Private-AI

Running

App Files Files Community

seawolf2357 commited on 9 days ago

Commit

f98e6c2

verified ·

1 Parent(s): e32f3fb

Update app.py

Browse files

Files changed (1) hide show

app.py +84 -9

app.py CHANGED Viewed

@@ -32,10 +32,17 @@ load_dotenv()
 SAMPLE_RATE = 24000
-# Use Persistent Storage path if available (Hugging Face Space)
-PERSISTENT_DIR = os.environ.get("PERSISTENT_DIR", "./data")
 os.makedirs(PERSISTENT_DIR, exist_ok=True)
 DB_PATH = os.path.join(PERSISTENT_DIR, "personal_assistant.db")
 # HTML content embedded as a string
 HTML_CONTENT = """<!DOCTYPE html>
@@ -397,12 +404,12 @@ HTML_CONTENT = """<!DOCTYPE html>
             background: linear-gradient(135deg, #27ae60, #229954);
         }
         #end-session-button {
-            background: linear-gradient(135deg, #e74c3c, #c0392b);
             padding: 8px 16px;
             font-size: 13px;
         }
         #end-session-button:hover {
-            background: linear-gradient(135deg, #c0392b, #a93226);
         }
         #audio-output {
             display: none;
@@ -562,7 +569,7 @@ HTML_CONTENT = """<!DOCTYPE html>
                 <div class="controls">
                     <button id="start-button">대화 시작</button>
-                    <button id="end-session-button" style="display: none;">세션 종료</button>
                 </div>
             </div>
@@ -685,7 +692,7 @@ HTML_CONTENT = """<!DOCTYPE html>
                 }
             } catch (error) {
                 console.error('Failed to end session:', error);
-                showError('세션 종료 중 오류가 발생했습니다.');
             }
         }
@@ -1489,11 +1496,20 @@ async def process_text_chat(message: str, web_search_enabled: bool, session_id:
                           user_name: str = "", memories: Dict = None) -> Dict[str, str]:
     """Process text chat using GPT-4o-mini model"""
     try:
         # Build system prompt with memories
         base_prompt = f"""You are a personal AI assistant for {user_name if user_name else 'the user'}.
 You remember all previous conversations and personal information about the user.
 Be friendly, helpful, and personalized in your responses.
-Always use the information you remember to make conversations more personal and relevant."""
         # Add memories to prompt
         if memories:
@@ -1577,6 +1593,8 @@ class OpenAIHandler(AsyncStreamHandler):
         self.session_id = session_id
         self.user_name = user_name
         self.memories = memories or {}
         print(f"[INIT] Handler created with web_search={web_search_enabled}, session_id={session_id}, user={user_name}")
@@ -1654,7 +1672,8 @@ class OpenAIHandler(AsyncStreamHandler):
         base_instructions = f"""You are a personal AI assistant for {self.user_name if self.user_name else 'the user'}.
 You remember all previous conversations and personal information about the user.
 Be friendly, helpful, and personalized in your responses.
-Always use the information you remember to make conversations more personal and relevant."""
         # Add memories to prompt
         if self.memories:
@@ -1695,7 +1714,12 @@ Always use the information you remember to make conversations more personal and
             model="gpt-4o-mini-realtime-preview-2024-12-17"
         ) as conn:
             session_update = {
-                "turn_detection": {"type": "server_vad"},
                 "instructions": instructions,
                 "tools": tools,
                 "tool_choice": "auto" if tools else "none",
@@ -1710,7 +1734,37 @@ Always use the information you remember to make conversations more personal and
             print(f"Connected with tools: {len(tools)} functions")
             async for event in self.connection:
                 if event.type == "response.audio_transcript.done":
                     print(f"[RESPONSE] Transcript: {event.transcript[:100]}...")
                     # Detect language
@@ -1731,7 +1785,17 @@ Always use the information you remember to make conversations more personal and
                     }
                     await self.output_queue.put(AdditionalOutputs(output_data))
                 elif event.type == "response.audio.delta":
                     await self.output_queue.put(
                         (
                             self.output_sample_rate,
@@ -1847,6 +1911,17 @@ async def startup_event():
     await PersonalAssistantDB.init()
     print(f"Database initialized at: {DB_PATH}")
     print(f"Persistent directory: {PERSISTENT_DIR}")
 # Intercept offer to capture settings
 @app.post("/webrtc/offer", include_in_schema=False)

