Tencent IMA Knowledge Base

Problem Solved

Documents, web pages, and WeChat articles are often scattered across local folders, browser bookmarks, and shared links, making them hard for models to ground answers in. This skill exposes the Tencent IMA Knowledge Base as a readable and writable interface so callers can search knowledge bases, browse folders and files, and add new files or web article URLs.

How It Works

It uses IMA OpenAPI, where every request is HTTP POST + JSON under /openapi/wiki/v1/. Credentials are handled by get-token.ps1 or get-token.sh, preferring a platform proxy, then environment variables, then a config file.

Key capabilities:

• Read: search content with search_knowledge, browse knowledge bases/folders with get_knowledge_list, and look up bases by name with search_knowledge_base.
• Write: upload files through create_media → COS upload → add_knowledge; add web or WeChat article URLs with import_urls.
• Batch upload: scripts can scan a directory or process a file list, with recursive scanning, extension filtering, and result summaries.

Boundaries

Supported uploads include .pdf, Office documents, .md, images, .txt, and audio files, with different size limits by type. Video files, some online video pages, and file:// local HTML are not supported. Output should show knowledge base names and file titles rather than internal IDs such as kb_id, media_id, or folder_id.