WikiMind¶

You never write the wiki. You feed it. Every question makes it smarter.

WikiMind is a personal LLM-powered knowledge OS. Feed it articles, PDFs, YouTube videos, podcasts, or papers -- it compiles them into a structured wiki and answers questions with full source attribution.

What WikiMind is¶

Not a note-taking app -- you never write
Not a chatbot -- it builds something persistent
Not a RAG tool -- the wiki is the product, not a retrieval layer

It is the synthesis layer that sits above everything you consume.

How it works¶

Feed --> Compile --> Query --> Answer files back --> Wiki gets smarter --> Repeat

Feed -- Ingest URLs, PDFs, raw text, or YouTube videos
Compile -- An LLM transforms each source into a structured wiki article with key claims, concepts, and backlinks
Query -- Ask questions against your wiki; the Q&A agent cites specific articles
File back -- High-confidence answers are filed back into the wiki, making it smarter over time

Quick links¶

:material-rocket-launch:{ .lg .middle } Quick Start

Get WikiMind running locally in 5 minutes.

:octicons-arrow-right-24: Quick Start
:material-file-document-multiple:{ .lg .middle } Ingesting Sources

Learn how to feed URLs, PDFs, text, and YouTube videos.

:octicons-arrow-right-24: Ingestion Guide
:material-chat-question:{ .lg .middle } Q&A Agent

Ask questions and get cited answers from your wiki.

:octicons-arrow-right-24: Q&A Guide
:material-cog:{ .lg .middle } Configuration

All environment variables and settings documented.

:octicons-arrow-right-24: Settings Reference

Tech stack¶

Layer	Technology
Backend gateway	Python 3.11+ / FastAPI
Job queue	ARQ + asyncio (in-process for dev, ARQ + Redis for prod)
Database	SQLite (dev) / PostgreSQL (prod) via SQLModel
LLM providers	Anthropic Claude, OpenAI GPT, Google Gemini, Ollama
PDF extraction	docling-serve sidecar; pymupdf fallback
Document ingest	trafilatura (URLs), youtube-transcript-api (YouTube)
Frontend	React 18 + TypeScript + Vite + TanStack Query + Zustand + Tailwind CSS
Testing	pytest + pytest-asyncio + httpx

Status¶

Phase 1 (Working Core) -- Done

Backend pipeline: ingest -> compile -> query -> file-back
React UI: Inbox + Wiki Explorer
LLM provider abstraction with auto-enable
Multi-format ingest (URL, PDF, text, YouTube)
Source provenance and citation chains

Phase 2 (Query Loop) -- In progress

Conversational Q&A agent with thread file-back
React UI: Ask view with conversation threads
Semantic search (ChromaDB + embeddings)
Knowledge graph view
Wiki linter and health dashboard