THOTH → dwtj → Djehuty ( pronounced Jeh - hoo - tee )
Thoth learns your voice, your knowledge, your publishing patterns; then helps you write original technical content that sounds like you, not an LLM.
Free forever with local Ollama; no API key required. Your data never leaves your machine.
Thoth builds a model of you in three dimensions.
- What you know
- How you communicate
- What you've published.
Then uses
that model to generate content that is unmistakably you.
What you know.
- Extracted from your repos, architecture docs, and technical notes.
- Powers factual accuracy and real examples.
How you communicate.
- Tone, pacing, vocabulary, storytelling rhythm; all extracted from
essays, journals, and reflections.
What you've built publicly.
- Tracks article relationships, series continuity, and your
publishing evolution over time.
Creates a local SQLite database at
~/.thoth/. No cloud. No account.
Point Thoth at your notes, repos, or articles. It chunks, embeds, and stores locally.
$ thoth import voice ./journalsThoth analyses your imports and builds the three identity profiles. Runs once, updates on demand.
$ thoth profile generateGive Thoth a topic. It researches, outlines, and generates an article in your voice — 8 to 10 minute reads, split automatically into series if needed.
$ thoth draft --topic "Offline First"
Thoth is a CLI tool.
Everything — your profiles, embeddings, drafts; lives in a SQLite
database on your own filesystem.
All data is stored at
~/.thoth/thoth.db. Nothing is sent to Thoth servers; there are none.
No API key, or cost, or data leaving your network. Runs entirely on your hardware. OpenAI and Gemini are opt in for higher quality.
If you opt into OpenAI or Gemini, only your embeddings and prompts are sent; never your raw source files.
The full source is on GitHub. Audit what runs. Fork it. Own it entirely.