Spaces:
Sleeping
Sleeping
Update main.py
Browse files
main.py
CHANGED
@@ -1,7 +1,7 @@
|
|
1 |
import uvicorn
|
2 |
from fastapi import FastAPI, HTTPException
|
3 |
-
from fastapi.middleware.cors import CORSMiddleware
|
4 |
from pydantic import BaseModel
|
|
|
5 |
from sentence_transformers import SentenceTransformer
|
6 |
from pinecone import Pinecone, ServerlessSpec
|
7 |
import uuid
|
@@ -11,31 +11,25 @@ from contextlib import asynccontextmanager
|
|
11 |
# --- Environment Setup ---
|
12 |
PINECONE_API_KEY = os.getenv("PINECONE_API_KEY")
|
13 |
PINECONE_INDEX_NAME = os.getenv("PINECONE_INDEX_NAME", "memoria-index")
|
14 |
-
|
15 |
-
CACHE_DIR = "/app/model_cache"
|
16 |
|
17 |
-
# --- Global
|
18 |
model = None
|
19 |
pc = None
|
20 |
index = None
|
21 |
|
22 |
@asynccontextmanager
|
23 |
async def lifespan(app: FastAPI):
|
24 |
-
"""
|
25 |
-
Handles startup and shutdown events for the FastAPI app.
|
26 |
-
Loads the model and connects to Pinecone on startup.
|
27 |
-
"""
|
28 |
global model, pc, index
|
29 |
print("Application startup...")
|
30 |
|
31 |
if not PINECONE_API_KEY:
|
32 |
raise ValueError("PINECONE_API_KEY environment variable not set.")
|
33 |
|
34 |
-
# 1. Load the
|
35 |
-
print(
|
36 |
-
# THE FINAL FIX: Explicitly tell the library where to save the model.
|
37 |
model = SentenceTransformer(
|
38 |
-
'sentence-transformers/
|
39 |
cache_folder=CACHE_DIR
|
40 |
)
|
41 |
print("Model loaded.")
|
@@ -44,12 +38,15 @@ async def lifespan(app: FastAPI):
|
|
44 |
print("Connecting to Pinecone...")
|
45 |
pc = Pinecone(api_key=PINECONE_API_KEY)
|
46 |
|
47 |
-
# 3. Get or create the Pinecone index
|
|
|
|
|
|
|
48 |
if PINECONE_INDEX_NAME not in pc.list_indexes().names():
|
49 |
-
print(f"Creating new Pinecone index: {PINECONE_INDEX_NAME}")
|
50 |
pc.create_index(
|
51 |
name=PINECONE_INDEX_NAME,
|
52 |
-
dimension=
|
53 |
metric="cosine",
|
54 |
spec=ServerlessSpec(cloud="aws", region="us-east-1")
|
55 |
)
|
@@ -58,30 +55,18 @@ async def lifespan(app: FastAPI):
|
|
58 |
yield
|
59 |
print("Application shutdown.")
|
60 |
|
61 |
-
#
|
62 |
-
|
63 |
-
# --- Pydantic Models ---
|
64 |
class Memory(BaseModel):
|
65 |
content: str
|
66 |
-
|
67 |
class SearchQuery(BaseModel):
|
68 |
query: str
|
69 |
|
70 |
-
# --- FastAPI App ---
|
71 |
app = FastAPI(
|
72 |
title="Memoria API",
|
73 |
-
|
74 |
-
version="1.0.1", # Final deployed version
|
75 |
lifespan=lifespan
|
76 |
)
|
77 |
-
|
78 |
-
app.add_middleware(
|
79 |
-
CORSMiddleware,
|
80 |
-
allow_origins=["*"],
|
81 |
-
allow_credentials=True,
|
82 |
-
allow_methods=["*"],
|
83 |
-
allow_headers=["*"],
|
84 |
-
)
|
85 |
|
86 |
# --- API Endpoints ---
|
87 |
@app.get("/")
|
@@ -89,28 +74,20 @@ def read_root():
|
|
89 |
return {"status": "ok", "message": "Welcome to the Memoria API!"}
|
90 |
|
91 |
@app.post("/save_memory")
|
92 |
-
def
|
93 |
-
|
94 |
-
|
95 |
-
|
96 |
-
|
97 |
-
|
98 |
-
return {"status": "success", "id": memory_id}
|
99 |
-
except Exception as e:
|
100 |
-
print(f"An error occurred during save: {e}")
|
101 |
-
raise HTTPException(status_code=500, detail=str(e))
|
102 |
|
103 |
@app.post("/search_memory")
|
104 |
-
def
|
105 |
-
|
106 |
-
|
107 |
-
|
108 |
-
|
109 |
-
|
110 |
-
return {"status": "success", "results": retrieved_documents}
|
111 |
-
except Exception as e:
|
112 |
-
print(f"An error occurred during search: {e}")
|
113 |
-
raise HTTPException(status_code=500, detail=str(e))
|
114 |
|
115 |
if __name__ == "__main__":
|
116 |
uvicorn.run("main:app", host="127.0.0.1", port=8000, reload=True)
|
|
|
1 |
import uvicorn
|
2 |
from fastapi import FastAPI, HTTPException
|
|
|
3 |
from pydantic import BaseModel
|
4 |
+
from fastapi.middleware.cors import CORSMiddleware
|
5 |
from sentence_transformers import SentenceTransformer
|
6 |
from pinecone import Pinecone, ServerlessSpec
|
7 |
import uuid
|
|
|
11 |
# --- Environment Setup ---
|
12 |
PINECONE_API_KEY = os.getenv("PINECONE_API_KEY")
|
13 |
PINECONE_INDEX_NAME = os.getenv("PINECONE_INDEX_NAME", "memoria-index")
|
14 |
+
CACHE_DIR = "/app/model_cache" # For Hugging Face caching
|
|
|
15 |
|
16 |
+
# --- Global Objects ---
|
17 |
model = None
|
18 |
pc = None
|
19 |
index = None
|
20 |
|
21 |
@asynccontextmanager
|
22 |
async def lifespan(app: FastAPI):
|
|
|
|
|
|
|
|
|
23 |
global model, pc, index
|
24 |
print("Application startup...")
