Refactor Unified API for stable MoE execution with Multi-Modal Routing and dynamic Expert Support

Author: harshdesaii

examples/data/sample_audio.mp3

@@ -6,21 +6,36 @@
 import base64
 import os
 from pathlib import Path
-from typing import Dict, Any
+from typing import Dict, Any, Optional
 import requests
 from multimind.api.unified_api import UnifiedRequest, ModalityInput
 
 def get_data_path(filename: str) -> Path:
     """Get the absolute path to a data file."""
-    return Path(os.path.join(os.path.dirname(__file__), "..", "..", "data", filename))
+    # examples/multi_modal/basic/process_request.py -> parents[2] == examples/
+    return Path(__file__).resolve().parents[2] / "data" / filename
+
+
+def _pick_first(data_dir: Path, patterns: list[str]) -> Optional[Path]:
+    """Pick first matching file in data_dir for given glob patterns."""
+    for pat in patterns:
+        matches = sorted(data_dir.glob(pat))
+        if matches:
+            return matches[0]
+    return None
 
 async def process_image_caption():
     """Process an image captioning request."""
     # Load and encode image
-    image_path = get_data_path("sample_image.jpg")
+    data_dir = Path(__file__).resolve().parents[2] / "data"
+    image_path = get_data_path("sample_image1.png")
     if not image_path.exists():
-        print(f"Error: Image file not found at {image_path}")
-        return
+        fallback = _pick_first(data_dir, ["*.png", "*.jpg", "*.jpeg", "*.webp"])
+        if fallback is None:
+            print(f"Error: Image file not found at {image_path}")
+            return
+        image_path = fallback
+        print(f"Warning: Using fallback image file {image_path}")
 
     with open(image_path, "rb") as f:
         image_data = base64.b64encode(f.read()).decode()
@@ -47,7 +62,7 @@ async def process_image_caption():
     # Send request to API
     response = requests.post(
         "http://localhost:8000/v1/process",
-        json=request.dict()
+        json=request.model_dump()
     )
 
     if response.status_code == 200:
@@ -63,10 +78,15 @@ async def process_image_caption():
 async def process_audio_transcription():
     """Process an audio transcription request."""
     # Load and encode audio
+    data_dir = Path(__file__).resolve().parents[2] / "data"
     audio_path = get_data_path("sample_audio.mp3")
     if not audio_path.exists():
-        print(f"Error: Audio file not found at {audio_path}")
-        return
+        fallback = _pick_first(data_dir, ["*.mp3", "*.wav", "*.m4a", "*.flac", "*.ogg"])
+        if fallback is None:
+            print(f"Error: Audio file not found at {audio_path}")
+            return
+        audio_path = fallback
+        print(f"Warning: Using fallback audio file {audio_path}")
 
     with open(audio_path, "rb") as f:
         audio_data = base64.b64encode(f.read()).decode()
@@ -93,7 +113,7 @@ async def process_audio_transcription():
     # Send request to API
     response = requests.post(
         "http://localhost:8000/v1/process",
-        json=request.dict()
+        json=request.model_dump()
     )
 
     if response.status_code == 200:
@@ -109,15 +129,24 @@ async def process_audio_transcription():
 async def process_multi_modal_analysis():
     """Process a complex multi-modal analysis request."""
     # Load and encode media
-    image_path = get_data_path("sample_image.jpg")
+    data_dir = Path(__file__).resolve().parents[2] / "data"
+    image_path = get_data_path("sample_image1.png")
     audio_path = get_data_path("sample_audio.mp3")
 
     if not image_path.exists():
-        print(f"Error: Image file not found at {image_path}")
-        return
+        fallback = _pick_first(data_dir, ["*.png", "*.jpg", "*.jpeg", "*.webp"])
+        if fallback is None:
+            print(f"Error: Image file not found at {image_path}")
+            return
+        image_path = fallback
+        print(f"Warning: Using fallback image file {image_path}")
     if not audio_path.exists():
-        print(f"Error: Audio file not found at {audio_path}")
-        return
+        fallback = _pick_first(data_dir, ["*.mp3", "*.wav", "*.m4a", "*.flac", "*.ogg"])
+        if fallback is None:
+            print(f"Error: Audio file not found at {audio_path}")
+            return
+        audio_path = fallback
+        print(f"Warning: Using fallback audio file {audio_path}")
 
     with open(image_path, "rb") as f:
         image_data = base64.b64encode(f.read()).decode()
@@ -150,12 +179,16 @@ async def process_multi_modal_analysis():
     # Send request to API
     response = requests.post(
         "http://localhost:8000/v1/process",
-        json=request.dict()
+        json=request.model_dump()
     )
 
     if response.status_code == 200:
         result = response.json()
         print("Multi-modal analysis completed successfully!")
+        if "image_text" in result.get("outputs", {}):
+            print(f"\nImage expert output:\n{result['outputs']['image_text']}")
+        if "audio_text" in result.get("outputs", {}):
+            print(f"\nAudio expert output:\n{result['outputs']['audio_text']}")
         print(f"Analysis: {result['outputs']['text']}")
         print("\nExpert weights:")
         for expert, weight in result['expert_weights'].items():
@@ -165,7 +198,7 @@ async def process_multi_modal_analysis():
 
 if __name__ == "__main__":
     # Create example data directory if it doesn't exist
-    data_dir = Path(os.path.join(os.path.dirname(__file__), "..", "..", "data"))
+    data_dir = Path(__file__).resolve().parents[2] / "data"
     data_dir.mkdir(parents=True, exist_ok=True)
 
     # Run examples