Summary

Thoth is a local-first AI assistant prioritizing personal sovereignty, integrating tools, a knowledge graph, and voice/vision. It targets users demanding data privacy and control, enabling local AI execution with optional cloud model integration, offering a powerful platform for personal productivity and knowledge management.

How It Works

Thoth uses a LangGraph-based ReAct agent orchestrating 27 core tools and auto-generated channel integrations. Its core is a persistent personal knowledge graph, enhanced by FAISS semantic search and graph traversal for context recall. Designed for local-first operation via Ollama, user data remains on-device, with flexible opt-in support for cloud LLMs.

Quick Start & Requirements

One-click installers for Windows/macOS automate Python, Ollama, and dependency setup. Local models require Python 3.11+, 8GB RAM (16-32GB recommended), optional GPU; disk space varies (5GB+). Cloud-only needs Python 3.11+, 4GB RAM, ~1GB disk, internet, and provider API keys. Official demos and architecture docs are linked.

Highlighted Details

• Personal Knowledge Graph: Stores entities/relations, with graph-enhanced recall (semantic + 1-hop expansion) and interactive visualization.
• Dream Cycle: Nightly background process refines the knowledge graph via deduplication, enrichment, and inference.
• Designer Studio: Creates/edits multi-page presentations, one-pagers, reports with export options.
• Toolset: Includes web search, shell, browser automation, vision, image generation, document extraction, and 5 messaging channels (Telegram, WhatsApp, Discord, Slack, SMS).
• Local Voice: Faster-whisper STT and Kokoro TTS run entirely on-device.
• Workflow Engine: Supports step-based pipelines with conditional branching and approval gates.

Maintenance & Community

An open-source project on GitHub, Thoth benefits from community contributions. Specific community channels or active maintainer details are not explicitly listed.

Licensing & Compatibility

Released under the permissive Apache 2.0 license, allowing commercial use and integration into closed-source applications.

Limitations & Caveats

Cloud models require provider API keys and incur per-token costs. Local model setup involves large downloads, demanding disk space and RAM. Many advanced tools require users to obtain and configure their own external API keys.