init
This commit is contained in:
16
.env.example
Normal file
16
.env.example
Normal file
@@ -0,0 +1,16 @@
|
||||
# ChromaDB Configuration
|
||||
CHROMA_DB_PATH=/app/chroma_db
|
||||
CHROMA_COLLECTION_NAME=memories
|
||||
|
||||
# SQLite Configuration
|
||||
SQLITE_DB_PATH=/app/graph_store.db
|
||||
|
||||
# Ollama Configuration
|
||||
# When using Docker Compose with Ollama service, use: http://ollama:11434
|
||||
# When running Ollama on host machine, use: http://host.docker.internal:11434
|
||||
OLLAMA_HOST=http://ollama:11434
|
||||
OLLAMA_LLM_MODEL=llama3.2
|
||||
OLLAMA_EMBEDDER_MODEL=nomic-embed-text
|
||||
|
||||
# History Database Configuration
|
||||
HISTORY_DB_PATH=/app/history/history.db
|
||||
27
Dockerfile
Normal file
27
Dockerfile
Normal file
@@ -0,0 +1,27 @@
|
||||
FROM python:3.12-slim
|
||||
|
||||
WORKDIR /app
|
||||
|
||||
COPY requirements.txt .
|
||||
|
||||
RUN pip install --no-cache-dir -r requirements.txt
|
||||
|
||||
COPY . .
|
||||
|
||||
# Create directories for ChromaDB, SQLite and history database
|
||||
RUN mkdir -p /app/chroma_db /app/history
|
||||
|
||||
EXPOSE 8000
|
||||
|
||||
ENV PYTHONUNBUFFERED=1
|
||||
|
||||
# Default environment variables for ChromaDB, SQLite and Ollama
|
||||
ENV CHROMA_DB_PATH=/app/chroma_db
|
||||
ENV CHROMA_COLLECTION_NAME=memories
|
||||
ENV SQLITE_DB_PATH=/app/graph_store.db
|
||||
ENV OLLAMA_HOST=http://host.docker.internal:11434
|
||||
ENV OLLAMA_LLM_MODEL=llama3.2
|
||||
ENV OLLAMA_EMBEDDER_MODEL=nomic-embed-text
|
||||
ENV HISTORY_DB_PATH=/app/history/history.db
|
||||
|
||||
CMD ["uvicorn", "main:app", "--host", "0.0.0.0", "--port", "8000", "--reload"]
|
||||
7
Makefile
Normal file
7
Makefile
Normal file
@@ -0,0 +1,7 @@
|
||||
build:
|
||||
docker build -t mem0-api-server .
|
||||
|
||||
run_local:
|
||||
docker run -p 8000:8000 -v $(shell pwd):/app mem0-api-server --env-file .env
|
||||
|
||||
.PHONY: build run_local
|
||||
493
OLLAMA_DOCKER.md
Normal file
493
OLLAMA_DOCKER.md
Normal file
@@ -0,0 +1,493 @@
|
||||
# Ollama Docker Compose 部署指南
|
||||
|
||||
## 📋 概述
|
||||
|
||||
现在 Mem0 Server 的 docker-compose.yaml 已经包含了 Ollama 服务，所有组件都可以通过 Docker Compose 统一管理。
|
||||
|
||||
---
|
||||
|
||||
## 🏗️ 新架构
|
||||
|
||||
### 技术架构
|
||||
```
|
||||
┌─────────────┐
|
||||
│ Mem0 App │ (FastAPI, port: 8888)
|
||||
└──────┬──────┘
|
||||
│
|
||||
├──→ ┌─────────────┐
|
||||
│ │ Ollama │ (port: 11434)
|
||||
│ │ LLM+AI │
|
||||
│ └─────────────┘
|
||||
│
|
||||
├──→ ┌─────────────┐
|
||||
│ │ ChromaDB │ (Embedded)
|
||||
│ └─────────────┘
|
||||
│
|
||||
└──→ ┌─────────────┐
|
||||
│ SQLite │ (Embedded)
|
||||
└─────────────┘
|
||||
```
|
||||
|
||||
### 服务列表
|
||||
- `mem0` - FastAPI 应用（端口 8888）
|
||||
- `ollama` - Ollama AI 服务（端口 11434）
|
||||
|
||||
---
|
||||
|
||||
## 🚀 快速开始
|
||||
|
||||
### 方式 1：使用 GPU（推荐）
|
||||
|
||||
如果您的机器有 NVIDIA GPU：
|
||||
|
||||
```bash
|
||||
# 启动所有服务
|
||||
docker-compose --profile gpu up -d
|
||||
|
||||
# 查看状态
|
||||
docker-compose ps
|
||||
|
||||
# 查看日志
|
||||
docker-compose logs -f ollama
|
||||
```
|
||||
|
||||
### 方式 2：仅使用 CPU
|
||||
|
||||
如果没有 GPU：
|
||||
|
||||
```bash
|
||||
# 启动所有服务（不使用 GPU profile）
|
||||
docker-compose up -d
|
||||
|
||||
# 注意：Ollama 在 CPU 模式下可能较慢
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 📊 docker-compose.yaml 配置说明
|
||||
|
||||
### Ollama 服务配置
|
||||
|
||||
```yaml
|
||||
ollama:
|
||||
image: ollama/ollama:latest
|
||||
container_name: ollama
|
||||
restart: on-failure
|
||||
networks:
|
||||
- mem0_network
|
||||
healthcheck:
|
||||
test: ["CMD", "curl", "-f", "http://localhost:11434/api/tags"]
|
||||
interval: 5s
|
||||
timeout: 5s
|
||||
retries: 10
|
||||
start_period: 60s # 给 Ollama 60 秒启动时间
|
||||
volumes:
|
||||
- ollama_data:/root/.ollama # 持久化模型数据
|
||||
ports:
|
||||
- "11434:11434" # 暴露 Ollama API
|
||||
profiles: ["gpu"] # GPU 配置
|
||||
deploy:
|
||||
resources:
|
||||
reservations:
|
||||
devices:
|
||||
- driver: nvidia
|
||||
count: all
|
||||
capabilities: [gpu]
|
||||
```
|
||||
|
||||
### Mem0 服务配置更新
|
||||
|
||||
```yaml
|
||||