 SAMPLE_RATE = 24000
+# Use Persistent Storage path for Hugging Face Space
+# In HF Spaces, persistent storage is at /data
+if os.path.exists("/data"):
+    PERSISTENT_DIR = "/data"
+else:
+    PERSISTENT_DIR = "./data"
 os.makedirs(PERSISTENT_DIR, exist_ok=True)
 DB_PATH = os.path.join(PERSISTENT_DIR, "personal_assistant.db")
+print(f"Using persistent directory: {PERSISTENT_DIR}")
+print(f"Database path: {DB_PATH}")
 # HTML content embedded as a string
 HTML_CONTENT = """<!DOCTYPE html>
             background: linear-gradient(135deg, #27ae60, #229954);
         }
         #end-session-button {
+            background: linear-gradient(135deg, #4a9eff, #3a7ed8);
             padding: 8px 16px;
             font-size: 13px;
         }
         #end-session-button:hover {
+            background: linear-gradient(135deg, #3a7ed8, #2a5eb8);
         }
         #audio-output {
             display: none;
                 <div class="controls">
                     <button id="start-button">대화 시작</button>
+                    <button id="end-session-button" style="display: none;">기억 업데이트</button>
                 </div>
             </div>
                 }
             } catch (error) {
                 console.error('Failed to end session:', error);
+                showError('기억 업데이트 중 오류가 발생했습니다.');
             }
         }
                           user_name: str = "", memories: Dict = None) -> Dict[str, str]:
     """Process text chat using GPT-4o-mini model"""
     try:
+        # Check for stop words
+        stop_words = ["중단", "그만", "스톱", "stop", "닥쳐", "멈춰", "중지"]
+        if any(word in message.lower() for word in stop_words):
+            return {
+                "response": "대화를 중단합니다.",
+                "detected_language": "ko"
+            }
         # Build system prompt with memories
         base_prompt = f"""You are a personal AI assistant for {user_name if user_name else 'the user'}.
 You remember all previous conversations and personal information about the user.
 Be friendly, helpful, and personalized in your responses.
+Always use the information you remember to make conversations more personal and relevant.
+IMPORTANT: Give only ONE response. Do not repeat or give multiple answers."""
         # Add memories to prompt
         if memories:
         self.session_id = session_id
         self.user_name = user_name
         self.memories = memories or {}
+        self.is_responding = False  # Track if already responding
+        self.should_stop = False  # Track if conversation should stop
         print(f"[INIT] Handler created with web_search={web_search_enabled}, session_id={session_id}, user={user_name}")
         base_instructions = f"""You are a personal AI assistant for {self.user_name if self.user_name else 'the user'}.
 You remember all previous conversations and personal information about the user.
 Be friendly, helpful, and personalized in your responses.
+Always use the information you remember to make conversations more personal and relevant.
+IMPORTANT: Give only ONE response per user input. Do not repeat yourself or give multiple answers."""
         # Add memories to prompt
         if self.memories:
             model="gpt-4o-mini-realtime-preview-2024-12-17"
         ) as conn:
             session_update = {
+                "turn_detection": {
+                    "type": "server_vad",
+                    "threshold": 0.5,
+                    "prefix_padding_ms": 300,
+                    "silence_duration_ms": 200
+                },
                 "instructions": instructions,
                 "tools": tools,
                 "tool_choice": "auto" if tools else "none",
             print(f"Connected with tools: {len(tools)} functions")
             async for event in self.connection:
+                # Handle user transcription for stop detection
+                if event.type == "conversation.item.created":
+                    if hasattr(event, 'item') and hasattr(event.item, 'role') and event.item.role == "user":
+                        if hasattr(event.item, 'content') and event.item.content:
+                            for content_item in event.item.content:
+                                if hasattr(content_item, 'transcript'):
+                                    user_text = content_item.transcript.lower()
+                                    stop_words = ["중단", "그만", "스톱", "stop", "닥쳐", "멈춰", "중지"]
+                                    if any(word in user_text for word in stop_words):
+                                        print(f"[STOP DETECTED] User said: {content_item.transcript}")
+                                        self.should_stop = True
+                                        # Cancel any ongoing response
+                                        if self.connection:
+                                            try:
+                                                await self.connection.response.cancel()
+                                            except:
+                                                pass
+                                        continue
+                                    # Save user message to database
+                                    if self.session_id:
+                                        await PersonalAssistantDB.save_message(self.session_id, "user", content_item.transcript)
                 if event.type == "response.audio_transcript.done":
+                    # Prevent multiple responses
+                    if self.is_responding:
+                        print("[DUPLICATE RESPONSE] Skipping duplicate response")
+                        continue
+                    self.is_responding = True
                     print(f"[RESPONSE] Transcript: {event.transcript[:100]}...")
                     # Detect language
                     }
                     await self.output_queue.put(AdditionalOutputs(output_data))
+                elif event.type == "response.done":
+                    # Reset responding flag when response is complete
+                    self.is_responding = False
+                    self.should_stop = False
+                    print("[RESPONSE DONE] Response completed")
                 elif event.type == "response.audio.delta":
+                    # Check if we should stop
+                    if self.should_stop:
+                        continue
                     await self.output_queue.put(
                         (
                             self.output_sample_rate,
     await PersonalAssistantDB.init()
     print(f"Database initialized at: {DB_PATH}")
     print(f"Persistent directory: {PERSISTENT_DIR}")
+    print(f"DB file exists: {os.path.exists(DB_PATH)}")
+    # Check if we're in Hugging Face Space
+    if os.path.exists("/data"):
+        print("Running in Hugging Face Space with persistent storage")
+        # List files in persistent directory
+        try:
+            files = os.listdir(PERSISTENT_DIR)
+            print(f"Files in persistent directory: {files}")
+        except Exception as e:
+            print(f"Error listing files: {e}")
 # Intercept offer to capture settings
 @app.post("/webrtc/offer", include_in_schema=False)