AINote — AI Agent Workflow Platform

An AI agent workflow platform for building intelligent applications with visual flow orchestration, knowledge bases, and multi-model LLM integration.

Features

Visual Workflow Editor — Drag-and-drop flow builder with branching, loops, and AI nodes
Knowledge Base — Document ingestion, vector search (pgvector), and RAG pipelines
AI Agent Studio — Multi-model LLM integration (OpenAI, DeepSeek, Qwen, etc.), tool orchestration
Block-Note Editor — Rich-text collaborative editor with AI assistance
Code Sandbox — Secure remote code execution via OpenSandbox
Multi-Tenant — Organization-based workspace with invitation codes and member management
Extensible Storage — Local, Qiniu, Aliyun OSS, or S3-compatible backends

Architecture

┌──────────────────────────────┐
│         Client (React)       │  Port 5000 (dev) / 8081 (prod)
│   Vite + Ant Design + Flow   │
└─────────────┬────────────────┘
              │ HTTP / WebSocket
┌─────────────┴────────────────┐
│     Server (Express + Node)  │  Port 5001
│   REST API + Workflow Engine │
└──────┬───────┬───────┬───────┘
       │       │       │
┌──────┴─┐ ┌───┴───┐ ┌─┴────────┐
│Postgres│ │Temporal│ │Markitdown │
│+vector │ │ 7233   │ │ 6010      │
│+graph  │ └───────┘ └──────────┘
└────────┘

Quick Start (GHCR Images) — Easiest

Deploy with pre-built images from GitHub Container Registry. No source code, no build tools, no Node.js needed. Only Docker required.

Private repository? Login first: docker login ghcr.io -u YOUR_USERNAME

# 1. Download deployment files
curl -O https://raw.githubusercontent.com/yangzc/ainote/main/docker-compose.ghcr.yml
curl -O https://raw.githubusercontent.com/yangzc/ainote/main/.env.example
mkdir -p server
curl -o server/.env.example https://raw.githubusercontent.com/yangzc/ainote/main/server/.env.example

# 2. Prepare configuration
cp .env.example .env
cp server/.env.example server/.env

# 3. Edit the configs — at minimum set JWT_SECRET and one LLM API key
#    .env          → infrastructure (ports, DB password)
#    server/.env   → application (LLM keys, JWT secret, etc.)

# 4. Start all services
docker compose -f docker-compose.ghcr.yml up -d

# 5. Access
# Frontend: http://localhost:8081
# Backend:  http://localhost:5001

You only need these 3 files:

File	Purpose
`docker-compose.ghcr.yml`	Docker orchestration with pre-built images
`.env`	Infrastructure variables (ports, DB credentials)
`server/.env`	Application config (LLM API keys, JWT secret)

The stack includes: PostgreSQL (with pgvector, AGE, zhparser), Temporal server, backend, frontend, and Python Markitdown service — all running as containers.

Quick Start (Docker Compose — from source)

Build images locally from source. Requires cloning the repository and Docker.

Prerequisites

Dependency	Version	Required	Note
Node.js	≥ 20	Required	Backend + Frontend
pnpm	≥ 8	Required	Server & Client package manager
Python	≥ 3.10	Optional	Document conversion service
PostgreSQL	≥ 15	Required	With pgvector extension
Temporal	1.24+	Recommended	Workflow engine

Docker Compose (from source)

The fastest way to get everything running:

# 1. Clone the repository
git clone <repository_url>
cd ainote

# 2. Prepare environment
cp .env.example .env
cp server/.env.example server/.env
# Edit server/.env — at minimum set JWT_SECRET and one LLM API key

# 3. Start all services
docker compose up -d

# 4. Access the application
# Frontend: http://localhost:8081
# Backend:  http://localhost:5001
# Temporal UI (optional): docker compose --profile debug up -d  →  http://localhost:8233

The docker stack includes: PostgreSQL (with pgvector, AGE, zhparser), Temporal server, backend, frontend, and Python Markitdown service. Temporal UI is optional (use --profile debug).

Manual Setup (Bare-Metal)

1. Infrastructure Services

You'll need these running before starting the app:

PostgreSQL (with extensions)

# Install PostgreSQL 15+, then enable extensions:
psql -U postgres -c "CREATE EXTENSION IF NOT EXISTS vector;"
psql -U postgres -c "CREATE EXTENSION IF NOT EXISTS age;"
psql -U postgres -c "CREATE EXTENSION IF NOT EXISTS zhparser;"

Temporal (optional, for workflows)

# Using the provided script:
bash start-temporal.sh

# Or manually:
temporal server start-dev --db-port 5432

Python Markitdown (optional, for document conversion)

cd python
pip install -r requirements.txt
python server.py  # starts on port 6010

2. Server Setup

cd server

# Install dependencies
pnpm install

# Initialize database extensions
pnpm run db:init

# Configure environment
cp .env.example .env
# Edit .env — set JWT_SECRET, database URL, LLM keys (see Configuration below)

# Start in development mode
pnpm run dev

# Or start with workflow worker
pnpm run dev:worker

The server runs on http://localhost:5001 by default.

