alibabacloud-rds-copilot

Alibaba Cloud RDS Copilot Intelligent Operations Assistant

This skill serves as an intelligent agent for Alibaba Cloud RDS Copilot in conversations, helping users with RDS-related intelligent Q&A, SQL optimization, instance operations, and troubleshooting.

Scenario Description

Architecture: Alibaba Cloud CLI + RdsAi OpenAPI

Main features:

• Understand user's natural language requests (Chinese or English), identify if related to RDS Copilot
• Directly call Alibaba Cloud CLI to execute aliyun rdsai chat-messages command for real-time RDS Copilot queries
• When receiving results or user-pasted error messages, further explain, diagnose, and provide recommendations

Agent Execution Contract

Before calling RDS Copilot, the agent must make the local environment ready by itself whenever command execution is available. Do not ask the user to install CLI or plugins first if the agent can install or upgrade them directly. If install or upgrade needs network access, sudo, or tool approval, request that approval with the concrete command and continue after approval.

Hard gate: Do not execute aliyun rdsai chat-messages until all readiness checks pass:

1. Alibaba Cloud CLI exists and aliyun version is >= 3.3.3.
2. Plugin auto-install is enabled and the rdsai product plugin/command is available.
3. Alibaba Cloud CLI credentials are configured and the selected profile is valid.

If any gate fails:

• Missing or old CLI: install or upgrade Alibaba Cloud CLI, then re-run aliyun version.
• Missing rdsai plugin: enable plugin auto-install and install or trigger installation of rdsai, then re-check the command.
• Missing/invalid credentials: stop before the RDS Copilot API call and guide the user to configure credentials. If the user asks the agent to configure them, request only the required fields, then configure through aliyun configure; never ask the user to paste secrets into environment variables.

Default region: If the user does not explicitly provide a region, always use cn-hangzhou. Natural language "Hangzhou" maps to cn-hangzhou. Keep this default in both --inputs RegionId=cn-hangzhou and any credential setup guidance unless the user specifies another region.

Installation

Agent preflight: Alibaba Cloud CLI must be installed or installed by the agent

This skill uses Alibaba Cloud CLI to call RdsAi OpenAPI. The agent must first check whether the CLI is present and usable. If the CLI is missing and command execution is available, install it before asking the user to take manual action.

Pre-check: Aliyun CLI >= 3.3.3 required

Run command -v aliyun and aliyun version to verify >= 3.3.3. If not installed or version is too low, install or update it before proceeding. CLI versions below 3.3.0 do not support the aliyun plugin command; upgrade the CLI first instead of running plugin commands repeatedly.

Pre-check: Aliyun CLI plugin update required

[MUST] run aliyun configure set --auto-plugin-install true to enable automatic plugin installation. [MUST] verify the rdsai plugin capability with aliyun plugin list, aliyun plugin search rdsai, or aliyun rdsai --help. [MUST] if rdsai is not installed, run aliyun plugin install --names rdsai and re-check. [MUST] run aliyun plugin update --name rdsai when the plugin is already installed; if the CLI only supports aliyun plugin update, use that and continue.

[MUST] CLI User-Agent — Every aliyun rdsai chat-messages invocation must include a per-command --user-agent value in this format: AlibabaCloud-Agent-Skills/alibabacloud-rds-copilot/{session-id}. See the Observability section for session-id generation rules. The auxiliary aliyun rdsai get-chat-model command, when used, follows the same User-Agent rule. Do not pass --user-agent to system commands such as aliyun configure, aliyun plugin, aliyun version, or install/upgrade commands.

macOS Installation

# Option 1: Install via Homebrew (recommended)
brew install aliyun-cli
 
# Option 2: Install via PKG package
curl -O https://aliyuncli.alicdn.com/aliyun-cli-latest.pkg
sudo installer -pkg

Linux Installation

Verify Installation

Credential Configuration

Option 1: Interactive Configuration (Recommended)

Follow the prompts to enter:

• Access Key Id: Your AccessKey ID
• Access Key Secret: Your AccessKey Secret
• Default Region Id: cn-hangzhou unless the user specifies another region

Before asking the user for secrets, explain that the recommended interactive command keeps the secret inside Alibaba Cloud CLI's credential store. Do not print or store the secret in chat history longer than needed.

