Anima Protocol

Your AI shouldn't forget who it is.

Anima Protocol is an open-source platform for creating persistent, evolving AI companions with memory, identity, personality, voice, multimodal generation, and multi-character crossover conversations.

Live app: https://anima-protocol.com

Traditional AI chats preserve context.

Anima Protocol is designed to preserve identity.

Why Anima?

Anima Protocol explores what happens when an AI companion is treated as a persistent character rather than a disposable chat session. The platform is being built around durable memory, distinct personalities, user-controlled resonance settings, and multi-character experiences where each companion can retain an individual voice and history.

Core capabilities

• 🧠 Persistent companion memory architecture
• 🎭 User-created AI identities and personalities
• 🌌 Multi-character crossover sessions
• 💬 Real-time conversational experiences
• 🎨 AI image generation and editing
• 🔊 Optional text-to-speech voices
• 🧬 Resonance controls for tone, intensity, memory depth, and boundaries
• 🏠 Self-hosted Anima LLM support through Ollama or vLLM
• 🔐 User-scoped persistence and authentication
• 📊 Consent-gated product analytics

Project status: active alpha development. The application foundation is in place, while persistent memory, companion creation, crossover orchestration, and resonance systems remain active product work.

Architecture

flowchart LR
  user[User] --> web[React + Vite app<br/>artifacts/anima-protocol]
  web --> api[Express API<br/>artifacts/api-server]
  web --> mixpanel[Mixpanel<br/>consent-gated analytics]
  web --> clerk[Clerk frontend auth]
  api --> clerk_api[Clerk session verification]
  api --> animaLlm[Anima LLM<br/>Ollama/vLLM open weights]
  api --> geminiImages[Gemini Flash Image<br/>preferred image generation/editing]
  api --> openaiImages[OpenAI gpt-image-1<br/>fallback image path]
  api --> eleven[ElevenLabs optional TTS]
  api --> db[(PostgreSQL<br/>Drizzle schema in lib/db)]
  mockup[Mockup sandbox<br/>artifacts/mockup-sandbox] --> web

Tech Stack

Layer	Technology
Frontend	React 19, Vite, Tailwind CSS, Radix UI, framer-motion, React Query, zod
Backend	Node 24, Express 5, Clerk middleware
Database	PostgreSQL, Drizzle ORM, drizzle-kit
Local AI	Ollama or vLLM with OpenAI-compatible endpoints
Image generation	Gemini Flash Image with OpenAI image fallback
Voice	ElevenLabs optional TTS
Analytics	Mixpanel through a consent-gated wrapper
Package manager	pnpm workspaces

Quick Start

Prerequisites

• Node 24
• pnpm
• Docker, if using the bundled local Postgres + Anima LLM development stack

Clone and install:

git clone https://github.com/davins56/anima-protocol.git
cd anima-protocol
pnpm install --frozen-lockfile

Start the bundled development infrastructure:

pnpm dev:infra:up

Initialize the local database:

export DATABASE_URL=postgresql://anima:anima_dev@localhost:5432/anima_dev
pnpm --filter @workspace/db run push

Start the API:

export DATABASE_URL=postgresql://anima:anima_dev@localhost:5432/anima_dev
export PORT=8080
export NODE_ENV=development
pnpm --filter @workspace/api-server run dev

Start the frontend in a second terminal:

export PORT=23660
export BASE_PATH=/
pnpm --filter @workspace/anima-protocol run dev

For environment-specific authentication, analytics, AI-provider, and deployment configuration, read AGENTS.md and the documentation under docs/ before changing runtime behavior.

Project Structure

Anima-Protocol/
|-- artifacts/
|   |-- anima-protocol/      # Main React + Vite app
|   |-- api-server/          # Express API for /api/*
|   `-- mockup-sandbox/      # Isolated UI previews at /__mockup
|-- lib/
|   |-- db/                  # Drizzle schema and database helpers
|   |-- api-client-react/    # Shared API client package
|   `-- api-spec/            # Shared API specification package
|-- docs/                    # Architecture and deployment documentation
|-- scripts/                 # Utility scripts and local AI infrastructure
|-- AGENTS.md                # Detailed development and analytics instructions
|-- package.json             # Root workspace scripts
|-- pnpm-workspace.yaml      # Workspace packages, catalog, and overrides
`-- README.md

Current Foundation

The repository already includes:

• React 19 + Vite frontend in artifacts/anima-protocol
• Express API in artifacts/api-server, mounted under /api
• Shared Drizzle/Postgres package in lib/db
• User-scoped entity persistence
• Conversation and message persistence foundations
• Clerk authentication for frontend and API routes
• Consent-gated Mixpanel analytics
• Local Anima LLM infrastructure through Ollama/vLLM
• Gemini and OpenAI image-generation paths
• Optional ElevenLabs TTS
• Isolated mockup sandbox for UI experimentation

The highest-value product loop is:

create companion → start chat → retrieve memory → generate in-character response → persist the turn → deepen continuity

Use Node 24 before running workspace commands:

export NVM_DIR="$HOME/.nvm"
. "$NVM_DIR/nvm.sh"
export PATH="$NVM_DIR/versions/node/v24.16.0/bin:$PATH"

Install dependencies with pnpm only:

pnpm install --frozen-lockfile

For local Postgres development:

export DATABASE_URL=postgresql://anima:anima_dev@localhost:5432/anima_dev
pnpm --filter @workspace/db run push

One-command dev infra (Postgres + Anima LLM)

docker-compose.dev.yml bundles the two infra pieces that aren't already a pnpm dev process — Postgres and the self-hosted Anima LLM (Ollama, branded anima-chat from open Qwen2.5 weights) — so both come up together instead of running pnpm llm:up separately:

여기까지가 README 앞부분입니다. 나머지는 원문에서 읽을 수 있습니다. README 전문 보기 ↗