Создание базы знаний с источниками для AI-агентов

★ 8.0 · knowledge-management

nvk/llm-wiki is an open-source tool that builds LLM-compiled knowledge bases and plugs into any AI agent as a first-class plugin. Reach for it when you need a traceable, source-backed knowledge base rather than a one-off model answer; when you're taking a rough idea through research and want to promote it into a structured delivery project; when the output must survive as a reusable artifact and not disappear into a chat thread; or when you're running thesis-driven research and need the wiki to confirm or challenge your starting claim. Under the hood it runs parallel multi-agent research, thesis and collector modes, URL ingestion, session memory, wiki compilation, audits, variable-depth querying, and knowledge-checkpoint exports. Built for Claude Code (native plugin via `claude plugin install wiki@llm-wiki`), OpenAI Codex, OpenCode, and portable agents; stores data in an Obsidian-compatible format. The direction is assembling a knowledge base from scratch around a specific question — not indexing an existing document corpus (use a dedicated RAG pipeline for that). Wiki quality depends on the quality of sources the agents find, so fact-checking remains the user's responsibility. Works well for research and product teams building with AI agents; not a drop-in replacement for a governed enterprise knowledge-management system without additional editorial oversight.