dierk
66f7db40b0
Three FastAPI backends comparing PostgreSQL/pgvector and Oracle 26ai for semantic image search using CLIP embeddings: Python-side embedding for both databases, plus Oracle in-database embedding via VECTOR_EMBEDDING(CLIP_TXT). Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
50 lines
1.4 KiB
Python
50 lines
1.4 KiB
Python
import os
|
|
import array
|
|
from fastapi import FastAPI, Query
|
|
from fastapi.middleware.cors import CORSMiddleware
|
|
from fastapi.responses import FileResponse
|
|
from dotenv import load_dotenv
|
|
from db_oracle import get_connection
|
|
from embedder import embed_text
|
|
|
|
load_dotenv()
|
|
|
|
PHOTOS_DIR = os.getenv("PHOTOS_DIR")
|
|
|
|
app = FastAPI()
|
|
app.add_middleware(CORSMiddleware, allow_origins=["*"], allow_methods=["*"], allow_headers=["*"])
|
|
|
|
@app.get("/search")
|
|
def search(q: str = Query(...), limit: int = Query(12)):
|
|
vec = array.array("f", embed_text(q))
|
|
conn = get_connection()
|
|
cur = conn.cursor()
|
|
cur.execute(
|
|
"""
|
|
SELECT filename, 1 - VECTOR_DISTANCE(embedding, :vec, COSINE) AS score
|
|
FROM images
|
|
ORDER BY VECTOR_DISTANCE(embedding, :vec, COSINE)
|
|
FETCH FIRST :lim ROWS ONLY
|
|
""",
|
|
{"vec": vec, "lim": limit},
|
|
)
|
|
rows = cur.fetchall()
|
|
cur.close()
|
|
conn.close()
|
|
return [{"filename": r[0], "score": round(r[1], 4)} for r in rows]
|
|
|
|
@app.get("/stats")
|
|
def stats():
|
|
conn = get_connection()
|
|
cur = conn.cursor()
|
|
cur.execute("SELECT COUNT(*) FROM images")
|
|
count = cur.fetchone()[0]
|
|
cur.close()
|
|
conn.close()
|
|
return {"count": count}
|
|
|
|
@app.get("/photos/{filename}")
|
|
def get_photo(filename: str):
|
|
path = os.path.join(PHOTOS_DIR, filename)
|
|
return FileResponse(path, media_type="image/jpeg")
|