IMA Notes and Knowledge Base API

Problem

IMA exposes notes and knowledge-base features through separate OpenAPI operations: creating notes, appending content, searching knowledge bases, uploading files, adding web links, and retrieving original content all require different parameters and flows. When calling these APIs directly, teams often hit routing mistakes, encoding corruption, accidental filename changes, missing credentials, and inconsistent error layers. In particular, notes writes can become unrecoverable garbled text if content or title is not valid UTF-8, and PowerShell 5.1 may silently convert the request body to GBK.

How It Works

The skill consolidates the IMA OpenAPI surface into two modules: notes and knowledge-base. The notes module covers note creation, appending, search, and content retrieval. The knowledge-base module covers file uploads, web-link additions, knowledge-base search, and original-content management. It routes by intent: use import_doc or append_doc when the target is note content; use create_media, add_knowledge, or related lookup APIs when the target is a knowledge-base entry; and read both submodules before chaining cross-module operations.
Before calling the API, the skill expects IMA Client ID and API Key credentials. Requests are sent as HTTP POST + JSON through ima_api.cjs. Errors are split into two layers: script execution errors appear in stderr with -100 or -200; backend business errors appear in stdout with code and msg. Before notes writes, string fields must be validated as UTF-8. File uploads must preserve binary content, and title must match the original filename.

Boundaries

This is useful for automating IMA note and knowledge-base workflows, but it does not cover every IMA capability and does not support video files, Bilibili/YouTube URLs, or file:// addresses. It should not be used to rename original files, skip encoding checks, or repeatedly retry API calls when credentials are missing.