𓁟

Meet Thoth

Your private knowledge agent. Upload documents, ask questions, get cited answers — all running locally on your machine.

⬇ Download for Windows ☆ View on GitHub

Windows 10/11 · 64-bit · Free & open source (MIT)

Everything you need, nothing you don't

💬

Multi-Turn Conversations

Persistent chat threads with full message history. Auto-naming, thread switching, and seamless resume across sessions.

📄

Document Intelligence

Upload PDF, DOCX, and TXT files. Documents are chunked, embedded, and indexed into a local FAISS vector store for instant retrieval.

🔍

Multi-Source Retrieval

Pulls context from your documents, Wikipedia, Arxiv, and the web. Toggle each source on or off from the settings panel.

📝

Source Citations

Every piece of information is cited — documents, Wikipedia articles, Arxiv papers, web links, or the model's own knowledge.

🤖

30+ LLM Models

Switch between Llama, Qwen, Gemma, Mistral, DeepSeek, Phi, and more. Models download automatically on demand via Ollama.

🔒

Fully Local & Private

All LLM inference runs on your machine. No cloud. No telemetry. Your data never leaves your computer.

🧠

Contextual Compression

Retrieved documents are compressed and filtered to extract only the content relevant to your query, improving answer quality and reducing noise.

🔄

Smart Query Rewriting

Follow-up questions with pronouns or references are automatically rewritten into standalone queries using conversation history for more accurate retrieval.

Parallel Retrieval

All enabled retrieval sources are queried simultaneously, reducing total retrieval time to the speed of the slowest source instead of the sum of all sources.

🧹

Context Deduplication

Embedding-based similarity deduplication removes near-duplicate results both within a single query and across conversation turns.

Up and running in 3 minutes

1

Download & install

Run the installer — it sets up Ollama, Python, and all dependencies automatically. No command line needed.

2

Launch Thoth

Open from the Start Menu or Desktop shortcut. The app opens in your browser. On first launch, the default LLM model downloads automatically.

3

Ask anything

Start a conversation, upload your documents, and get answers with full source citations. Configure retrieval sources and API keys from the Settings panel.

🔒 Privacy by design

Thoth runs entirely on your machine. Your documents, conversations, and API keys are stored locally and never sent to any cloud service. External calls are only made to the retrieval sources you explicitly enable.

🖥️ Local LLM inference
💾 Local data storage
🚫 No telemetry
📜 MIT License

Ready to try Thoth?

Free, open source, and yours to keep.

⬇ Download for Windows ☆ Star on GitHub