AliCloud Milvus (Serverless) via PyMilvus

---

name: alicloud-ai-search-milvus

description: Use AliCloud Milvus (serverless) with PyMilvus to create collections, insert vectors, and run filtered similarity search. Optimized for Claude Code/Codex vector retrieval flows.

version: 1.0.0

---

Category: provider

# AliCloud Milvus (Serverless) via PyMilvus

This skill uses standard PyMilvus APIs to connect to AliCloud Milvus and run vector search.

Prerequisites

python3 -m venv .venv
. .venv/bin/activate
python -m pip install --upgrade pymilvus

- `MILVUS_URI` (e.g. `http://<host>:19530`)

- `MILVUS_TOKEN` (`<username>:<password>`)

- `MILVUS_DB` (default: `default`)

Quickstart (Python)

import os
from pymilvus import MilvusClient

client = MilvusClient(
    uri=os.getenv("MILVUS_URI"),
    token=os.getenv("MILVUS_TOKEN"),
    db_name=os.getenv("MILVUS_DB", "default"),
)

# 1) Create a collection
client.create_collection(
    collection_name="docs",
    dimension=768,
)

# 2) Insert data
items = [
    {"id": 1, "vector": [0.01] * 768, "source": "kb", "chunk": 0},
    {"id": 2, "vector": [0.02] * 768, "source": "kb", "chunk": 1},
]
client.insert(collection_name="docs", data=items)

# 3) Search
query_vectors = [[0.01] * 768]
res = client.search(
    collection_name="docs",
    data=query_vectors,
    limit=5,
    filter='source == "kb" and chunk >= 0',
    output_fields=["source", "chunk"],
)
print(res)

Script quickstart

python skills/ai/search/alicloud-ai-search-milvus/scripts/quickstart.py

Environment variables:

Optional args: `--collection`, `--dimension`, `--limit`, `--filter`.

Notes for Claude Code/Codex

Error handling

Validation

mkdir -p output/alicloud-ai-search-milvus
for f in skills/ai/search/alicloud-ai-search-milvus/scripts/*.py; do
  python3 -m py_compile "$f"
done
echo "py_compile_ok" > output/alicloud-ai-search-milvus/validate.txt

Pass criteria: command exits 0 and `output/alicloud-ai-search-milvus/validate.txt` is generated.

Output And Evidence

Workflow

1) Confirm user intent, region, identifiers, and whether the operation is read-only or mutating.

2) Run one minimal read-only query first to verify connectivity and permissions.

3) Execute the target operation with explicit parameters and bounded scope.

4) Verify results and save output/evidence files.

References