Add HeAR audio support

medgemma-hf-space/server.py

@@ -18,6 +18,7 @@ from pydantic import BaseModel
 
 # Configuration
 LLAMA_SERVER_URL = os.getenv("LLAMA_SERVER_URL", "http://localhost:8081")
+HEAR_SERVER_URL = os.getenv("HEAR_SERVER_URL", "")  # Empty = disabled
 DB_PATH = os.getenv("DB_PATH", "data/fhir.db")
 
 # Headers for LLM requests (ngrok requires this)
@@ -440,17 +441,61 @@ async def agent_chat_endpoint(request: ChatRequest):
     return StreamingResponse(generate(), media_type="text/event-stream")
 
 # ============================================================================
-# Audio Status (disabled on HF Spaces)
+# Audio Analysis (proxies to remote HeAR server)
 # ============================================================================
 
+from fastapi import File, UploadFile
+
 @app.get("/api/audio/status")
 async def audio_analyzer_status():
-    return {
-        "available": False,
-        "model": None,
-        "message": "Audio analysis not available on HuggingFace Spaces.",
-        "capabilities": []
-    }
+    if not HEAR_SERVER_URL:
+        return {
+            "available": False,
+            "model": None,
+            "message": "Audio analysis not configured. Set HEAR_SERVER_URL.",
+            "capabilities": []
+        }
+    
+    # Check remote HeAR server
+    async with httpx.AsyncClient(timeout=5.0) as client:
+        try:
+            resp = await client.get(f"{HEAR_SERVER_URL}/status")
+            if resp.status_code == 200:
+                data = resp.json()
+                return {
+                    "available": data.get("available", True),
+                    "model": "HeAR (Remote)",
+                    "model_type": "HeAR (Health Acoustic Representations)",
+                    "message": "Connected to remote HeAR server",
+                    "capabilities": data.get("capabilities", ["cough_detection", "covid_risk_screening", "tb_risk_screening"])
+                }
+        except Exception as e:
+            return {
+                "available": False,
+                "model": None,
+                "message": f"Cannot connect to HeAR server: {str(e)}",
+                "capabilities": []
+            }
+
+@app.post("/api/audio/analyze")
+async def analyze_audio(audio: UploadFile = File(...)):
+    if not HEAR_SERVER_URL:
+        return {"success": False, "error": "Audio analysis not configured"}
+    
+    try:
+        audio_bytes = await audio.read()
+        
+        async with httpx.AsyncClient(timeout=60.0) as client:
+            files = {"audio": ("recording.webm", audio_bytes, "audio/webm")}
+            resp = await client.post(f"{HEAR_SERVER_URL}/analyze", files=files)
+            
+            if resp.status_code == 200:
+                result = resp.json()
+                return result
+            else:
+                return {"success": False, "error": f"HeAR server error: {resp.status_code}"}
+    except Exception as e:
+        return {"success": False, "error": str(e)}
 
 if __name__ == "__main__":
     import uvicorn

server.py

@@ -18,6 +18,7 @@ from pydantic import BaseModel
 
 # Configuration
 LLAMA_SERVER_URL = os.getenv("LLAMA_SERVER_URL", "http://localhost:8081")
+HEAR_SERVER_URL = os.getenv("HEAR_SERVER_URL", "")  # Empty = disabled
 DB_PATH = os.getenv("DB_PATH", "data/fhir.db")
 
 # Headers for LLM requests (ngrok requires this)
@@ -440,17 +441,61 @@ async def agent_chat_endpoint(request: ChatRequest):
     return StreamingResponse(generate(), media_type="text/event-stream")
 
 # ============================================================================
-# Audio Status (disabled on HF Spaces)
+# Audio Analysis (proxies to remote HeAR server)
 # ============================================================================
 
+from fastapi import File, UploadFile
+
 @app.get("/api/audio/status")
 async def audio_analyzer_status():
-    return {
-        "available": False,
-        "model": None,
-        "message": "Audio analysis not available on HuggingFace Spaces.",
-        "capabilities": []
-    }
+    if not HEAR_SERVER_URL:
+        return {
+            "available": False,
+            "model": None,
+            "message": "Audio analysis not configured. Set HEAR_SERVER_URL.",
+            "capabilities": []
+        }
+    
+    # Check remote HeAR server
+    async with httpx.AsyncClient(timeout=5.0) as client:
+        try:
+            resp = await client.get(f"{HEAR_SERVER_URL}/status")
+            if resp.status_code == 200:
+                data = resp.json()
+                return {
+                    "available": data.get("available", True),
+                    "model": "HeAR (Remote)",
+                    "model_type": "HeAR (Health Acoustic Representations)",
+                    "message": "Connected to remote HeAR server",
+                    "capabilities": data.get("capabilities", ["cough_detection", "covid_risk_screening", "tb_risk_screening"])
+                }
+        except Exception as e:
+            return {
+                "available": False,
+                "model": None,
+                "message": f"Cannot connect to HeAR server: {str(e)}",
+                "capabilities": []
+            }
+
+@app.post("/api/audio/analyze")
+async def analyze_audio(audio: UploadFile = File(...)):
+    if not HEAR_SERVER_URL:
+        return {"success": False, "error": "Audio analysis not configured"}
+    
+    try:
+        audio_bytes = await audio.read()
+        
+        async with httpx.AsyncClient(timeout=60.0) as client:
+            files = {"audio": ("recording.webm", audio_bytes, "audio/webm")}
+            resp = await client.post(f"{HEAR_SERVER_URL}/analyze", files=files)
+            
+            if resp.status_code == 200:
+                result = resp.json()
+                return result
+            else:
+                return {"success": False, "error": f"HeAR server error: {resp.status_code}"}
+    except Exception as e:
+        return {"success": False, "error": str(e)}
 
 if __name__ == "__main__":
     import uvicorn