Digital Replicant is an open-source platform for creating a highly accurate cognitive replication of a person based on interviews, narratives, memories, and personalized moral models. The goal of the project is to create a digital twin as close to the original person as possible, enabling third parties to interact as if they were speaking to the actual person, not an algorithm.
Digital Replicant — это open-source проект, задача которого создать максимально точный, глубокий и когнитивно достоверный цифровой портрет личности человека (Оригинала). Полученный портрет используется в качестве системного промпта для GPT, что позволяет другим людям (Третьим лицам) общаться с моделью так, как если бы они общались с этим человеком напрямую.
For local use:
# install dependencies
pnpm install
# initialize DB
pnpm init-db
# run server in dev mode
pnpm back
# run front in develop mode
pnpm front
# open web
http://localhost:9000/#/Before running the project, create a .env file in the root directory and add the following environment variables:
OPENAI_API_KEY=your_api_key_here
OPENAI_BASE_URL=https://api.openai.io
OPENAI_API_MODEL=gpt-4o-mini- `OPENAI_API_KEY` — your API key for accessing the language model.
- `OPENAI_BASE_URL` — the base URL of your API endpoint.
- `OPENAI_API_MODEL` — the model identifier you want to use for generating responses.
Note: Replace your_api_key_here with your actual API key.
Digital Replicant aims to go beyond simple "chatbots" that mimic conversational style and move towards building a deep holistic cognitive portrait of a personality, taking into account:
- beliefs and core values,
- emotions and reaction patterns,
- past experiences and memories,
- evolution of views over time,
- moral structures and inner contradictions.
The project is based on the idea that a person is not a sum of answers, but a dynamic system of memory, emotions, experience, speech patterns, and the connections between them. In this sense, Digital Replicant is not just a chatbot but a reproducible model of a personality, which can be applied in:
- digital legacy projects,
- cognitive research,
- humanistic AI architecture,
- healthcare (e.g. cognitive therapy, dementia support),
- dialogue agents shaped by the cognitive and emotional profile of a specific person.
- Original — the person whose cognitive portrait is being recorded.
- Replicant — a GPT model + system prompt + assembled cognitive portrait of the Original.
- Third Party — a user interacting with the Replicant, expecting realistic, human-like behavior.
- A Replicant is created via a "Create Replicant" interface.
- Interviews are the primary data collection format.
- Each interview consists of topics.
- Each topic contains questions, answers, and emotions.
- There are required topics: Introduction, Close Ones, Authorities, Politics, Religion, and others.
- The Original can add custom topics that are personally important (e.g. "Marxism," "Military Service," "Crisis 2014," etc.).
- Each question is accompanied by: question text, answer, and selected emotion.
- GPT generates a question based on previous answers and context.
- The Original answers and selects an emotion.
- The question, answer, and emotion are recorded.
- Option to skip a question with a reason.
- Ability to switch between topics.
- Ability to add new topics.
After completing the interview process, a step-by-step holistic personality portrait is generated.
For each topic, a report is generated containing:
- Core beliefs and views
- Evolution of views
- Emotional background
- Key memories
- Causal links and motivations
- Speech patterns
- Internal contradictions
- Personality summary for the topic
Each topic summary is used as context for the next, ensuring cumulative cohesion in the final model.
A unified cognitive portrait is assembled from all topics, providing a more compact but deeply structured overview:
- Core beliefs
- Evolutions of views
- Emotional background
- Key events and memories
- Causal connections
- Speech patterns
- Inner contradictions
- Biographical summary
- Final personality summary
This result becomes the Replicant’s main system prompt and is used in conversations with third parties.
- The holistic portrait is used as a system prompt.
- A third party sends a message.
- GPT generates a response as close as possible to the Original’s personality.
- The response includes an emotion, which can be visualized.
- The interface displays text + emotion.
- Optionally: voice-over, video avatar, synchronous articulation (external modules).
- Ability to specify who is interacting with the Replicant (e.g. Wife Anna, Brother Peter).
- These details are injected into the prompt and can adjust the tone or content of the conversation.
- Topics related to close people are linked to individual third-party profiles.
- Daily monitoring of news relevant to the Original’s core beliefs.
- When potentially significant events are detected — an option to update or adjust opinions and values.
- All mentioned stories and events are collected in a dedicated database.
- RAG-based access to memories is available during response generation.
- Before each response, an "evaluator" layer runs:
- Analyzes the query for the need to fetch current news.
- Checks for relevant memories.
- Suggests contextual elements that can be added to the response (weather, mood, references to past events, etc.).
- Backend: Node.js, Express, TypeScript, Zod
- Frontend: Vue 3, Quasar
- Database: Plasma DB + SQLite
Digital Replicant strictly follows ethical and privacy principles:
- All interviews, memories, and replicant data are stored exclusively on the user’s side.
- Data is not transmitted or stored on third-party servers or cloud services without explicit consent.
- Users have full control over their data, with options to delete or export at any time.
- The project is committed to respecting personality, protecting privacy, and transparency in data handling.
- Any use of the model for third-party interactions must comply with ethical standards and data protection laws.
- Replicating a personality raises important issues of privacy, consent, and autonomy.
- A digital twin is not just data, but a reflection of someone’s unique identity.
- It is critical to strictly respect the rights and freedoms of the original person, and ensure user control over their data.
- The technology must not be used for manipulation, deception, or breaching trust.
- All information is stored solely on the user’s side and is not shared with third parties without explicit consent.
Any developer, researcher, philosopher, or psychologist interested in digital twins, cognitive modeling, and humanistic AI is welcome.
We would appreciate:
- new feature proposals,
- UI/UX improvements,
- hypothesis validation based on collected data,
- integration development (voice-over, video, voice avatars, etc.).
Create issue