s3vectors_demo.sh

#!/bin/bash

set -e

RGW_ENDPOINT="http://localhost:8000"
VECTOR_BUCKET="demo-vector-bucket"
INDEX_NAME="demo-index"
DIMENSION=4

# --------------------------------------------------------------------------
# 1. Create a Vector Bucket
# --------------------------------------------------------------------------
aws --endpoint-url $RGW_ENDPOINT s3vectors create-vector-bucket \
    --vector-bucket-name "$VECTOR_BUCKET" | jq .

# --------------------------------------------------------------------------
# 2. Create an Index
# --------------------------------------------------------------------------
aws --endpoint-url $RGW_ENDPOINT s3vectors create-index \
    --vector-bucket-name "$VECTOR_BUCKET" \
    --index-name "$INDEX_NAME" \
    --data-type float32 \
    --dimension $DIMENSION \
    --distance-metric cosine | jq .

# --------------------------------------------------------------------------
# 3. Upload 20 vectors
# --------------------------------------------------------------------------
VECTORS="["
for i in $(seq 1 20); do
    v1="0.$((i * 17 % 100))"
    v2="0.$((i * 31 % 100))"
    v3="0.$((i * 47 % 100))"
    v4="0.$((i * 61 % 100))"

    ENTRY="{\"key\": \"vec-$(printf '%03d' $i)\", \"data\": {\"float32\": [$v1, $v2, $v3, $v4]}, \"metadata\": \"{\\\"label\\\": \\\"item-$i\\\"}\"}"
    if [ "$i" -lt 20 ]; then
        ENTRY="$ENTRY,"
    fi
    VECTORS="$VECTORS $ENTRY"
done
VECTORS="$VECTORS ]"

aws --endpoint-url $RGW_ENDPOINT s3vectors put-vectors \
    --vector-bucket-name "$VECTOR_BUCKET" \
    --index-name "$INDEX_NAME" \
    --vectors "$VECTORS"

# --------------------------------------------------------------------------
# 4. Query vectors _ find top 10 nearest to a probe vector
# --------------------------------------------------------------------------
aws --endpoint-url $RGW_ENDPOINT s3vectors query-vectors \
    --vector-bucket-name "$VECTOR_BUCKET" \
    --index-name "$INDEX_NAME" \
    --query-vector '{"float32": [0.5, 0.5, 0.5, 0.5]}' \
    --top-k 10 \
    --return-metadata \
    --return-distance | jq .

# --------------------------------------------------------------------------
# 5. List all vectors in the index
# --------------------------------------------------------------------------
RESULT=$(aws --endpoint-url $RGW_ENDPOINT s3vectors list-vectors \
    --vector-bucket-name "$VECTOR_BUCKET" \
    --index-name "$INDEX_NAME" \
    --return-metadata \
    --return-data)
echo "$RESULT" | jq .

KEYS=$(echo "$RESULT" | jq -r '[.vectors[].key]')
echo "Found $(echo "$KEYS" | jq 'length') vectors to delete"

# --------------------------------------------------------------------------
# 6. Delete vectors based on the list
# --------------------------------------------------------------------------
aws --endpoint-url $RGW_ENDPOINT s3vectors delete-vectors \
    --vector-bucket-name "$VECTOR_BUCKET" \
    --index-name "$INDEX_NAME" \
    --keys "$KEYS"

# --------------------------------------------------------------------------
# 7. Delete the index
# --------------------------------------------------------------------------
aws --endpoint-url $RGW_ENDPOINT s3vectors delete-index \
    --vector-bucket-name "$VECTOR_BUCKET" \
    --index-name "$INDEX_NAME"

# --------------------------------------------------------------------------
# 8. Delete the vector bucket
# --------------------------------------------------------------------------
aws --endpoint-url $RGW_ENDPOINT s3vectors delete-vector-bucket \
    --vector-bucket-name "$VECTOR_BUCKET"