|
25 |
|
26 |
if not PINECONE_API_KEY:
|
27 |
raise ValueError("PINECONE_API_KEY environment variable not set.")
|
28 |
|
29 |
+
# 1. Load the official, industry-standard lightweight model.
|
30 |
+
print("Loading sentence-transformers/all-MiniLM-L6-v2 model...")
|
|
|
31 |
model = SentenceTransformer(
|
32 |
+
'sentence-transformers/all-MiniLM-L6-v2',
|
33 |
cache_folder=CACHE_DIR
|
34 |
)
|
35 |
print("Model loaded.")
|
|
|
38 |
print("Connecting to Pinecone...")
|
39 |
pc = Pinecone(api_key=PINECONE_API_KEY)
|
40 |
|
41 |
+
# 3. Get or create the Pinecone index with the correct dimension.
|
42 |
+
model_dimension = model.get_sentence_embedding_dimension()
|
43 |
+
print(f"Model dimension is: {model_dimension}")
|
44 |
+
|
45 |
if PINECONE_INDEX_NAME not in pc.list_indexes().names():
|
46 |
+
print(f"Creating new Pinecone index: {PINECONE_INDEX_NAME} with dimension {model_dimension}")
|
47 |
pc.create_index(
|
48 |
name=PINECONE_INDEX_NAME,
|
49 |
+
dimension=model_dimension,
|
50 |
metric="cosine",
|
51 |
spec=ServerlessSpec(cloud="aws", region="us-east-1")
|
52 |
)
|
|
|
55 |
yield
|
56 |
print("Application shutdown.")
|
57 |
|
58 |
+
# --- Pydantic Models & FastAPI App ---
|
|
|
|
|
59 |
class Memory(BaseModel):
|
60 |
content: str
|
|
|
61 |
class SearchQuery(BaseModel):
|
62 |
query: str
|
63 |
|
|
|
64 |
app = FastAPI(
|
65 |
title="Memoria API",
|
66 |
+
version="1.1.0",
|
|
|
67 |
lifespan=lifespan
|
68 |
)
|
69 |
+
app.add_middleware(CORSMiddleware, allow_origins=["*"], allow_credentials=True, allow_methods=["*"], allow_headers=["*"])
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
70 |
|
71 |
# --- API Endpoints ---
|
72 |
@app.get("/")
|
|
|
74 |
return {"status": "ok", "message": "Welcome to the Memoria API!"}
|
75 |
|
76 |
@app.post("/save_memory")
|
77 |
+
def save_memory_endpoint(memory: Memory):
|
78 |
+
embedding = model.encode(memory.content).tolist()
|
79 |
+
memory_id = str(uuid.uuid4())
|
80 |
+
index.upsert(vectors=[{"id": memory_id, "values": embedding, "metadata": {"text": memory.content}}])
|
81 |
+
print(f"Saved memory: {memory_id}")
|
82 |
+
return {"status": "success", "id": memory_id}
|
|
|
|
|
|
|
|
|
83 |
|
84 |
@app.post("/search_memory")
|
85 |
+
def search_memory_endpoint(search: SearchQuery):
|
86 |
+
query_embedding = model.encode(search.query).tolist()
|
87 |
+
results = index.query(vector=query_embedding, top_k=5, include_metadata=True)
|
88 |
+
retrieved_documents = [match['metadata']['text'] for match in results['matches']]
|
89 |
+
print(f"Found {len(retrieved_documents)} results for query: '{search.query}'")
|
90 |
+
return {"status": "success", "results": retrieved_documents}
|
|
|
|
|
|
|
|
|
91 |
|
92 |
if __name__ == "__main__":
|
93 |
uvicorn.run("main:app", host="127.0.0.1", port=8000, reload=True)
|