3. Client Setup

cd client

# Install dependencies
pnpm install

# Start development server
pnpm run dev

The client runs on http://localhost:5000 by default. No .env file is needed — all defaults work out of the box.

4. Verify

# Check backend health
curl http://localhost:5001/api/v1/health

# Open frontend
open http://localhost:5000

Minimal Configuration

Copy-paste the block below as server/.env. Replace <...> placeholders with your own values.
Client needs no .env file — VITE_API_URL defaults to /api/v1.

# ============================================================
# Server
# ============================================================
PORT=5001
CLIENT_ORIGIN=http://localhost:5173,http://localhost:5000,http://localhost:8081
MAX_FILE_SIZE_MB=10
MAX_ATTACHMENT_FILE_SIZE_MB=20
DEFAULT_TOKEN_BALANCE=100000

# ============================================================
# Security
# ============================================================
# Generate with: openssl rand -hex 64
JWT_SECRET=<your-jwt-secret>

# ============================================================
# Database (PostgreSQL + pgvector)
# ============================================================
VECTOR_POSTGRES_URL=postgresql://postgres:postgres@localhost:5432/ainote

# ============================================================
# Workflow Engine (Temporal)
# ============================================================
TEMPORAL_SERVER_URL=localhost:7233
TEMPORAL_NAMESPACE=default
TEMPORAL_TASK_QUEUE=ainote-workflows
START_TEMPORAL_WORKER=true

# ============================================================
# Storage (local | qiniu | oss | s3)
# ============================================================
STORAGE_PROVIDER=local
# QINIU_ACCESS_KEY=<your-qiniu-access-key>
# QINIU_SECRET_KEY=<your-qiniu-secret-key>
# QINIU_BUCKET=<your-bucket>
# QINIU_DOMAIN=<your-domain>

# ============================================================
# LLM Providers (at least one required)
#   Pattern: LLM_{PROVIDER}_{SETTING}
# ============================================================
# Provider: OpenAI / compatible API (required)
LLM_OPENAI_API_KEY=sk-<your-openai-key>
LLM_OPENAI_BASE_URL=https://api.openai.com/v1
LLM_OPENAI_MODEL=gpt-4o

# Provider: DeepSeek (optional)
# LLM_DEEPSEEK_API_KEY=sk-<your-deepseek-key>
# LLM_DEEPSEEK_BASE_URL=https://api.deepseek.com/v1
# LLM_DEEPSEEK_MODEL=deepseek-chat

# Provider: Qwen / DashScope (optional)
# LLM_QWEN_API_KEY=sk-<your-qwen-key>
# LLM_QWEN_BASE_URL=https://dashscope.aliyuncs.com/compatible-mode/v1
# LLM_QWEN_MODEL=qwen-plus

# Provider: OneAPI / proxy aggregator (optional)
# LLM_ONEAPI_API_KEY=sk-<your-oneapi-key>
# LLM_ONEAPI_BASE_URL=https://your-oneapi-host/v1
# LLM_ONEAPI_MODEL=gpt-4o,claude-3.5-sonnet

LLM_DEFAULT_PROVIDER=openai

# ============================================================
# Embedding (uses LLM provider API by default)
# ============================================================
EMBEDDING_PROVIDER=openai
EMBEDDING_API_URL=https://api.openai.com/v1
EMBEDDING_API_KEY=sk-<your-embedding-key>
EMBEDDING_MODEL_NAME=text-embedding-3-small
EMBEDDING_DIMENSION=1536

# ============================================================
# Vector Memory (mem0ai)
# ============================================================
MEMORY_PROVIDER=pgvector

# ============================================================
# Document Conversion (Markitdown)
# ============================================================
MARKITDOWN_SERVICE_URL=http://127.0.0.1:6010/v1/convert

# ============================================================
# Invitation
# ============================================================
FIXED_INVITATION_CODE=SIT2024
DEFAULT_INVITATION_SLOTS=5

# ============================================================
# WeTinker Gateway (optional, for enterprise)
# ============================================================
WETINKER_API_BASE_URL=