mem0:
|
||||
environment:
|
||||
- OLLAMA_HOST=http://ollama:11434 # 使用 Docker 网络访问 Ollama
|
||||
- OLLAMA_LLM_MODEL=llama3.2
|
||||
- OLLAMA_EMBEDDER_MODEL=nomic-embed-text
|
||||
depends_on:
|
||||
ollama:
|
||||
condition: service_healthy # 等待 Ollama 健康检查通过
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 🔧 使用步骤
|
||||
|
||||
### 1. 首次启动
|
||||
|
||||
```bash
|
||||
# 启动服务
|
||||
docker-compose --profile gpu up -d
|
||||
|
||||
# 等待 Ollama 启动完成（约 60 秒）
|
||||
docker-compose logs -f ollama
|
||||
|
||||
# 查看已加载的模型
|
||||
docker exec ollama ollama list
|
||||
```
|
||||
|
||||
### 2. 下载模型
|
||||
|
||||
首次使用需要下载模型：
|
||||
|
||||
```bash
|
||||
# 进入 Ollama 容器
|
||||
docker exec -it ollama bash
|
||||
|
||||
# 下载 LLM 模型
|
||||
ollama pull llama3.2
|
||||
|
||||
# 下载嵌入模型
|
||||
ollama pull nomic-embed-text
|
||||
|
||||
# 退出
|
||||
exit
|
||||
```
|
||||
|
||||
或者一次性下载：
|
||||
|
||||
```bash
|
||||
docker exec ollama ollama pull llama3.2
|
||||
docker exec ollama ollama pull nomic-embed-text
|
||||
```
|
||||
|
||||
### 3. 测试 Ollama
|
||||
|
||||
```bash
|
||||
# 测试 Ollama API
|
||||
curl http://localhost:11434/api/tags
|
||||
|
||||
# 测试生成
|
||||
curl http://localhost:11434/api/generate -d '{
|
||||
"model": "llama3.2",
|
||||
"prompt": "Hello, how are you?"
|
||||
}'
|
||||
```
|
||||
|
||||
### 4. 测试 Mem0 API
|
||||
|
||||
```bash
|
||||
# 访问 API 文档
|
||||
http://localhost:8888/docs
|
||||
|
||||
# 创建记忆
|
||||
curl -X POST http://localhost:8888/memories \
|
||||
-H "Content-Type: application/json" \
|
||||
-d '{
|
||||
"messages": [
|
||||
{"role": "user", "content": "你好"},
|
||||
{"role": "assistant", "content": "你好！有什么可以帮助你的？"}
|
||||
],
|
||||
"user_id": "test_user"
|
||||
}'
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 💾 数据持久化
|
||||
|
||||
### Ollama 数据卷
|
||||
- **名称**: `ollama_data`
|
||||
- **路径**: `/root/.ollama`
|
||||
- **内容**: 下载的模型文件、配置等
|
||||
|
||||
### 查看数据卷
|
||||
```bash
|
||||
# 查看数据卷信息
|
||||
docker volume inspect ollama_data
|
||||
|
||||
# 查看数据卷位置
|
||||
docker volume ls | grep ollama_data
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## ⚙️ 配置选项
|
||||
|
||||
### 使用不同的模型
|
||||
|
||||
编辑 `.env` 文件：
|
||||
|
||||
```bash
|
||||
OLLAMA_LLM_MODEL=mistral
|
||||
OLLAMA_EMBEDDER_MODEL=mxbai-embed-large
|
||||
```
|
||||
|
||||
重启服务：
|
||||
```bash
|
||||
docker-compose --profile gpu up -d
|
||||
```
|
||||
|
||||
### 调整 Ollama 资源
|
||||
|
||||
如果需要更多 GPU 资源：
|
||||
|
||||
```yaml
|
||||
deploy:
|
||||
resources:
|
||||
reservations:
|
||||
devices:
|
||||
- driver: nvidia
|
||||
device_ids: ['0', '1'] # 使用特定 GPU
|
||||
capabilities: [gpu]
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 🔍 故障排查
|
||||
|
||||
### Ollama 启动失败
|
||||
|
||||
```bash
|
||||
# 查看详细日志
|
||||
docker-compose logs ollama
|
||||
|
||||
# 检查健康状态
|
||||
docker-compose ps ollama
|
||||
|
||||
# 进入容器调试
|
||||
docker exec -it ollama bash
|
||||
```
|
||||
|
||||
### 模型下载慢
|
||||
|
||||
```bash
|
||||
# 设置国内镜像（如果可用）
|
||||
export OLLAMA_MODELS=https://mirror.ghproxy.com/https://github.com/ollama/ollama/releases/download
|
||||
|
||||
# 或者手动下载模型文件后挂载
|
||||
```
|
||||
|
||||
### GPU 不可用
|
||||
|
||||
```bash
|
||||
# 检查 NVIDIA Docker 是否安装
|
||||
nvidia-smi
|
||||
|
||||
# 检查容器是否识别 GPU
|
||||
docker exec ollama nvidia-smi
|
||||
|
||||
# 如果没有 GPU，使用 CPU 模式
|
||||
docker-compose up -d
|
||||
```
|
||||
|
||||
### Mem0 无法连接 Ollama
|
||||
|
||||
```bash
|
||||
# 检查网络连通性
|
||||
docker exec mem0 ping ollama
|
||||
|
||||
# 检查 Ollama 是否运行
|
||||
docker exec mem0 curl http://ollama:11434/api/tags
|
||||
|
||||
# 查看依赖关系
|
||||
docker-compose ps
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 🎯 优势特点
|
||||
|
||||
### 统一管理
|
||||
- ✅ 所有服务通过 docker-compose 管理
|
||||
- ✅ 一键启动/停止
|
||||
- ✅ 自动处理依赖关系
|
||||
|
||||
### 数据隔离
|
||||
- ✅ Ollama 数据独立存储
|
||||
- ✅ 不会污染宿主机环境
|
||||
- ✅ 易于备份和迁移
|
||||
|
||||
### 网络自动化
|
||||
- ✅ Docker 网络自动配置
|
||||
- ✅ 服务间通过服务名通信
|
||||
- ✅ 无需手动配置 IP
|
||||
|
||||
### 健康检查
|
||||
- ✅ 自动检测 Ollama 状态
|
||||
- ✅ Mem0 等待 Ollama 就绪
|
||||
- ✅ 提高系统可靠性
|
||||
|
||||
---
|
