convez.ai
AgenteProdutividadeAvançado
mrdaemoni

myalicia

mrdaemoni/myalicia

Companheiro IA personalizado que cresce com você em múltiplos ritmos

Ver no GitHub
BomPluginScore · de 0 a 100

Um agente IA que se adapta à forma de quem o utiliza, implementando humorfismo através de três loops de reflexão (Listen, Notice, Know) em cadências diferentes. Cria um companheiro inteligente personalizado que aprende padrões sobre você ao longo do tempo, melhora autonomamente e se estende com novas habilidades conforme necessário.

Instalação

$npx convez-ai mrdaemoni/myalicia

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
7
Forks
1
Acompanhando
0
Saúde
85%
Último commit
há 6 meses
Versão
v0.1.5
Issues abertas
1
Contribuidores
1

Casos de uso

  • Criar um assistente IA customizado que aprende seus padrões e preferências
  • Automatizar reflexão pessoal e síntese de conhecimento em diferentes escalas de tempo
  • Integrar um agente IA nos seus fluxos existentes (Telegram, CLI, Claude Code)

Veredito

Projeto ambicioso que rejeita design genérico em favor de relacionamento profundo com um usuário. A arquitetura de três loops é sólida, mas exige disposição para experimentar e curar a própria IA — não é um plug-and-play. Vale para quem quer entender como IA personalizável funciona por dentro.

README

A tradução para o português deste README ainda está na fila.

myalicia

An AI teammate that grows into the shape of the person it serves.

myalicia is the tangible example of humorphism — the design philosophy that technology should take the shape of the human, not the other way around. Where humorphism gives the philosophy, myalicia gives the architecture you can run on your own machine.

The project is called myalicia because every running instance is someone's own. There is no shared "Alicia" served to everyone. The framework teaches you to build your myalicia — one relationship, deepened over time, shaped by you, by design.


Three loops, three depths of attention

myalicia is built around three relationship loops, each running at a different cadence, all reading from and writing to a shared memory substrate (your vault, your notes, your archetype):

Listen — seconds. Be present in the conversation in front of you. Cheap, fast model. The only loop most AI products have.

Notice — minutes to hours. Catch patterns across moments. Triggered by events: you finished a book, you closed a project, you came back from a trip. Mid-tier model, wider context, asynchronous.

Know — days to weeks. Come to know the person, over time. Scheduled autonomous reflection. Most expensive, runs on its own, produces synthesis the user didn't ask for but is glad to have.

Awareness compounds because three different cadences of reflection are constantly metabolizing input into structure. That compounding is what makes a teammate, not a tool.


What makes it different

It is shaped to one person, on purpose. Most AI products try to be useful to everyone, and end up useful to no one in particular. myalicia gets sharper at one relationship, week over week.

It is self-healing. When a skill errors, myalicia notices, logs the trajectory, and proposes a fix in the next reflection loop. You don't have to babysit it.

It is self-extending. myalicia can author new skills based on observed gaps in her own behavior. Every instance grows.

It is self-aware. Every action runs through metacognitive confidence assessment; uncertain actions escalate to deeper models automatically.

It is plural by design. Reach myalicia through Telegram, Claude Code, or the CLI. The same teammate, different surface for different parts of your life.


Quickstart

With pipx (recommended — handles the venv for you):

brew install pipx       # macOS; or apt/dnf install pipx on Linux
pipx install myalicia
myalicia init

With pip + venv (if you want to hack on the code):

git clone https://github.com/mrdaemoni/myalicia.git
cd myalicia
python3 -m venv .venv
source .venv/bin/activate
pip install -e .
myalicia init

The venv step is required on macOS Homebrew Python and most modern Linux distros (PEP 668 blocks system-wide pip installs). pipx avoids this by creating a dedicated venv per package.

See docs/QUICKSTART.md for the full walkthrough and docs/FAQ.md for setup questions.

The init flow is more than configuration — it's a guided introduction to relationship-shaped AI. Each setup step is paired with a one-screen explanation of which loop it activates and why. By the time the install completes, you understand the architecture, not just have it running.

See docs/QUICKSTART.md for the full walkthrough, docs/FAQ.md for setup questions, and docs/PAPERS.md for the intellectual lineage.


Architecture

myalicia/
├── core/                    # the thin harness (loops, scheduler, routing, memory)
├── skills/                  # the fat skills (~80 modules, each composable)
├── surfaces/                # telegram, claude_code, cli adapters
├── config.py                # typed config — the single source of truth
└── defaults.yaml            # shipped defaults
examples/
├── archetypes/              # curious-builder, patient-mentor, ariadne, …
├── memory_seeds/            # starter memory files
├── vault_layouts/           # example Obsidian vault layouts
└── deployments/             # launchd, systemd, docker

See ARCHITECTURE.md for the detailed view. See PHILOSOPHY.md for the humorphism framing in depth.


Contributing

The contribution surface we are most excited about is awareness primitives — small, opinionated modules that turn some slice of life into structured memory the loops can metabolize. A reading-pattern primitive. A calendar-rhythm primitive. A relationship-graph primitive. A mood-arc primitive.

Don't think of contributions as "writing code." Think of them as "designing perception." That framing is what humorphism asks of designers in the agentic era.

See CONTRIBUTING.md for how to write one.


Acknowledgments

myalicia stands on the shoulders of humorphism, Garry Tan's thin-harness/fat-skills pattern, and Andrej Karpathy's autoresearch pattern. See ACKNOWLEDGMENTS.md.

License

MIT — see LICENSE.

Ver o original no GitHub

Bom para

Relacionados