Personas — clone a personality, grounded in real knowledge
A Persona is a character agent that talks in the voice of a real person, a historical figure, a brand, or a fictional character — and stays grounded in a body of source material you give it. Feed it documents, books, transcripts, or URLs; NodeGraph chunks and embeds that material into a vector database and the persona answers using retrieval-augmented generation (RAG) — every reply is backed by the actual sources, with verifiable citations rather than free-floating hallucination.
It works over text and voice, and a persona can be made public so anyone can talk to it directly — on the web or as a Telegram Mini App.
Talk to Aristotle
Aristotle is our flagship persona — a digital twin grounded in his actual works. Ask him a question and he answers in his own voice, citing the passage he's drawing from. It's the canonical example of what a persona clone feels like: an immersive conversation with a historical mind, not a generic chatbot. Open the app and start the Aristotle scenario to try it.
How a persona clone works
- Create a persona (name, voice, system character).
- Feed it sources — documents, URLs, or pasted text — via the Persona Feed block. NodeGraph runs the ingest pipeline: download → chunk → embed → index into a vector store.
- Talk to it. Each question runs a semantic search over the corpus; the most relevant passages are retrieved and the model answers in-character, grounded in (and citing) those passages.
- Publish it — make it public on the web, or wire it into a Telegram AI Bot as a Mini App.
The persona block family
| Block | Type ID | Description |
|---|---|---|
| Persona Twin | persona.twin | The character agent itself — bound to its corpus, answers in-voice with RAG grounding and citations. Text + voice. |
| Persona Feed | persona.feed | Add a new source to the corpus by document, URL, or text — automatic license check, then chunk + embed + index. |
| Persona Library | persona.library | Live inventory of the corpus — sources, licenses, indexed chunk counts. Drop a source to remove its vectors. |
| Persona Activity | persona.activity | Live event tail — ingestion jobs, queries, updates. |
| Persona Monitor | persona.monitor | Visual presence dashboard — a particle pulse or a 3D avatar with audio-driven lip-sync while the persona speaks. |
Bring your own personality
Aristotle is just the showcase. Point the Feed block at any corpus — a founder's talks and writing, a brand's style guide, a fictional universe's lore — and you've cloned that personality into a talkable, cite-everything agent.
How they connect
| From | To | What happens |
|---|---|---|
| Persona Feed | Persona Twin | New sources are ingested into the persona's corpus |
| Persona Twin | Persona Monitor | The monitor shows the persona's avatar + live activity |
| Persona Twin | Telegram AI Bot | The persona becomes a Telegram Mini App ("Make Mini App") |