Overview

Full name
docker.io/jokerwho/bge-large-api
Registry
docker.io
Namespace
jokerwho
Repository
bge-large-api
Cached tags
1
Last synced
07/06/2026, 07:21 AM

Introduction

BGE-Large 中文向量嵌入 API 服务镜像,提供 OpenAI 兼容 /embeddings 接口,适合 RAG 检索增强内网部署。

Details

jokerwho/bge-large-api 加载 BAAI/bge-large-zh 模型暴露 HTTP embedding API。定位为向量推理 HTTP 服务,CPU 可运行但 GPU 更快;K8s Deployment 供 RAG 流水线调用,模型 cache 用 PVC。

快速启动

docker run -d --name bge-api \
  -p 8000:8000 \
  docker.io/jokerwho/bge-large-api:latest

推荐实践

docker run -d --name bge-api \
  --restart unless-stopped \
  -p 8000:8000 \
  -v /data/bge-cache:/root/.cache \
  docker.io/jokerwho/bge-large-api:1.0.0

核心参数说明

  • -p 8000:8000 — Embedding API 端口
  • -v .../bge-cache:... — HuggingFace 模型缓存
  • --restart unless-stopped — API 持续在线
  • 固定 tag — 模型版本 pin

Kubernetes

apiVersion: apps/v1
kind: Deployment
metadata:
  name: bge-large-api
spec:
  replicas: 2
  selector:
    matchLabels:
      app: bge-large-api
  template:
    metadata:
      labels:
        app: bge-large-api
    spec:
      containers:
        - name: api
          image: docker.io/jokerwho/bge-large-api:1.0.0
          ports:
            - containerPort: 8000
          volumeMounts:
            - name: cache
              mountPath: /root/.cache
      volumes:
        - name: cache
          persistentVolumeClaim:
            claimName: bge-cache
---
apiVersion: v1
kind: Service
metadata:
  name: bge-large-api
spec:
  selector:
    app: bge-large-api
  ports:
    - port: 8000
      targetPort: 8000

RAG 应用经 ClusterIP 调用。

Latest tags (20)

View all 1 tags →
TagArchPushed AtSizeDigestSyncWishAction
docker.io/jokerwho/bge-large-api
:latest
linux/amd64
2023/12/05 20:454.1 GBsha256:352c4235d2a5