||||
|
||||
## 📈 性能优化
|
||||
|
||||
### GPU 加速
|
||||
```yaml
|
||||
# 启用 GPU 支持
|
||||
profiles: ["gpu"]
|
||||
deploy:
|
||||
resources:
|
||||
reservations:
|
||||
devices:
|
||||
- driver: nvidia
|
||||
count: all
|
||||
capabilities: [gpu]
|
||||
```
|
||||
|
||||
### 内存优化
|
||||
```yaml
|
||||
# 限制最大内存使用
|
||||
deploy:
|
||||
resources:
|
||||
limits:
|
||||
memory: 8G
|
||||
reservations:
|
||||
memory: 4G
|
||||
```
|
||||
|
||||
### 并发优化
|
||||
```bash
|
||||
# 设置 Ollama 并发数
|
||||
docker exec ollama sh -c 'echo "OLLAMA_NUM_PARALLEL=4" >> /root/.ollama/config.json'
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 🔄 常用命令
|
||||
|
||||
### 启动服务
|
||||
```bash
|
||||
# 带 GPU 支持
|
||||
docker-compose --profile gpu up -d
|
||||
|
||||
# 仅 CPU
|
||||
docker-compose up -d
|
||||
```
|
||||
|
||||
### 停止服务
|
||||
```bash
|
||||
docker-compose --profile gpu down
|
||||
```
|
||||
|
||||
### 查看状态
|
||||
```bash
|
||||
# 查看所有服务
|
||||
docker-compose ps
|
||||
|
||||
# 查看 Ollama 日志
|
||||
docker-compose logs -f ollama
|
||||
|
||||
# 查看 Mem0 日志
|
||||
docker-compose logs -f mem0
|
||||
```
|
||||
|
||||
### 重启服务
|
||||
```bash
|
||||
# 重启 Ollama
|
||||
docker-compose restart ollama
|
||||
|
||||
# 重启 Mem0
|
||||
docker-compose restart mem0
|
||||
|
||||
# 重启所有
|
||||
docker-compose restart
|
||||
```
|
||||
|
||||
### 进入容器
|
||||
```bash
|
||||
# 进入 Ollama 容器
|
||||
docker exec -it ollama bash
|
||||
|
||||
# 进入 Mem0 容器
|
||||
docker exec -it mem0 bash
|
||||
```
|
||||
|
||||
### 管理模型
|
||||
```bash
|
||||
# 列出已安装的模型
|
||||
docker exec ollama ollama list
|
||||
|
||||
# 下载新模型
|
||||
docker exec ollama ollama pull mistral
|
||||
|
||||
# 删除模型
|
||||
docker exec ollama ollama rm llama3.2
|
||||
|
||||
# 查看模型信息
|
||||
docker exec ollama ollama show llama3.2
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 💡 最佳实践
|
||||
|
||||
### 1. 首次部署
|
||||
```bash
|
||||
# 1. 启动服务
|
||||
docker-compose --profile gpu up -d
|
||||
|
||||
# 2. 等待 Ollama 就绪
|
||||
sleep 60
|
||||
|
||||
# 3. 下载所需模型
|
||||
docker exec ollama ollama pull llama3.2
|
||||
docker exec ollama ollama pull nomic-embed-text
|
||||
|
||||
# 4. 测试 API
|
||||
curl http://localhost:8888/docs
|
||||
```
|
||||
|
||||
### 2. 日常使用
|
||||
```bash
|
||||
# 启动
|
||||
docker-compose --profile gpu up -d
|
||||
|
||||
# 停止
|
||||
docker-compose --profile gpu down
|
||||
|
||||
# 重启
|
||||
docker-compose --profile gpu restart
|
||||
```
|
||||
|
||||
### 3. 数据备份
|
||||
```bash
|
||||
# 备份 Ollama 数据
|
||||
docker run --rm \
|
||||
-v ollama_data:/data \
|
||||
-v $(pwd):/backup \
|
||||
alpine tar czf /backup/ollama_backup.tar.gz /data
|
||||
```
|
||||
|
||||
### 4. 恢复数据
|
||||
```bash
|
||||
# 恢复 Ollama 数据
|
||||
docker run --rm \
|
||||
-v ollama_data:/data \
|
||||
-v $(pwd):/backup \
|
||||
alpine tar xzf /backup/ollama_backup.tar.gz -C /
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 🆘 常见问题
|
||||
|
||||
### Q1: Ollama 启动很慢？
|
||||
**A:** Ollama 首次启动需要加载模型，可能需要 60 秒。健康检查配置了 `start_period: 60s` 来等待。
|
||||
|
||||
### Q2: 如何切换模型？
|
||||
**A:** 修改 `.env` 中的 `OLLAMA_LLM_MODEL`，然后重启服务，并确保新模型已下载。
|
||||
|
||||
### Q3: GPU 内存不足？
|
||||
**A:** 使用更小的模型（如 `phi3`），或限制并发数。
|
||||
|
||||
### Q4: 如何完全清理？
|
||||
**A:**
|
||||
```bash
|
||||
docker-compose --profile gpu down -v # 删除所有数据卷
|
||||
```
|
||||
|
||||
### Q5: 可以在没有 GPU 的情况下运行吗？
|
||||
**A:** 可以，但速度较慢。使用 `docker-compose up -d`（不带 gpu profile）。
|
||||
|
||||
---
|
||||
|
||||
## 📚 相关文档
|
||||
|
||||
- [README.md](./README.md) - 项目主文档
|
||||
- [TECH_STACK.md](./TECH_STACK.md) - 技术架构说明
|
||||
- [QUICK_START.md](./QUICK_START.md) - 快速入门
|
||||
- [DEPLOYMENT.md](./DEPLOYMENT.md) - 详细部署指南
|
||||
|
||||
---
|
||||
|
||||
现在所有服务都集成在 docker-compose.yaml 中，管理更加简单统一！🎉
|
||||
105
README.md
Normal file
105
README.md
Normal file
@@ -0,0 +1,105 @@
|
||||
# Mem0 REST API Server
|
||||
|
||||
Mem0 provides a REST API server (written using FastAPI). Users can perform all operations through REST endpoints. The API also includes OpenAPI documentation, accessible at `/docs` when the server is running.
