2 else None
+    
+    # Start daemon if needed
     ensure_daemon()
-    sys.exit(query_daemon(audio_path))
+    
+    # Get transcription
+    transcript = query_daemon(audio_path)
+    
+    if transcript:
+        # Output transcript to stdout (OpenClaw reads stdout for CLI transcribers)
+        print(transcript)
+        sys.exit(0)
+    else:
+        print("Transcription failed", file=sys.stderr)
+        sys.exit(1)