Option 2: Non-interactive Configuration

Command Format

Basic Command Structure

Parameter Description

IMPORTANT: Parameter Confirmation — Before executing any command, Determine user intent: SQL writing/optimization, SQL diagnosis, instance parameter tuning, troubleshooting, performance analysis, query instance list, etc. Collect necessary parameters (use default values if not specified).

RAM Permissions

This skill requires the following RAM permissions. See references/ram-policies.md for details.

Core Workflow

0. Environment Readiness Preflight (Before Calling RDS Copilot)

Run this preflight once at the start of every RDS Copilot task. Do not skip it because the user only asked a simple query.

If aliyun is missing, install it by choosing the safest available local method:

If aliyun version is lower than 3.3.3, upgrade it before running plugin commands:

After install or upgrade, re-run aliyun version.

If aliyun plugin returns an error such as 'plugin' is not a valid command or product, the CLI is too old. Upgrade the CLI first; do not keep retrying plugin commands.

Treat these credential states as not ready:

• No profile is listed.
• The selected or default profile has empty Credential.
• Valid is Invalid or not shown as valid.
• The command reports unknown profile, InvalidAccessKeyId, SignatureDoesNotMatch, missing AccessKey, expired STS token, or similar authentication errors.

When credentials are not ready, stop before aliyun rdsai chat-messages and answer with credential guidance:

If the user wants the agent to configure credentials, ask for:

• Credential mode: default to AK unless the user says OAuth, STS, RAM role, EcsRamRole, or another supported mode.
• Profile name: default to rdsai.
• AccessKeyId and AccessKeySecret for AK mode, or the mode-specific fields for STS/RAM-role modes.
• STS token if using temporary credentials.
• RegionId: default cn-hangzhou if not specified.

Never use export ALIBABA_CLOUD_ACCESS_KEY_ID=... or export ALIBABA_CLOUD_ACCESS_KEY_SECRET=... as the normal setup path. Use aliyun configure so credentials stay in the CLI credential store.

1. Confirm Task Type and Parameters

Determine user intent: SQL writing/optimization, SQL diagnosis, instance parameter tuning, troubleshooting, performance analysis, query instance list, etc.

Collect necessary parameters (use default values if not specified):

• RegionId: Region ID (default cn-hangzhou; use this if omitted)
• Language: Language (default zh-CN)
• Timezone: Timezone (default Asia/Shanghai)
• CustomAgentId: Custom Agent ID (optional)
• EnableThinking: Deep thinking mode switch, true or false (optional; omit unless the user asks for deep thinking or a quick answer)
• ThinkEffort: Thinking depth (optional; default auto, high deeper reasoning for complex diagnosis/optimization, low faster responses; the values each model supports are returned in ThinkingLevels by get-chat-model)
• ModelId: Model ID (optional; use only when the user explicitly specifies a model). When explicitly set, [MUST] first call the auxiliary GetChatModel API and confirm the ID appears in Data[].ModelId before calling ChatMessages — this applies to every user-specified model ID, including well-known ones such as qwen3.8-max; never skip the check or rely on a memorized/hardcoded model list. Omitting ModelId needs no lookup: the service then uses the model marked Default=true in the GetChatModel response. Forward compatibility: new model IDs appear in the GetChatModel response automatically once the API supports them — verify them with GetChatModel, then pass them through; no skill configuration or code change is required
• --conversation-id: Conversation ID for multi-turn dialogue (optional)

GetChatModel is an auxiliary check of the main call flow: ChatMessages itself works without it, but the check becomes [MUST] before aliyun rdsai chat-messages whenever any of the following holds:

1. The user explicitly specifies a model ID — any model ID, including well-known ones such as qwen3.8-max. Verify that the ID appears in Data[].ModelId. Never skip this check because you believe the model is already supported; never rely on a memorized or hardcoded model list.
2. The user asks which models are available.
3. The user requests deep thinking or a specific thinking depth (EnableThinking/ThinkEffort) — supported ThinkEffort values are model-specific and must be read from ThinkingLevels in the get-chat-model response.

Only when none of the above holds (no model specified, no model-list question, no thinking-depth request) may you skip get-chat-model:

The response Data array lists every currently supported model with ModelId, Default, ThinkingLevels, Features, and ContextWindow. If the user's model ID is not in Data[].ModelId, show the available models and ask the user to pick one.

2. Construct Command and Call CLI

Only run this step after the environment readiness preflight succeeds, and — when the user specified a model ID or thinking depth — only after the mandatory aliyun rdsai get-chat-model verification succeeded. Always include RegionId; if the user does not provide a region, use cn-hangzhou. Before the first business API call, generate one session-id using the Observability rules below and reuse it for all aliyun rdsai chat-messages calls in this RDS Copilot task.

