llm_wiki

LLM Wiki builds a persistent, interlinked wiki out of your documents instead of re-answering the same questions from scratch. Most retrieval tools reason again on every query; this one reasons once, writes wiki pages with source traceability and cross-references, then maintains them as you add material. Feed it PDFs, Word files, or web clips. GPL-3.0, desktop app, macOS/Windows/Linux.

It needs an LLM provider configured before it does anything: OpenAI, Anthropic, Google, Ollama, or any OpenAI-compatible endpoint. Point it at Ollama and the whole thing runs local and costs nothing. Point it at a hosted API and ingest is where the money goes, because the two-step pass reasons over content before generating pages.

The build has grown well past a document indexer: a Rust backend agent with tool use, multimodal ingest that pulls images out of PDFs, a Deep Research mode running multi-query web searches, a Chrome clipper, and a knowledge graph view with community detection. Solo researchers and anyone sitting on a document pile get real value. Teams get less, since there is no shared instance.

The catch is that GPL-3.0 is a deliberate choice with consequences. Build a product on this and your product inherits the license. For personal use, irrelevant; for anything commercial, it is the first thing to check. The quieter catch: a wiki generated by an LLM is a wiki full of LLM claims, and traceability tells you where a sentence came from, not whether it is correct.