Synthadoc
An LLM knowledge-compilation engine that reads raw documents, such as PDFs, spreadsheets and transcripts, and synthesizes them into a persistent, cross-referenced wiki stored as local Markdown files, as an alternative to RAG.
Synthadoc reads raw source documents, including PDFs, spreadsheets, presentations, web pages, images, videos, Word files, text files and AI session transcripts, and uses an LLM to synthesize them into a structured wiki instead of just chunking them for retrieval. Cross-references between pages are built automatically, contradictions between sources are detected and surfaced, orphan pages are flagged, and every answer cites the documents it came from.
Outputs are stored as local Markdown files, so the wiki integrates directly into Obsidian or any other Markdown-based wiki, and can be worked with through a CLI, an Obsidian plugin, a web UI or an MCP server. It is aimed at solo researchers as well as larger organizations, can run for free on Gemini Flash or a local Ollama model, and is written in Python under the AGPL-3.0 license.