Unified IMA API Assistant

Problem

The IMA open API separates notes and knowledge bases into different modules, but real user requests are often ambiguous. For example, “save the knowledge-base content into a note” requires cross-module execution, while “add this to note YY inside knowledge base XX” may mean appending to an existing note rather than creating a knowledge-base entry. Misrouting the API call can write content to the wrong place. In Windows PowerShell 5.1, HTTP request bodies may also be silently converted to GBK, producing unrecoverable Chinese text corruption.

How It Works

The skill targets the official ima.qq.com OpenAPI and uses HTTP POST + JSON Body consistently. Core capabilities include:

• Notes module: search notes, browse notebooks, read content, create new notes, and append content.
• Knowledge-base module: upload files, add web links, search knowledge bases, browse knowledge-base information, and list addable targets.
• Routing rules: use notes when the target is note content; use knowledge-base when the target is a knowledge-base entry; execute multi-module tasks in the order implied by the user intent.

Key steps include confirming Client ID and API Key, then selecting the relevant module documentation. For notes write operations, it enforces UTF-8 validation for fields such as content and title; for knowledge-base file uploads, it preserves the original binary content and avoids unauthorized transcoding.

Boundaries

This skill depends on official IMA credentials and is intended for users who already have an IMA account and are comfortable configuring API credentials. It does not invent new business logic; instead, it constrains call paths, encoding checks, and module selection. If the intent is unclear, such as “remember this” without specifying whether to create a new note or append to an existing one, the skill should confirm first rather than guessing the endpoint.