Aliyun Qwen Rerank

Use when reranking search candidates is needed with Alibaba Cloud Model Studio rerank models, including hybrid retrieval, top-k refinement, and multilingual...

Security Scan

The skill is internally consistent: it only prepares a local JSON payload for Alibaba Cloud Model Studio rerank, with no credentials, network calls, or surprising behaviors.

Purpose & Capability

The name/description promise (reranking with Alibaba Cloud Model Studio rerank models) matches the files: SKILL.md documents usage and the script constructs a rerank request payload and lists valid model names. There are no unrelated environment variables, binaries, or config paths requested.

Instruction Scope

SKILL.md instructions and the included Python script are narrowly scoped: they validate Python compilation, create an output directory, and write a minimal request.json containing query, model, and candidate documents. The instructions do not read other system files, call external endpoints, or access environment variables.

Install Mechanism

There is no install spec (instruction-only with a small script). Nothing is downloaded or extracted, and no external packages or package manager installs are requested.

Credentials

The skill declares no required environment variables or credentials. The behavior (preparing a local payload) does not require secrets. Note: actually sending the payload to Alibaba Cloud would require credentials, but this skill does not perform that network call or request those credentials.

Persistence & Privilege

always is false and the skill doesn't request persistent privileges or modify other skills or system-wide settings. It only writes its own output file under output/aliyun-qwen-rerank.

Assessment

This skill is lightweight and simply builds a local request.json for Alibaba Cloud rerank models — it does not contact any external service or request credentials itself. Before using it in a pipeline, verify the component that will send the request to Model Studio: that component will need Alibaba Cloud credentials and will actually transmit your queries/documents (so avoid including sensitive secrets or PII in documents unless you trust that connector). If you expect the skill to perform the network call, note it currently does not; you'll need a separate, authenticated integration to call Model Studio.

Category: provider

Model Studio Rerank

Validation

mkdir -p output/aliyun-qwen-rerank
python -m py_compile skills/ai/search/aliyun-qwen-rerank/scripts/prepare_rerank_request.py && echo "py_compile_ok" > output/aliyun-qwen-rerank/validate.txt

Pass criteria: command exits 0 and output/aliyun-qwen-rerank/validate.txt is generated.

Critical model names

Use one of these exact model strings:

• gte-rerank-v2
• gte-rerank
• gte-multilingual-rerank
• qwen3-reranker-8b
• qwen3-reranker-4b
• qwen3-reranker-0.6b

Quick start

python skills/ai/search/aliyun-qwen-rerank/scripts/prepare_rerank_request.py \
  --query "cloud vector database" \
  --output output/aliyun-qwen-rerank/request.json

Notes

• Use after embedding/vector retrieval to reorder candidates.
• Prefer multilingual rerankers when query/document languages differ.

References

• references/sources.md