Alibaba Adb Knowledge

ADB Knowledge

Run aliyun commands directly. Do not create, invoke, or depend on a wrapper script.

Read README.md when the user needs CLI installation, credential setup, permissions, or a complete end-to-end walkthrough.

End-to-end workflow

Every task that involves uploading or ingesting files must follow these four steps in strict order:

1. Environment check — Verify the CLI, ADB plugin, and required parameters are available (see "Install and verify the CLI").
2. Bucket selection — Execute list-lake-storages, extract Items[].LakeStorageId, present those values to the user, and let them choose exactly one bucket. This is a mandatory prerequisite; do not skip it or assume a bucket.
3. Upload + Ingest — Upload the file(s) to OSS → verify with aliyun oss stat → immediately execute add-knowledge-file. These three actions are a single atomic sequence; do not wait for a separate user instruction between them.
4. Recall — get-knowledge-recall with full raw output returned verbatim.

Step 2 (Bucket selection) is an independent prerequisite that must complete before any upload or ingestion operation begins.

Path validation

Every OSS target path (for upload) and every --file-location value (for ingestion) must start with oss://adb-lake-. This ensures only lake-storage buckets are used.

Before executing aliyun oss cp or aliyun adb add-knowledge-file, validate the resolved OSS URI:

• ✅ oss://adb-lake-cn-beijing-9746ll8740/knowledge/report.pdf — valid
• ❌ oss://my-custom-bucket/knowledge/report.pdf — rejected

If the path does not match the oss://adb-lake- prefix, stop immediately and inform the user that only lake-storage buckets (prefixed with adb-lake-) are supported. Do not proceed with upload or ingestion.

Constants and configuration

Use API version 2021-12-01 for every ADB knowledge command.

Require the ADB cluster ID. Prefer ADB_INSTANCE_ID when it is available:

export ADB_INSTANCE_ID="amv-xxxxxxxxxxxxxx"

Resolve the business region for lake-bucket discovery automatically from the active aliyun CLI profile with aliyun configure get region. Do not ask the user to supply a separate business-region value when the CLI region is configured. This command returns only the configured Region; do not use bare aliyun configure get, which may expose credentials.

Treat the standard region override as an optional customer-supplied value for ingestion and recall:

export ADB_REGION="<region-id>"

When constructing any ADB command:

• Append --region "$ADB_REGION" only when ADB_REGION is non-empty.
• Omit --region entirely when the variable is absent.
• Do not add --endpoint in the customer skill.
• Do not invent a region default.

Install and verify the CLI

Install the ADB plugin from the public plugin registry:

aliyun plugin install --name aliyun-cli-adb

Update an existing installation when necessary:

aliyun plugin update --name aliyun-cli-adb

Verify the required commands and parameters:

aliyun version
aliyun plugin show --name aliyun-cli-adb
aliyun adb list-lake-storages --api-version 2021-12-01 --help
aliyun adb add-knowledge-file --api-version 2021-12-01 --help
aliyun adb get-knowledge-recall --api-version 2021-12-01 --help

Require list-lake-storages to expose --biz-region-id, --page-number, and --page-size. Require add-knowledge-file to expose --is-dir and --upload-user. Require get-knowledge-recall to expose --topk and the global --read-timeout option.

Resolve the lake-storage bucket

Complete this workflow before every upload or add-knowledge-file request:

1. Check OSS_BUCKET first.
2. If OSS_BUCKET is non-empty, use it as the preconfigured lake-storage bucket. It must have been selected from list-lake-storages; never accept a custom bucket name.
3. If OSS_BUCKET is empty, resolve the business region automatically from the active aliyun CLI profile:

BIZ_REGION_ID="$(aliyun configure get region 2>/dev/null)"

4. If BIZ_REGION_ID is empty, stop before calling list-lake-storages and ask the user to configure the CLI default region once with aliyun configure set --region <region-id>.
5. Issue the following command without --endpoint:

aliyun adb list-lake-storages \
  --api-version 2021-12-01 \
  --db-cluster-id "$ADB_INSTANCE_ID" \
  --biz-region-id "$BIZ_REGION_ID"

Accept the response only when Success is true, Code is ok, and HttpStatusCode is 200. Extract each non-empty Items[].LakeStorageId; every extracted LakeStorageId is a selectable OSS bucket name. Do not use RegionId, account, permission, or other response fields as the bucket name.

Use PageNumber, PageSize, and TotalCount to detect more pages. While PageNumber * PageSize < TotalCount, request the next page by appending --page-number <next-page-number> --page-size <current-page-size> and collect its Items[].LakeStorageId. Show the complete collected list and ask the user to choose exactly one. Set OSS_BUCKET to the selected LakeStorageId only; never synthesize or accept a value that is not in the list.