|
||||
|
||||
## Features
|
||||
|
||||
- **Create memories:** Create memories based on messages for a user, agent, or run.
|
||||
- **Retrieve memories:** Get all memories for a given user, agent, or run.
|
||||
- **Search memories:** Search stored memories based on a query.
|
||||
- **Update memories:** Update an existing memory.
|
||||
- **Delete memories:** Delete a specific memory or all memories for a user, agent, or run.
|
||||
- **Reset memories:** Reset all memories for a user, agent, or run.
|
||||
- **OpenAPI Documentation:** Accessible via `/docs` endpoint.
|
||||
|
||||
## Technology Stack
|
||||
|
||||
- **Vector Store**: ChromaDB - 高性能向量数据库
|
||||
- **Graph Store**: SQLite - 轻量级图存储
|
||||
- **LLM & Embedder**: Ollama - 本地大模型服务
|
||||
|
||||
## Quick Start
|
||||
|
||||
### 方式 1: Docker Compose（最简单）
|
||||
|
||||
```bash
|
||||
# 1. 复制环境变量配置
|
||||
cp .env.example .env
|
||||
|
||||
# 2. 编辑 .env 文件，根据需要修改配置
|
||||
# vi .env 或 nano .env
|
||||
|
||||
# 3. 启动所有服务（包括 Ollama）
|
||||
docker-compose --profile gpu up -d
|
||||
|
||||
# 4. 访问 API 文档
|
||||
open http://localhost:8888/docs
|
||||
```
|
||||
|
||||
**注意**: 首次启动需要下载 Ollama 模型，请等待约 60 秒。
|
||||
|
||||
### 方式 2: Docker 命令部署
|
||||
|
||||
```bash
|
||||
# 1. 构建镜像
|
||||
docker build -t mem0-server .
|
||||
|
||||
# 2. 运行容器（通过环境变量传入配置）
|
||||
docker run -d \
|
||||
--name mem0-server \
|
||||
-p 8000:8000 \
|
||||
-v chroma_data:/app/chroma_db \
|
||||
-v sqlite_data:/app/graph_store.db \
|
||||
-v history_data:/app/history \
|
||||
--add-host=host.docker.internal:host-gateway \
|
||||
-e CHROMA_DB_PATH=/app/chroma_db \
|
||||
-e CHROMA_COLLECTION_NAME=memories \
|
||||
-e SQLITE_DB_PATH=/app/graph_store.db \
|
||||
-e OLLAMA_HOST=http://host.docker.internal:11434 \
|
||||
-e OLLAMA_LLM_MODEL=llama3.2 \
|
||||
-e OLLAMA_EMBEDDER_MODEL=nomic-embed-text \
|
||||
-e REDIS_HOST=host.docker.internal \
|
||||
-e REDIS_PORT=6379 \
|
||||
mem0-server
|
||||
```
|
||||
|
||||
## 环境变量配置
|
||||
|
||||
所有配置都可以通过环境变量传入：
|
||||
|
||||
| 变量名 | 说明 | 默认值 |
|
||||
|--------|------|--------|
|
||||
| `CHROMA_DB_PATH` | ChromaDB 数据路径 | `/app/chroma_db` |
|
||||
| `CHROMA_COLLECTION_NAME` | ChromaDB 集合名称 | `memories` |
|
||||
| `SQLITE_DB_PATH` | SQLite 数据库路径 | `/app/graph_store.db` |
|
||||
| `OLLAMA_HOST` | Ollama 服务地址 | `http://localhost:11434` |
|
||||
| `OLLAMA_LLM_MODEL` | LLM 模型名称 | `llama3.2` |
|
||||
| `OLLAMA_EMBEDDER_MODEL` | Embedder 模型名称 | `nomic-embed-text` |
|
||||
| `HISTORY_DB_PATH` | 历史数据库路径 | `/app/history/history.db` |
|
||||
|
||||
详细部署文档请查看 [DEPLOYMENT.md](./DEPLOYMENT.md)
|
||||
|
||||
## Prerequisites
|
||||
|
||||
Before running the server, ensure you have:
|
||||
|
||||
1. **Ollama installed and running** on your host machine
|
||||
```bash
|
||||
# Install Ollama from https://ollama.ai
|
||||
ollama pull llama3.2
|
||||
ollama pull nomic-embed-text
|
||||
```
|
||||
|
||||
2. **Docker and Docker Compose** installed
|
||||
|
||||
## API Endpoints
|
||||
|
||||
- `POST /memories` - Create new memories
|
||||
- `GET /memories` - Retrieve all memories
|
||||
- `GET /memories/{memory_id}` - Get a specific memory
|
||||
- `POST /search` - Search memories
|
||||
- `PUT /memories/{memory_id}` - Update a memory
|
||||
- `DELETE /memories/{memory_id}` - Delete a memory
|
||||
- `DELETE /memories` - Delete all memories for an identifier
|
||||
- `POST /reset` - Reset all memories
|
||||
- `GET /docs` - OpenAPI documentation
|
||||
284
TECH_STACK.md
Normal file
284
TECH_STACK.md
Normal file
@@ -0,0 +1,284 @@
