Update main.py

main.py

@@ -1,5 +1,5 @@
 import uvicorn
-from fastapi import FastAPI, HTTPException, Depends
+from fastapi import FastAPI, HTTPException
 from fastapi.middleware.cors import CORSMiddleware
 from pydantic import BaseModel
 from sentence_transformers import SentenceTransformer
@@ -9,13 +9,12 @@ import os
 from contextlib import asynccontextmanager
 
 # --- Environment Setup ---
-# It's best practice to get sensitive keys from environment variables
-# We will set these up in Hugging Face Spaces Secrets
 PINECONE_API_KEY = os.getenv("PINECONE_API_KEY")
 PINECONE_INDEX_NAME = os.getenv("PINECONE_INDEX_NAME", "memoria-index")
+# Define a writable cache directory inside our container
+CACHE_DIR = "/app/model_cache"
 
 # --- Global objects ---
-# We load these once at startup to save time and memory
 model = None
 pc = None
 index = None
@@ -33,8 +32,12 @@ async def lifespan(app: FastAPI):
         raise ValueError("PINECONE_API_KEY environment variable not set.")
 
     # 1. Load the AI Model
-    print("Loading lightweight sentence transformer model...")
-    model = SentenceTransformer('sentence-transformers/paraphrase-albert-small-v2')
+    print(f"Loading model and setting cache to: {CACHE_DIR}")
+    # THE FINAL FIX: Explicitly tell the library where to save the model.
+    model = SentenceTransformer(
+        'sentence-transformers/paraphrase-albert-small-v2',
+        cache_folder=CACHE_DIR
+    )
     print("Model loaded.")
 
     # 2. Connect to Pinecone
@@ -47,15 +50,16 @@ async def lifespan(app: FastAPI):
         pc.create_index(
             name=PINECONE_INDEX_NAME,
             dimension=model.get_sentence_embedding_dimension(),
-            metric="cosine", # Cosine similarity is great for sentence vectors
+            metric="cosine",
             spec=ServerlessSpec(cloud="aws", region="us-east-1")
         )
     index = pc.Index(PINECONE_INDEX_NAME)
     print("Pinecone setup complete.")
     yield
-    # Cleanup logic can go here if needed on shutdown
     print("Application shutdown.")
 
+# ... (The rest of the file remains exactly the same) ...
+
 # --- Pydantic Models ---
 class Memory(BaseModel):
     content: str
@@ -67,13 +71,13 @@ class SearchQuery(BaseModel):
 app = FastAPI(
     title="Memoria API",
     description="API for storing and retrieving memories.",
-    version="1.0.0",
-    lifespan=lifespan # Use the lifespan context manager
+    version="1.0.1", # Final deployed version
+    lifespan=lifespan
 )
 
 app.add_middleware(
     CORSMiddleware,
-    allow_origins=["*"], # Allow all origins for simplicity
+    allow_origins=["*"],
     allow_credentials=True,
     allow_methods=["*"],
     allow_headers=["*"],
@@ -89,10 +93,7 @@ def save_memory(memory: Memory):
     try:
         embedding = model.encode(memory.content).tolist()
         memory_id = str(uuid.uuid4())
-        
-        # Upsert (update or insert) the vector into Pinecone
         index.upsert(vectors=[{"id": memory_id, "values": embedding, "metadata": {"text": memory.content}}])
-        
         print(f"Successfully saved memory with ID: {memory_id}")
         return {"status": "success", "id": memory_id}
     except Exception as e:
@@ -103,13 +104,8 @@ def save_memory(memory: Memory):
 def search_memory(search: SearchQuery):
     try:
         query_embedding = model.encode(search.query).tolist()
-        
-        # Query Pinecone for the most similar vectors
         results = index.query(vector=query_embedding, top_k=5, include_metadata=True)
-        
-        # Extract the original text from the metadata
         retrieved_documents = [match['metadata']['text'] for match in results['matches']]
-        
         print(f"Found {len(retrieved_documents)} results for query: '{search.query}'")
         return {"status": "success", "results": retrieved_documents}
     except Exception as e:
@@ -117,7 +113,4 @@ def search_memory(search: SearchQuery):
         raise HTTPException(status_code=500, detail=str(e))
 
 if __name__ == "__main__":
-    uvicorn.run("main:app", host="127.0.0.1", port=8000, reload=True)
-
-
-
+    uvicorn.run("main:app", host="127.0.0.1", port=8000, reload=True)