3. Parse Results and Follow-up Processing

• Only when the response or user-pasted error contains No valid order found, stop normal result analysis and directly return this guidance to the user:
  • The current call did not find a valid RDS AI Assistant Professional Edition order. Enable RDS AI Assistant Professional Edition for the current Alibaba Cloud account, then retry the failed call.
  • Activation page: https://rdsnext.console.aliyun.com/rdsCopilotProfessional/cn-hangzhou
  • Operation guide: https://help.aliyun.com/zh/rds/apsaradb-rds-for-mysql/manage-rds-colipot-professional-edition
  • According to the operation guide, create RDS AI Assistant Professional Edition from the RDS console by choosing RDS AI Assistant > Professional Edition and clicking Activate Now. After activation succeeds, retry the failed call with the enabled account.
• When the response fails with another error code, map it to user-facing guidance before retrying:
  • ContentModeration (HTTP 451): the request contains non-compliant content; ask the user to rephrase the query.
  • InvalidParameter (HTTP 400): one or more parameters are invalid; check the --inputs values such as EnableThinking, ThinkEffort, and ModelId, fix them, then retry. If an unsupported ModelId is the cause, run aliyun rdsai get-chat-model to fetch the current supported models, guide the user to pick one from the list, or contact technical support to confirm the latest supported models.
  • UserNotFound (HTTP 404): the user does not exist; verify the CLI credential configuration and the account.
  • TooManyRequests (HTTP 429): the maximum concurrent request threshold is exceeded; wait before retrying, or suggest purchasing an AI capacity pack to raise the concurrency limit.
  • MoQuotaExceeded (HTTP 429): the MO token quota of the routed Standard instance is exhausted and the request is rate limited; wait for quota recovery or suggest purchasing a capacity pack.
• Explain RDS Copilot's response to the user in natural language
• If the response contains SQL or operational steps, assess risks and warn:
  • Avoid executing high-risk statements directly in production (e.g., large table DELETE / UPDATE / schema changes)
  • Recommend validating in test environment or adding backup/condition restrictions
• If continuing the conversation, record the ConversationId from the response for the next query

Observability

Generate one session-id per RDS Copilot task/session before the first business API call. The session-id must be a 32-character lowercase hexadecimal string, generated once and reused for every aliyun rdsai chat-messages command in the same task, including multi-turn follow-ups. The auxiliary aliyun rdsai get-chat-model command, when used, reuses the same session-id.

Recommended generation methods:

Fallback:

User-Agent template:

Apply this User-Agent only as a per-command --user-agent argument on business API commands:

Do not use deprecated global User-Agent configuration mechanisms. Do not add --user-agent to system commands such as aliyun configure, aliyun plugin, aliyun version, command -v aliyun, install commands, or upgrade commands.

Output Format

Alibaba Cloud CLI returns JSON format responses (streaming multiple JSON events):

Key fields:

• ConversationId: Conversation ID (for multi-turn dialogue)
• Answer: AI assistant's response content
• Event: Event type (workflow_started, message, workflow_finished)

Success Verification

1. CLI installation successful: aliyun version shows version number
2. Credential configured correctly: aliyun configure list shows configured credentials
3. Observability configured correctly: Business API calls include --user-agent 'AlibabaCloud-Agent-Skills/alibabacloud-rds-copilot/{session-id}' with a 32-character hex session-id
4. API call successful: Response contains ConversationId and Answer in JSON format
5. Response content valid: Answer is relevant to the query content

See references/verification-method.md for detailed verification steps.

Cleanup

This skill only performs read-only query operations, does not create any cloud resources, no cleanup required.

API and Command List

See references/related-apis.md for details.

Best Practices

1. Use multi-turn dialogue: For complex issues, use --conversation-id for context-aware multi-turn conversations
2. Specify correct region: Set RegionId parameter based on the RDS instance's region
3. Be cautious in production: SQL recommendations from RDS Copilot should be validated in test environment first
4. Save conversation ID: Save the returned ConversationId if you need to follow up or continue analysis
5. Use configuration file: Recommend using aliyun configure to configure credentials, avoid exposing sensitive information in command line
6. Use --profile: You can configure multiple credential profiles and switch between accounts using --profile

Reference Links