|
||||
# Mem0 Server 技术架构
|
||||
|
||||
## 📋 技术栈
|
||||
|
||||
Mem0 Server 采用简洁高效的技术架构，所有组件均通过环境变量配置。
|
||||
|
||||
### 核心组件
|
||||
|
||||
| 组件 | 技术 | 用途 | 存储方式 |
|
||||
|------|------|------|----------|
|
||||
| **Vector Store** | ChromaDB | 向量存储，用于语义搜索 | 文件系统 |
|
||||
| **Graph Store** | SQLite | 图存储，管理实体关系 | 文件系统 |
|
||||
| **LLM** | Ollama | 大语言模型，文本生成 | 外部服务 |
|
||||
| **Embedder** | Ollama | 嵌入模型，向量化文本 | 外部服务 |
|
||||
|
||||
---
|
||||
|
||||
## 🏗️ 架构图
|
||||
|
||||
```
|
||||
┌─────────────┐
|
||||
│ Mem0 App │ (FastAPI, port: 8888)
|
||||
│ main.py │
|
||||
└──────┬──────┘
|
||||
│
|
||||
├──→ ┌─────────────┐
|
||||
│ │ ChromaDB │ (Embedded)
|
||||
│ │ Vectors │
|
||||
│ └─────────────┘
|
||||
│
|
||||
├──→ ┌─────────────┐
|
||||
│ │ SQLite │ (Embedded)
|
||||
│ │ Graph │
|
||||
│ └─────────────┘
|
||||
│
|
||||
└──→ ┌─────────────┐
|
||||
│ Ollama │ (External Service)
|
||||
│ LLM + AI │
|
||||
└─────────────┘
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 🔧 技术选型理由
|
||||
|
||||
### ChromaDB（向量存储）
|
||||
- ✅ **轻量级**：无需独立数据库服务
|
||||
- ✅ **易部署**：嵌入式设计，开箱即用
|
||||
- ✅ **高性能**：专为向量相似度搜索优化
|
||||
- ✅ **持久化**：数据直接存储到文件系统
|
||||
|
||||
### SQLite（图存储）
|
||||
- ✅ **零配置**：无需安装和配置
|
||||
- ✅ **单文件**：数据存储在单个文件中
|
||||
- ✅ **成熟稳定**：经过广泛验证的数据库
|
||||
- ✅ **事务支持**：ACID 事务保证数据一致性
|
||||
|
||||
### Ollama（AI 服务）
|
||||
- ✅ **本地运行**：数据隐私和安全
|
||||
- ✅ **模型丰富**：支持多种开源模型
|
||||
- ✅ **易于使用**：简单的 API 接口
|
||||
- ✅ **性能优秀**：优化的推理引擎
|
||||
|
||||
---
|
||||
|
||||
## 📊 环境变量配置
|
||||
|
||||
所有配置通过环境变量传入，实现配置与代码分离。
|
||||
|
||||
### 必需的环境变量
|
||||
|
||||
```bash
|
||||
# ChromaDB 配置
|
||||
CHROMA_DB_PATH=/app/chroma_db
|
||||
CHROMA_COLLECTION_NAME=memories
|
||||
|
||||
# SQLite 配置
|
||||
SQLITE_DB_PATH=/app/graph_store.db
|
||||
|
||||
# Ollama 配置
|
||||
OLLAMA_HOST=http://host.docker.internal:11434
|
||||
OLLAMA_LLM_MODEL=llama3.2
|
||||
OLLAMA_EMBEDDER_MODEL=nomic-embed-text
|
||||
|
||||
# 历史数据库配置
|
||||
HISTORY_DB_PATH=/app/history/history.db
|
||||
```
|
||||
|
||||
### 环境变量说明
|
||||
|
||||
| 变量名 | 说明 | 默认值 | 示例 |
|
||||
|--------|------|--------|------|
|
||||
| `CHROMA_DB_PATH` | ChromaDB 数据存储路径 | `/app/chroma_db` | `/data/chroma` |
|
||||
| `CHROMA_COLLECTION_NAME` | ChromaDB 集合名称 | `memories` | `my_memories` |
|
||||
| `SQLITE_DB_PATH` | SQLite 数据库路径 | `/app/graph_store.db` | `/data/graph.db` |
|
||||
| `OLLAMA_HOST` | Ollama 服务地址 | `http://localhost:11434` | `http://ollama:11434` |
|
||||
| `OLLAMA_LLM_MODEL` | LLM 模型名称 | `llama3.2` | `mistral`, `llama3.1` |
|
||||
| `OLLAMA_EMBEDDER_MODEL` | Embedder 模型名称 | `nomic-embed-text` | `mxbai-embed-large` |
|
||||
| `HISTORY_DB_PATH` | 历史数据库路径 | `/app/history/history.db` | `/data/history.db` |
|
||||
|
||||
---
|
||||
|
||||
## 🚀 部署方式
|
||||
|
||||
### Docker Compose（推荐）
|
||||
|
||||
```yaml
|
||||
services:
|
||||
mem0:
|
||||
build: .
|
||||
ports:
|
||||
- "8888:8000"
|
||||
volumes:
|
||||
- ./chroma_db:/app/chroma_db
|
||||
- ./history:/app/history
|
||||
environment:
|
||||
- CHROMA_DB_PATH=/app/chroma_db
|
||||
- CHROMA_COLLECTION_NAME=memories
|
||||
- SQLITE_DB_PATH=/app/graph_store.db
|
||||
- OLLAMA_HOST=http://host.docker.internal:11434
|
||||
- OLLAMA_LLM_MODEL=llama3.2
|
||||
- OLLAMA_EMBEDDER_MODEL=nomic-embed-text
|
||||
```
|
||||
|
||||
启动命令：
|
||||
```bash
|
||||
docker-compose up -d
|
||||
```
|
||||
|
||||
### Docker 命令
|
||||
|
||||
```bash
|
||||
docker run -d \
|
||||
--name mem0-server \
|
||||
-p 8888:8000 \
|
||||
-v $(pwd)/chroma_db:/app/chroma_db \
|
||||
-v $(pwd)/history:/app/history \