# ============================================================
# Sandbox (code execution, optional)
# ============================================================
SANDBOX_ENABLED=false
# SANDBOX_SERVER_URL=http://localhost:5002
# SANDBOX_API_KEY=<your-sandbox-key>
# SANDBOX_USE_SERVER_PROXY=false
# SANDBOX_IMAGE=python:3.12-alpine
# SANDBOX_TIMEOUT=300

完整变量说明见下方 Full Configuration Reference。

Full Configuration Reference

Server Environment Variables (`server/.env`)

All configuration lives in server/.env. Copy from the template:

cp server/.env.example server/.env

Required

Variable	Description
`JWT_SECRET`	Secret for signing JWT tokens. Generate: `openssl rand -hex 64`
`VECTOR_POSTGRES_URL`	PostgreSQL connection: `postgresql://user:pass@host:5432/ainote`
`LLM_OPENAI_API_KEY`	At least one LLM provider API key

LLM Providers

Supported providers: OpenAI, DeepSeek, Qwen, OneAPI (proxy aggregator).

LLM_OPENAI_API_KEY=sk-xxx
LLM_OPENAI_BASE_URL=https://api.openai.com/v1
LLM_OPENAI_MODEL=gpt-4o
LLM_DEFAULT_PROVIDER=openai

Additional providers can be configured with the pattern LLM_{PROVIDER}_{SETTING} — see .env.example for all options.

Database

Variable	Default	Description
`VECTOR_POSTGRES_URL`	—	Full PostgreSQL connection string

Storage

Variable	Default	Description
`STORAGE_PROVIDER`	`local`	`local`, `qiniu`, `oss`, or `s3`
`QINIU_ACCESS_KEY`	—	Qiniu access key
`QINIU_SECRET_KEY`	—	Qiniu secret key
`QINIU_BUCKET`	—	Qiniu bucket name
`QINIU_DOMAIN`	—	Qiniu CDN domain

Workflow Engine (Temporal)

Variable	Default
`TEMPORAL_SERVER_URL`	`localhost:7233`
`TEMPORAL_NAMESPACE`	`default`
`TEMPORAL_TASK_QUEUE`	`ainote-workflows`
`START_TEMPORAL_WORKER`	`true`

Sandbox (Code Execution)

Variable	Default	Description
`SANDBOX_ENABLED`	`false`	Enable remote code execution
`SANDBOX_SERVER_URL`	`localhost:5002`	Sandbox server address
`SANDBOX_API_KEY`	—	Authentication key

See sandbox setup guide for deployment instructions.

Other

Variable	Default	Description
`CLIENT_ORIGIN`	`http://localhost:5173,http://localhost:8081`	CORS origins (comma-separated)
`FIXED_INVITATION_CODE`	`SIT2024`	Registration invitation code
`DEFAULT_TOKEN_BALANCE`	`100000`	Initial token balance for new users
`EMBEDDING_PROVIDER`	`openai`	Embedding provider
`EMBEDDING_MODEL_NAME`	`text-embedding-3-small`	Embedding model
`MEMORY_PROVIDER`	`pgvector`	Vector memory backend (uses PostgreSQL pgvector)
`MARKITDOWN_SERVICE_URL`	`http://127.0.0.1:6010/v1/convert`	Document conversion endpoint

Project Structure

ainote/
├── client/                    # React frontend (Vite + Ant Design)
│   └── src/
│       ├── api/               # API client
│       ├── components/        # Shared UI components
│       ├── pages/             # Route pages
│       │   └── workflow/      # Visual workflow editor
│       ├── i18n.js            # Internationalization
│       └── sdk/               # Embeddable SDK
├── server/                    # Node.js/Express backend
│   ├── index.js               # App entry point
│   ├── config/                # Configuration (env, db, logger)
│   ├── routes/                # API routes
│   ├── services/              # Business logic services
│   ├── controllers/           # Request handlers
│   ├── middleware/             # Express middleware
│   ├── temporal/              # Temporal workflows & activities
│   └── scripts/               # Utility scripts (db init, seeds)
├── python/                    # Markitdown document conversion service
├── docker/                    # PostgreSQL custom image build
├── 1shared/                   # Shared code (client ↔ server)
├── sandbox/                   # OpenSandbox deployment guide
├── docs/                      # Documentation
├── docker-compose.yml         # Full-stack Docker orchestration
└── *.sh                       # Infrastructure start scripts

Development

Server

cd server
pnpm run dev         # Start with hot reload (nodemon + tsx)
pnpm run test        # Run tests (vitest)
pnpm run lint        # ESLint
pnpm run format      # Prettier
pnpm run db:push     # Push schema changes to database

Client

cd client
pnpm run dev         # Start dev server (port 5000)
pnpm run build       # Production build
pnpm run preview     # Preview production build

Viewing Logs

cat app.1.log | pino-pretty

License

MIT