Engram-platform

Engram——人工智能记忆与智能平台

生产级多层AI存储平台。Engram为AI助手提供了跨项目的持久、可搜索的内存，内置了OSINT爬虫、知识图和统一的仪表板。

______________________________________________________________________

快速开始

# First-time setup (interactive env configuration + deploy)
./scripts/deploy-unified.sh init

# Or step by step:
./scripts/deploy-unified.sh setup    # configure .env interactively
./scripts/deploy-unified.sh up       # build and start all services

# Services:
#   Platform UI    http://localhost:3002
#   Memory API     http://localhost:8000
#   Crawler API    http://localhost:11235
#   MCP Server     http://localhost:3000
#   Weaviate       http://localhost:8080

完整部署指南 -> docs/01-deployment-manual.md

______________________________________________________________________

什么是Engram？

数据流： Crawler发现和抓取->内存API存储向量->MCP向AI客户端公开工具->平台提供UI。

______________________________________________________________________

建筑

+---------------------------------------------+
|  Engram-Platform  (Next.js 15, port 3002)   |
|  Dashboard . Memory browser . Crawler UI    |
+--------------+--------------+---------------+
|  Engram-MCP  | Crawler API  |  Memory API   |
|  port 3000   | port 11235   |  port 8000    |
+--------------+--------------+---------------+
|  Weaviate :8080  |  Redis x2 :6379/:6380    |
+------------------+--------------------------+

项目结构

Engram/
+-- Engram-AiMemory/        Python FastAPI memory system + CLI
+-- Engram-AiCrawler/       Python FastAPI OSINT crawler
+-- Engram-MCP/             TypeScript MCP server (canonical)
+-- Engram-Platform/        Next.js 15 dashboard (canonical UI)
|   +-- docker-compose.yml  Master orchestration file
+-- scripts/
|   +-- deploy-unified.sh   Canonical deployment entry point
+-- docs/                   Consolidated documentation
+-- engram-shared/          Shared Python utilities
+-- archive/                Retired artifacts and session docs
+-- plans/                  Implementation roadmaps

______________________________________________________________________

统一部署

所有部署都通过一个入口点：

./scripts/deploy-unified.sh 

通过委托可以访问每个子项目的遗留脚本：

./scripts/deploy-unified.sh deploy:production    # full production deploy
./scripts/deploy-unified.sh deploy:devnode       # devnode-optimized deploy
./scripts/deploy-unified.sh deploy:memory        # memory system deploy

______________________________________________________________________

文档

从这里开始

参考

运营

项目

______________________________________________________________________

开发设置

先决条件

• Docker&Docker编写
• Python 3.11+（用于AiMemory和AiCrawler开发）
• Node.js 20+（用于MCP和平台开发）

每个子项目开发

# AiMemory -- Python memory system + CLI
cd Engram-AiMemory
python3.11 -m venv .venv && source .venv/bin/activate
pip install -e ".[dev]"
make test

# AiCrawler -- FastAPI + React
cd Engram-AiCrawler/01_devroot
source .venv/bin/activate
uvicorn app.main:app --reload --port 11235

# MCP Server -- TypeScript (381 tests, OAuth 2.1)
cd Engram-MCP
npm install && npm run dev

# Platform -- Next.js 15 + React 19
cd Engram-Platform/frontend
npm install && npm run dev    # http://localhost:3002

共享公用设施

pip install -e ./engram-shared

______________________________________________________________________

关键端口

______________________________________________________________________

环境变量

跑 ./scripts/deploy-unified.sh setup 用于交互式配置或复制 .env.example:

cp Engram-Platform/.env.example Engram-Platform/.env

关键变量：

# Auth
NEXT_PUBLIC_CLERK_PUBLISHABLE_KEY=pk_...
CLERK_SECRET_KEY=sk_...
JWT_SECRET=your-secret-min-32-chars

# Embeddings
EMBEDDING_PROVIDER=deepinfra   # openai | deepinfra | nomic | local

# Network (never use 0.0.0.0 in production)
BIND_ADDRESS=127.0.0.1
TAILSCALE_HOSTNAME=your-host.tail1234.ts.net

# MCP
MCP_AUTH_TOKEN=your-mcp-token

看 docs/reference/environment-variables.md 查看完整列表。

______________________________________________________________________

测试

# AiMemory (Python, 80% coverage threshold)
cd Engram-AiMemory && make test

# AiCrawler (Python)
cd Engram-AiCrawler/01_devroot && pytest tests/ -v

# MCP (TypeScript, 381 tests)
cd Engram-MCP && npm test

# Platform (Vitest + Playwright)
cd Engram-Platform/frontend && npm run test:run && npm run test:e2e

______________________________________________________________________

生产

Engram部署在Tailscale上(*.tail4da6b7.ts.net).永远不要直接在公共IP上公开服务。

./scripts/deploy-unified.sh deploy             # full deploy with pre-flight
./scripts/deploy-unified.sh deploy --dry-run   # validate without changes
./scripts/deploy-unified.sh health             # verify all services

看 docs/01-deployment-manual.md 和 PRODUCTION_SETUP.md.

______________________________________________________________________

状态

当前： 约75%已完成——正在积极开发中（基线2026-03-17） 路线图： PROJECT_ROADMAP.md