|
||||
-e CHROMA_DB_PATH=/app/chroma_db \
|
||||
-e CHROMA_COLLECTION_NAME=memories \
|
||||
-e SQLITE_DB_PATH=/app/graph_store.db \
|
||||
-e OLLAMA_HOST=http://host.docker.internal:11434 \
|
||||
-e OLLAMA_LLM_MODEL=llama3.2 \
|
||||
-e OLLAMA_EMBEDDER_MODEL=nomic-embed-text \
|
||||
mem0-server
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 💾 数据持久化
|
||||
|
||||
### ChromaDB 数据
|
||||
- **位置**：`./chroma_db/`
|
||||
- **内容**：向量数据、集合元数据
|
||||
- **格式**：ChromaDB 内部格式
|
||||
|
||||
### SQLite 数据
|
||||
- **位置**：`./graph_store.db`（或自定义路径）
|
||||
- **内容**：图结构数据、实体关系
|
||||
- **格式**：SQLite 数据库文件
|
||||
|
||||
### 历史数据
|
||||
- **位置**：`./history/history.db`
|
||||
- **内容**：记忆变更历史记录
|
||||
- **格式**：SQLite 数据库文件
|
||||
|
||||
---
|
||||
|
||||
## 🔍 工作流程
|
||||
|
||||
### 1. 创建记忆
|
||||
```
|
||||
用户消息 → FastAPI → Memory.add()
|
||||
↓
|
||||
Ollama Embedder → 生成向量
|
||||
↓
|
||||
ChromaDB → 存储向量
|
||||
↓
|
||||
SQLite → 存储实体关系
|
||||
↓
|
||||
返回结果
|
||||
```
|
||||
|
||||
### 2. 搜索记忆
|
||||
```
|
||||
搜索查询 → FastAPI → Memory.search()
|
||||
↓
|
||||
Ollama Embedder → 查询向量化
|
||||
↓
|
||||
ChromaDB → 向量相似度搜索
|
||||
↓
|
||||
返回匹配的记忆
|
||||
```
|
||||
|
||||
### 3. 获取记忆
|
||||
```
|
||||
用户 ID → FastAPI → Memory.get_all()
|
||||
↓
|
||||
SQLite → 查询图关系
|
||||
↓
|
||||
ChromaDB → 获取相关向量
|
||||
↓
|
||||
返回结构化记忆
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 🎯 优势特点
|
||||
|
||||
### 简洁性
|
||||
- ✅ 无外部数据库依赖
|
||||
- ✅ 配置文件少
|
||||
- ✅ 易于理解和维护
|
||||
|
||||
### 灵活性
|
||||
- ✅ 所有配置通过环境变量
|
||||
- ✅ 支持自定义模型
|
||||
- ✅ 可切换不同组件
|
||||
|
||||
### 可扩展性
|
||||
- ✅ ChromaDB 支持集群部署
|
||||
- ✅ SQLite 可替换为 PostgreSQL
|
||||
- ✅ Ollama 可替换为其他 LLM 服务
|
||||
|
||||
### 数据隐私
|
||||
- ✅ 本地运行 AI 模型
|
||||
- ✅ 数据不上传第三方
|
||||
- ✅ 完全可控的数据流
|
||||
|
||||
---
|
||||
|
||||
## 📈 性能特性
|
||||
|
||||
### ChromaDB
|
||||
- 毫秒级向量相似度搜索
|
||||
- 支持百万级向量规模
|
||||
- 内存映射加速查询
|
||||
|
||||
### SQLite
|
||||
- 微秒级简单查询
|
||||
- 事务安全保证
|
||||
- 自动索引优化
|
||||
|
||||
### Ollama
|
||||
- GPU 加速推理（如可用）
|
||||
- 批量处理优化
|
||||
- 模型缓存机制
|
||||
|
||||
---
|
||||
|
||||
## 🛠️ 开发建议
|
||||
|
||||
### 本地开发
|
||||
```bash
|
||||
# 使用开发模式运行
|
||||
docker-compose -f docker-compose.yaml up --build
|
||||
|
||||
# 查看日志
|
||||
docker-compose logs -f mem0
|
||||
```
|
||||
|
||||
### 生产部署
|
||||
- 使用固定版本镜像
|
||||
- 配置数据卷备份
|
||||
- 监控资源使用
|
||||
- 设置健康检查
|
||||
|
||||
### 性能优化
|
||||
- 调整 ChromaDB 批次大小
|
||||
- 优化 SQLite 索引
|
||||
- 使用更快的存储介质（SSD）
|
||||
- 增加 Ollama 并发配置
|
||||
|
||||
---
|
||||
|
||||
## 📚 相关文档
|
||||
|
||||
- [README.md](./README.md) - 项目主文档
|
||||
- [DEPLOYMENT.md](./DEPLOYMENT.md) - 详细部署指南
|
||||
- [QUICK_START.md](./QUICK_START.md) - 快速入门
|
||||
- [ENV_VARIABLES.md](./ENV_VARIABLES.md) - 环境变量详解
|
||||
|
||||
---
|
||||
|
||||
这个简洁的技术架构专注于核心功能，避免了不必要的复杂性，同时保持了强大的扩展能力！🎉
|
||||
25
dev.Dockerfile
Normal file
25
dev.Dockerfile
Normal file
@@ -0,0 +1,25 @@
|
||||
FROM python:3.12
|
||||
|
||||
WORKDIR /app
|
||||
|
||||
# Install Poetry
|
||||
RUN curl -sSL https://install.python-poetry.org | python3 -
|
||||
ENV PATH="/root/.local/bin:$PATH"
|
||||
|
||||
# Copy requirements first for better caching
|
||||
COPY server/requirements.txt .
|
||||
RUN pip install -r requirements.txt
|
||||
|
||||
# Install mem0 in editable mode using Poetry
|
||||
WORKDIR /app/packages
|
||||
COPY pyproject.toml .
|
||||
COPY poetry.lock .
|
||||
COPY README.md .
|
||||
COPY mem0 ./mem0
|
||||
RUN pip install -e .[graph]
|
||||
|
||||
# Return to app directory and copy server code
|
||||
WORKDIR /app
|
||||
COPY server .
