What to note when using OpenAPI Explorer
Calling an API in OpenAPI Explorer — whether from a web page or Cloud Shell — performs real operations on your resources. Review your configuration carefully before clicking Initiate Call to avoid unintended changes to live resources.
Is OpenAPI Explorer free of charge?
OpenAPI Explorer and all cloud product APIs are free to use. Charges apply in two cases:
Resources or services created through API calls are billed at standard rates. For example, calling the
CreateInstanceAPI of ECS provisions an instance and incurs ECS service fees. Check the billing documentation of the relevant product for pricing details.APIs from the API marketplace must be purchased before use, and each call is charged separately.
What to prepare before using OpenAPI Explorer
An Alibaba Cloud account with the required permissions for the cloud products whose APIs you want to call.
What is the difference between Cloud Shell and CLI?
Cloud Shell is built on Cloud Assistant CLI. The command format and syntax are the same for both. Cloud Shell runs directly in your browser with no local installation required, while the CLI requires a local environment setup.
FAQ for AI Assistant
Cannot find the corresponding API
The current beta version covers APIs that account for 95% of typical usage. If an API is missing, it may fall outside that range — or the functionality may not have a corresponding API in Alibaba Cloud. Full OpenAPI portal catalog coverage is planned for a future release.
Why does the AI Assistant keep refusing to answer my questions?
The assistant is scoped to API-related questions. Questions outside that topic are declined by design.
Does the model support out-of-domain questions?
The current beta is optimized for two tasks: finding APIs and generating SDK code samples. Its intent recognition is tuned for these use cases and cannot accurately answer questions on other topics.
Why does the AI Assistant suddenly say "I cannot answer" during streaming responses?
Large language models (LLMs) can generate plausible but factually incorrect information — a phenomenon called hallucination. To guard against this, each response goes through a fact-check before it is fully displayed. If fabricated content is detected, the response is blocked and replaced with "I cannot answer."
Why do I encounter "Please try again later"?
The LLM backend supports up to 15 concurrent requests. During peak usage, additional requests are throttled. Wait a moment and retry.
Q&A contextual relevance is poor
Start a new session. Context from earlier turns can interfere with later questions, and a fresh session clears that state. Multi-turn conversation quality is an active area of improvement.
What large language model does OpenAPI AI Assistant use?
It uses a model from the Qwen family, specifically trained for API discovery and SDK code generation tasks.