Optimizing Index Performance

# target 1500 QPS on p1.x1 (~100 QPS each)
required_replicas = math.ceil(1500 / 100)  # 15
pc.configure_index("docs", replicas=15)

from pinecone import PineconeAsyncio
import asyncio

async def upsert_all(batches):
    async with PineconeAsyncio(api_key="...") as pc:
        async with pc.IndexAsyncio(name="docs") as idx:
            await asyncio.gather(*(idx.upsert(vectors=b) for b in batches))

import hashlib, json, redis

r = redis.Redis()
def cached_query(qv, top_k, ns):
    key = "q:" + hashlib.sha256(
        json.dumps([qv, top_k, ns]).encode()).hexdigest()
    if (v := r.get(key)): return json.loads(v)
    res = index.query(vector=qv, top_k=top_k, namespace=ns)
    r.set(key, json.dumps(res), ex=300)
    return res

import time

def timed(fn, *args, **kw):
    t = time.perf_counter()
    res = fn(*args, **kw)
    print(f"{fn.__name__}: {(time.perf_counter()-t)*1000:.1f}ms")
    return res

timed(index.query, vector=qv, top_k=10)