|
||||
|
||||
CMD ["uvicorn", "main:app", "--host", "0.0.0.0", "--port", "8000", "--reload"]
|
||||
57
docker-compose.yaml
Normal file
57
docker-compose.yaml
Normal file
@@ -0,0 +1,57 @@
|
||||
name: mem0-dev
|
||||
|
||||
services:
|
||||
mem0:
|
||||
build:
|
||||
context: .. # Set context to parent directory
|
||||
dockerfile: server/dev.Dockerfile
|
||||
ports:
|
||||
- "8888:8000"
|
||||
env_file:
|
||||
- .env
|
||||
networks:
|
||||
- mem0_network
|
||||
volumes:
|
||||
- ./chroma_db:/app/chroma_db # ChromaDB storage
|
||||
- ./history:/app/history # History db location
|
||||
- .:/app # Server code. This allows to reload the app when the server code is updated
|
||||
- ../mem0:/app/packages/mem0 # Mem0 library. This allows to reload the app when the library code is updated
|
||||
command: uvicorn main:app --host 0.0.0.0 --port 8000 --reload # Enable auto-reload
|
||||
environment:
|
||||
- PYTHONDONTWRITEBYTECODE=1 # Prevents Python from writing .pyc files
|
||||
- PYTHONUNBUFFERED=1 # Ensures Python output is sent straight to terminal
|
||||
- CHROMA_DB_PATH=/app/chroma_db
|
||||
- CHROMA_COLLECTION_NAME=memories
|
||||
- SQLITE_DB_PATH=/app/graph_store.db
|
||||
- OLLAMA_HOST=http://ollama:11434
|
||||
- OLLAMA_LLM_MODEL=llama3.2
|
||||
- OLLAMA_EMBEDDER_MODEL=nomic-embed-text
|
||||
- HISTORY_DB_PATH=/app/history/history.db
|
||||
depends_on:
|
||||
ollama:
|
||||
condition: service_healthy
|
||||
|
||||
ollama:
|
||||
image: ollama/ollama:latest
|
||||
container_name: ollama
|
||||
restart: on-failure
|
||||
networks:
|
||||
- mem0_network
|
||||
healthcheck:
|
||||
test: ["CMD", "curl", "-f", "http://localhost:11434/api/tags"]
|
||||
interval: 5s
|
||||
timeout: 5s
|
||||
retries: 10
|
||||
start_period: 60s
|
||||
volumes:
|
||||
- ollama_data:/root/.ollama
|
||||
ports:
|
||||
- "11434:11434"
|
||||
profiles: ["gpu"]
|
||||
|
||||
volumes:
|
||||
ollama_data: {}
|
||||
|
||||
networks:
|
||||
mem0_network:
|
||||
driver: bridge
|
||||
229
main.py
Normal file
229
main.py
Normal file
@@ -0,0 +1,229 @@
|
||||
import logging
|
||||
import os
|
||||
from typing import Any, Dict, List, Optional
|
||||
|
||||
from dotenv import load_dotenv
|
||||
from fastapi import FastAPI, HTTPException
|
||||
from fastapi.responses import JSONResponse, RedirectResponse
|
||||
from pydantic import BaseModel, Field
|
||||
|
||||
from mem0 import Memory
|
||||
|
||||
logging.basicConfig(level=logging.INFO, format="%(asctime)s - %(levelname)s - %(message)s")
|
||||
|
||||
# Load environment variables
|
||||
load_dotenv()
|
||||
|
||||
# ChromaDB configuration
|
||||
CHROMA_DB_PATH = os.environ.get("CHROMA_DB_PATH", "/app/chroma_db")
|
||||
CHROMA_COLLECTION_NAME = os.environ.get("CHROMA_COLLECTION_NAME", "memories")
|
||||
|
||||
# SQLite configuration for graph store
|
||||
SQLITE_DB_PATH = os.environ.get("SQLITE_DB_PATH", "/app/graph_store.db")
|
||||
|
||||
# Ollama configuration
|
||||
OLLAMA_HOST = os.environ.get("OLLAMA_HOST", "http://localhost:11434")
|
||||
OLLAMA_LLM_MODEL = os.environ.get("OLLAMA_LLM_MODEL", "llama3.2")
|
||||
OLLAMA_EMBEDDER_MODEL = os.environ.get("OLLAMA_EMBEDDER_MODEL", "nomic-embed-text")
|
||||
|
||||
HISTORY_DB_PATH = os.environ.get("HISTORY_DB_PATH", "/app/history/history.db")
|
||||
|
||||
DEFAULT_CONFIG = {
|
||||
"version": "v1.1",
|
||||
"vector_store": {
|
||||
"provider": "chroma",
|
||||
"config": {
|
||||
"path": CHROMA_DB_PATH,
|
||||
"collection_name": CHROMA_COLLECTION_NAME,
|
||||
},
|
||||
},
|
||||
"graph_store": {
|
||||
"provider": "sqlite",
|
||||
"config": {"db_path": SQLITE_DB_PATH},
|
||||
},
|
||||
"llm": {
|
||||
"provider": "ollama",
|
||||
"config": {
|
||||
"host": OLLAMA_HOST,
|
||||
"model": OLLAMA_LLM_MODEL,
|
||||
"temperature": 0.2,
|
||||
},
|
||||
},
|
||||
"embedder": {
|
||||
"provider": "ollama",
|
||||
"config": {
|
||||
"host": OLLAMA_HOST,
|
||||
"model": OLLAMA_EMBEDDER_MODEL,
|
||||
},
|
||||
},
|
||||
"history_db_path": HISTORY_DB_PATH,
|
||||
}
|
||||
|
||||
|
||||
MEMORY_INSTANCE = Memory.from_config(DEFAULT_CONFIG)
|
||||
|
||||
app = FastAPI(
|
||||
title="Mem0 REST APIs",
|
||||
description="A REST API for managing and searching memories for your AI Agents and Apps.",
|
||||
version="1.0.0",
|
||||
)
|
||||
|
||||
|
||||
class Message(BaseModel):
|
||||
role: str = Field(..., description="Role of the message (user or assistant).")
|
||||
content: str = Field(..., description="Message content.")
|
||||
|
||||
|
||||
class MemoryCreate(BaseModel):
|
||||
messages: List[Message] = Field(..., description="List of messages to store.")
|
||||
user_id: Optional[str] = None
|
||||
agent_id: Optional[str] = None
|
||||
run_id: Optional[str] = None
|
||||
metadata: Optional[Dict[str, Any]] = None
|
||||
|
||||
|
||||
class SearchRequest(BaseModel):
|
||||
query: str = Field(..., description="Search query.")
|
||||
user_id: Optional[str] = None
|
||||
run_id: Optional[str] = None
|
||||
agent_id: Optional[str] = None
|
||||
filters: Optional[Dict[str, Any]] = None
|
||||
|
||||
|
||||
@app.post("/configure", summary="Configure Mem0")
|
||||
def set_config(config: Dict[str, Any]):
|
||||
"""Set memory configuration."""
|
||||
global MEMORY_INSTANCE
|
||||
MEMORY_INSTANCE = Memory.from_config(config)
|
||||
return {"message": "Configuration set successfully"}
|
||||
|
||||
|
||||
@app.post("/memories", summary="Create memories")
|
||||
def add_memory(memory_create: MemoryCreate):
|
||||
"""Store new memories."""
|
||||
if not any([memory_create.user_id, memory_create.agent_id, memory_create.run_id]):
|
||||
raise HTTPException(status_code=400, detail="At least one identifier (user_id, agent_id, run_id) is required.")
