Unverified50% confidenceDecision RuleExact time
Selection advice: use RAG or ordinary agentic retrieval when the answers are in the source files and the source files are stable (API documentation, the library's own code, specifications) and the corpus is large and read-oriented; use an LLM wiki when part of the knowledge is in the source files and part is in people's heads/reviews/incidents/chat logs and cross-session reuse and an audit trail are needed; when the corpus is large, the two can be used together (the wiki manages decisions and synthesis, retrieval manages raw material).
1
Sources
50%
Confidence
Long-term
Relevance
10/14/2026
First Seen
Sources
Related Entities
Related Claims
Unverifiedllm-wiki模式在实践中需要解决知识冲突和知识腐化两大工程挑战,通常需引入版本管理和定期知识审计机制69% similarUnverified对于'本周最该关注的三件事'这类需要综合判断的问题,语义搜索或关键词匹配无法给出答案,而结构良好的LLM Wiki能让Agent跨文件收集信息整合67% similarUnverifiedThe knowledge boundary of an Agent in RAG-based products depends on the quality of the content provided, and outdated or incorrect information in the knowledge base will cause the model to cite errors.66% similarUnverified工具核查质量高度依赖所接入数据库的覆盖范围,冷门期刊、非英语文献或最新预印本可能存在覆盖盲区66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1032441API
curl https://kongchang.com/api/v1/knowledge/claims/1032441MCP
get_claim(id=1032441)