Deling AI Knowledge Base

Problem

When drafting research answers, models may rely on general knowledge or stop after one search, producing claims without enough facts, cases, data, or traceable sources. Deling AI Knowledge Base targets this deep-research scenario: it confirms whether the user means the Deling platform, then runs a structured evidence-collection workflow before content generation.

How It Works

The skill starts with intent clarification. For ambiguous requests such as searching documents or finding materials, it avoids guessing and asks whether the user wants Deling knowledge-base research. Once confirmed, it proceeds in stages:
- Knowledge-base first: use sessions_search for broad exploration and identify dimensions such as products, features, architecture, and cases.
- Vertical mining: run targeted searches for each important dimension and follow linked resources mentioned in documents.
- Web supplement: use web search only when knowledge-base coverage is insufficient, for trends, external authorities, or statistics.
- Diversity validation: cover facts, cases, expert views, trends, comparisons, and risks.
- Source tracing: record source names, file IDs or links, quoted snippets, and distinguish knowledge-base sources from web sources.

Note: it is not a generic search API; it is a pre-generation research and quality check. If the user has not specified the platform, clarify first; if the knowledge base has relevant resources, avoid over-relying on web search.