|
||||
|
||||
params = {k: v for k, v in memory_create.model_dump().items() if v is not None and k != "messages"}
|
||||
try:
|
||||
response = MEMORY_INSTANCE.add(messages=[m.model_dump() for m in memory_create.messages], **params)
|
||||
return JSONResponse(content=response)
|
||||
except Exception as e:
|
||||
logging.exception("Error in add_memory:") # This will log the full traceback
|
||||
raise HTTPException(status_code=500, detail=str(e))
|
||||
|
||||
|
||||
@app.get("/memories", summary="Get memories")
|
||||
def get_all_memories(
|
||||
user_id: Optional[str] = None,
|
||||
run_id: Optional[str] = None,
|
||||
agent_id: Optional[str] = None,
|
||||
):
|
||||
"""Retrieve stored memories."""
|
||||
if not any([user_id, run_id, agent_id]):
|
||||
raise HTTPException(status_code=400, detail="At least one identifier is required.")
|
||||
try:
|
||||
params = {
|
||||
k: v for k, v in {"user_id": user_id, "run_id": run_id, "agent_id": agent_id}.items() if v is not None
|
||||
}
|
||||
return MEMORY_INSTANCE.get_all(**params)
|
||||
except Exception as e:
|
||||
logging.exception("Error in get_all_memories:")
|
||||
raise HTTPException(status_code=500, detail=str(e))
|
||||
|
||||
|
||||
@app.get("/memories/{memory_id}", summary="Get a memory")
|
||||
def get_memory(memory_id: str):
|
||||
"""Retrieve a specific memory by ID."""
|
||||
try:
|
||||
return MEMORY_INSTANCE.get(memory_id)
|
||||
except Exception as e:
|
||||
logging.exception("Error in get_memory:")
|
||||
raise HTTPException(status_code=500, detail=str(e))
|
||||
|
||||
|
||||
@app.post("/search", summary="Search memories")
|
||||
def search_memories(search_req: SearchRequest):
|
||||
"""Search for memories based on a query."""
|
||||
try:
|
||||
params = {k: v for k, v in search_req.model_dump().items() if v is not None and k != "query"}
|
||||
return MEMORY_INSTANCE.search(query=search_req.query, **params)
|
||||
except Exception as e:
|
||||
logging.exception("Error in search_memories:")
|
||||
raise HTTPException(status_code=500, detail=str(e))
|
||||
|
||||
|
||||
@app.put("/memories/{memory_id}", summary="Update a memory")
|
||||
def update_memory(memory_id: str, updated_memory: Dict[str, Any]):
|
||||
"""Update an existing memory with new content.
|
||||
|
||||
Args:
|
||||
memory_id (str): ID of the memory to update
|
||||
updated_memory (str): New content to update the memory with
|
||||
|
||||
Returns:
|
||||
dict: Success message indicating the memory was updated
|
||||
"""
|
||||
try:
|
||||
return MEMORY_INSTANCE.update(memory_id=memory_id, data=updated_memory)
|
||||
except Exception as e:
|
||||
logging.exception("Error in update_memory:")
|
||||
raise HTTPException(status_code=500, detail=str(e))
|
||||
|
||||
|
||||
@app.get("/memories/{memory_id}/history", summary="Get memory history")
|
||||
def memory_history(memory_id: str):
|
||||
"""Retrieve memory history."""
|
||||
try:
|
||||
return MEMORY_INSTANCE.history(memory_id=memory_id)
|
||||
except Exception as e:
|
||||
logging.exception("Error in memory_history:")
|
||||
raise HTTPException(status_code=500, detail=str(e))
|
||||
|
||||
|
||||
@app.delete("/memories/{memory_id}", summary="Delete a memory")
|
||||
def delete_memory(memory_id: str):
|
||||
"""Delete a specific memory by ID."""
|
||||
try:
|
||||
MEMORY_INSTANCE.delete(memory_id=memory_id)
|
||||
return {"message": "Memory deleted successfully"}
|
||||
except Exception as e:
|
||||
logging.exception("Error in delete_memory:")
|
||||
raise HTTPException(status_code=500, detail=str(e))
|
||||
|
||||
|
||||
@app.delete("/memories", summary="Delete all memories")
|
||||
def delete_all_memories(
|
||||
user_id: Optional[str] = None,
|
||||
run_id: Optional[str] = None,
|
||||
agent_id: Optional[str] = None,
|
||||
):
|
||||
"""Delete all memories for a given identifier."""
|
||||
if not any([user_id, run_id, agent_id]):
|
||||
raise HTTPException(status_code=400, detail="At least one identifier is required.")
|
||||
try:
|
||||
params = {
|
||||
k: v for k, v in {"user_id": user_id, "run_id": run_id, "agent_id": agent_id}.items() if v is not None
|
||||
}
|
||||
MEMORY_INSTANCE.delete_all(**params)
|
||||
return {"message": "All relevant memories deleted"}
|
||||
except Exception as e:
|
||||
logging.exception("Error in delete_all_memories:")
|
||||
raise HTTPException(status_code=500, detail=str(e))
|
||||
|
||||
|
||||
@app.post("/reset", summary="Reset all memories")
|
||||
def reset_memory():
|
||||
"""Completely reset stored memories."""
|
||||
try:
|
||||
MEMORY_INSTANCE.reset()
|
||||
return {"message": "All memories reset"}
|
||||
except Exception as e:
|
||||
logging.exception("Error in reset_memory:")
|
||||
raise HTTPException(status_code=500, detail=str(e))
|
||||
|
||||
|
||||
@app.get("/", summary="Redirect to the OpenAPI documentation", include_in_schema=False)
|
||||
def home():
|
||||
"""Redirect to the OpenAPI documentation."""
|
||||
return RedirectResponse(url="/docs")
|
||||
6
requirements.txt
Normal file
6
requirements.txt
Normal file
@@ -0,0 +1,6 @@
|
||||
fastapi==0.115.8
|
||||
uvicorn==0.34.0
|
||||
pydantic==2.10.4
|
||||
mem0ai>=0.1.48
|
||||
python-dotenv==1.0.1
|
||||
chromadb>=0.4.22
|
||||
Reference in New Issue
Block a user