If any success-status check fails, stop and report the response Code and Message. If all pages contain no non-empty Items[].LakeStorageId, treat the lake-storage list as empty.

If the command fails, stop and return the error before uploading or reporting ingestion. If the list is empty, stop before upload or ingestion and direct the user to the AnalyticDB for MySQL console: open the target cluster, choose Data Management > Lake Storage Management, create lake storage, and add read/write authorization for the current RAM user or role. Do not continue until a lake bucket is available.

If the user supplies an existing oss:// URI, require its bucket component to match the resolved OSS_BUCKET.

Upload local content to OSS

Path prefix constraint: the target OSS path must start with oss://adb-lake-; see Path validation.

Upload one file:

aliyun oss cp ./report.pdf "oss://$OSS_BUCKET/knowledge/report.pdf"

Upload a directory recursively:

aliyun oss cp ./data/ "oss://$OSS_BUCKET/knowledge/data/" --recursive

Verify a known object with aliyun oss stat. Do not rely on bucket listing permission being available.

After upload and stat verification succeed, immediately proceed to execute add-knowledge-file (see next section). Do not wait for a separate user instruction. The sequence upload → stat → add-knowledge-file is a single atomic flow.

Add one knowledge file

Path prefix constraint: --file-location must start with oss://adb-lake-; see Path validation.

Require a complete oss:// location. Add optional parameters only when the user supplies values:

aliyun adb add-knowledge-file \
  --api-version 2021-12-01 \
  --db-cluster-id "$ADB_INSTANCE_ID" \
  --file-location "oss://$OSS_BUCKET/knowledge/report.pdf" \
  --file-type pdf \
  --upload-user alice

Omit --file-type, --upload-user, or --region when the corresponding value is absent. Do not pass --is-dir false; omission preserves single-file behavior.

Treat add-knowledge-file as non-idempotent. Never retry it automatically.

Add an OSS directory

Require the directory itself to use an oss:// URI. Set --is-dir true explicitly:

aliyun adb add-knowledge-file \
  --api-version 2021-12-01 \
  --db-cluster-id "$ADB_INSTANCE_ID" \
  --file-location "oss://$OSS_BUCKET/knowledge/data/" \
  --is-dir true \
  --upload-user alice

Do not expand a directory into repeated file-level ingestion calls unless the user explicitly requests that behavior.

Recall knowledge

Issue exactly one CLI request. Always pass the 300-second timeout as the CLI parameter --read-timeout 300:

aliyun adb get-knowledge-recall \
  --api-version 2021-12-01 \
  --read-timeout 300 \
  --db-cluster-id "$ADB_INSTANCE_ID" \
  --question "是否有产品授权书"

Append --topk <positive-integer> only when requested. Append --region only when the customer supplies it.

Do not retry recall automatically, including on EOF, timeout, 429/5xx, or empty output. Wait up to the CLI timeout unless the user requests cancellation.

Return CLI stdout, stderr, and the exit status verbatim. Do not parse, normalize, summarize, redact, or assume any response fields because the response schema may change.

Optional region example

When the customer explicitly provides a region, append it to the command:

aliyun adb get-knowledge-recall \
  --api-version 2021-12-01 \
  --read-timeout 300 \
  --db-cluster-id "$ADB_INSTANCE_ID" \
  --question "是否有产品授权书" \
  --region "$ADB_REGION"

Final response

For upload, verification, and ingestion operations, return only whether the user-requested file or directory succeeded. Use exactly one applicable one-line shape:

• 上传成功：<用户请求的文件或目录>。
• 入库成功：<用户请求的文件或目录>。
• 上传并入库成功：<用户请求的文件或目录>。
• 上传失败：<用户请求的文件或目录>。
• 入库失败：<用户请求的文件或目录>。

Use the path or label from the user's request; never embed a concrete case in this skill. When add-knowledge-file exits successfully and returns {}, treat ingestion as successful and suppress the {} response body.

Do not include environment checks, CLI/plugin versions, region, bucket-selection steps, object counts, file-size observations, OSS metadata, response bodies, exit status, asynchronous-indexing explanations, warnings, full commands, or follow-up offers. On failure, append one short error reason only when it is necessary for the user's next action.

• Do not apply this compact-summary rule to recall. Recall must continue to return the complete raw stdout, stderr, and exit status as specified above.

Safety

• Never expose AccessKey secrets or STS tokens.
• Redact the OSS bucket only in upload or ingestion summaries; do not modify raw recall output.
• Treat successful ingestion as request acceptance without claiming asynchronous indexing has completed.
• Use --cli-dry-run when command construction needs inspection without sending a request.