more minor bux fixedd

requirements.txt

@@ -11,4 +11,6 @@ numpy
 python-dotenv
 together
 einops
-opencv-python
+opencv-python
+timm
+json

src/embedder.py

@@ -6,8 +6,8 @@ import decord
 class InternVLEmbedder:
     def __init__(self):
         self.device = "cuda" if torch.cuda.is_available() else "cpu"
-        self.model = AutoModel.from_pretrained("OpenGVLab/InternVL2_5-8B-MPO", trust_remote_code=True).to(self.device)
-        self.processor = AutoProcessor.from_pretrained("OpenGVLab/InternVL2_5-8B-MPO")
+        self.model = AutoModel.from_pretrained("OpenGVLab/InternVL2_5-1B-MPO", trust_remote_code=True).to(self.device)
+        self.processor = AutoProcessor.from_pretrained("OpenGVLab/InternVL2_5-1B-MPO", trust_remote_code=True)
 
     def embed_video(self, video_path):
         vr = decord.VideoReader(video_path)

src/pinecone_store.py

@@ -1,4 +1,4 @@
-import pinecone
+from pinecone import Pinecone, ServerlessSpec
 import os
 from dotenv import load_dotenv
 
@@ -6,27 +6,24 @@ load_dotenv()
 
 class PineconeStore:
     def __init__(self):
-        api_key = os.getenv("PINECONE_API_KEY")
-        environment = os.getenv("PINECONE_ENV")
-        pinecone.init(api_key=api_key, environment=environment)
+        self.pc = Pinecone(api_key=os.getenv("PINECONE_API_KEY"))
 
-        self.index_name = "soccer-highlights"
-        if self.index_name not in pinecone.list_indexes():
-            pinecone.create_index(
-                name=self.index_name,
-                dimension=1024,  # Ensure this matches your embedding model's output dimension
-                metric="cosine"  # Choose the appropriate metric (e.g., cosine, euclidean)
+        if 'blip-soccer-highlights' not in self.pc.list_indexes().names():
+            self.pc.create_index(
+                name='blip-soccer-highlights',
+                dimension=1024,  # This must match InternVL2_5-8B-MPO
+                metric='cosine',
+                spec=ServerlessSpec(cloud='aws', region=os.getenv("PINECONE_ENV", "us-east-1"))
             )
-        self.index = pinecone.Index(self.index_name)
+
+        self.index = self.pc.Index('blip-soccer-highlights')
 
     def upsert(self, id, vector, metadata):
         self.index.upsert([(id, vector.tolist(), metadata)])
 
     def query(self, vector, filter_key, top_k):
+        results = self.index.query(vector.tolist(), top_k=top_k, include_metadata=True)
         return [
-            m["metadata"] for m in self.index.query(
-                vector.tolist(),
-                top_k=top_k,
-                include_metadata=True
-            )["matches"] if filter_key in m["id"]
+            match["metadata"] for match in results["matches"]
+            if filter_key in match["id"]
         ]