Skip to content

Commit 7d53da4

Browse files
danielaskdddanielaskdd
committed
Add full Docker Compose deployment sample with GPU support
1 parent 7aacabb commit 7d53da4

File tree

1 file changed

+235
-0
lines changed

1 file changed

+235
-0
lines changed

‎docker-compose-full.yml‎

Lines changed: 235 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,235 @@
1+
# Full Docker Compose Deployment Sample Generated by Setup Wizard: `make env` and `make storage`
2+
# This Sample File requires NVIDIA GPU for Milvus and VLLM services.
3+
# You can customize your setup using the Setup Wizard; for detailed instructions, please refer to docs/InteractiveSetup.md
4+
services:
5+
lightrag:
6+
image: ghcr.io/hkuds/lightrag:latest
7+
build:
8+
context: .
9+
dockerfile: Dockerfile
10+
tags:
11+
- ghcr.io/hkuds/lightrag:latest
12+
ports:
13+
- "${HOST:-0.0.0.0}:${PORT:-9621}:9621"
14+
volumes:
15+
- ./data/rag_storage:/app/data/rag_storage
16+
- ./data/inputs:/app/data/inputs
17+
- ./config.ini:/app/config.ini
18+
- ./.env:/app/.env
19+
restart: unless-stopped
20+
extra_hosts:
21+
- "host.docker.internal:host-gateway"
22+
environment:
23+
MILVUS_URI: "http://milvus:19530"
24+
NEO4J_URI: "neo4j://neo4j:7687"
25+
POSTGRES_HOST: "postgres"
26+
POSTGRES_PORT: "5432"
27+
EMBEDDING_BINDING_HOST: "http://vllm-embed:8001/v1"
28+
RERANK_BINDING_HOST: "http://vllm-rerank:8000/rerank"
29+
WORKING_DIR: "/app/data/rag_storage"
30+
INPUT_DIR: "/app/data/inputs"
31+
MEMGRAPH_URI: "bolt://host.docker.internal:7687"
32+
HOST: "0.0.0.0"
33+
PORT: "9621"
34+
depends_on:
35+
vllm-embed:
36+
condition: service_healthy
37+
vllm-rerank:
38+
condition: service_healthy
39+
postgres:
40+
condition: service_healthy
41+
neo4j:
42+
condition: service_healthy
43+
milvus:
44+
condition: service_healthy
45+
46+
milvus:
47+
image: milvusdb/milvus:v2.6.11-gpu
48+
command: ["milvus", "run", "standalone"]
49+
security_opt:
50+
- seccomp:unconfined
51+
environment:
52+
ETCD_ENDPOINTS: milvus-etcd:2379
53+
MINIO_ADDRESS: milvus-minio:9000
54+
MINIO_ACCESS_KEY_ID: "${MINIO_ACCESS_KEY_ID:?missing}"
55+
MINIO_SECRET_ACCESS_KEY: "${MINIO_SECRET_ACCESS_KEY:?missing}"
56+
# ports:
57+
# - "19530:19530"
58+
# - "9091:9091"
59+
volumes:
60+
- milvus_data:/var/lib/milvus
61+
deploy:
62+
resources:
63+
reservations:
64+
devices:
65+
- driver: nvidia
66+
capabilities: ["gpu"]
67+
healthcheck:
68+
test:
69+
- CMD-SHELL
70+
- 'PORT_HEX="$(printf ''%04X'' 19530)"; cat /proc/net/tcp /proc/net/tcp6 2>/dev/null | grep -q ":$${PORT_HEX} "'
71+
interval: 10s
72+
timeout: 3s
73+
retries: 120
74+
start_period: 10s
75+
depends_on:
76+
milvus-etcd:
77+
condition: service_healthy
78+
milvus-minio:
79+
condition: service_healthy
80+
restart: unless-stopped
81+
82+
milvus-etcd:
83+
image: quay.io/coreos/etcd:v3.5.25
84+
environment:
85+
ETCD_AUTO_COMPACTION_MODE: revision
86+
ETCD_AUTO_COMPACTION_RETENTION: "1000"
87+
ETCD_QUOTA_BACKEND_BYTES: "4294967296"
88+
ETCD_SNAPSHOT_COUNT: "50000"
89+
volumes:
90+
- milvus-etcd_data:/etcd
91+
command: >
92+
etcd
93+
-advertise-client-urls=http://0.0.0.0:2379
94+
-listen-client-urls=http://0.0.0.0:2379
95+
-data-dir /etcd
96+
healthcheck:
97+
test: ["CMD", "etcdctl", "endpoint", "health"]
98+
interval: 20s
99+
timeout: 20s
100+
retries: 3
101+
restart: unless-stopped
102+
103+
milvus-minio:
104+
image: minio/minio:RELEASE.2025-09-07T16-13-09Z
105+
environment:
106+
MINIO_ROOT_USER: "${MINIO_ACCESS_KEY_ID:?missing}"
107+
MINIO_ROOT_PASSWORD: "${MINIO_SECRET_ACCESS_KEY:?missing}"
108+
volumes:
109+
- milvus-minio_data:/minio_data
110+
command: minio server /minio_data --console-address ":9001"
111+
healthcheck:
112+
test: ["CMD", "curl", "-f", "http://localhost:9000/minio/health/live"]
113+
interval: 30s
114+
timeout: 20s
115+
retries: 3
116+
restart: unless-stopped
117+
118+
neo4j:
119+
image: neo4j:5-community
120+
# ports:
121+
# - "7474:7474"
122+
# - "${NEO4J_BOLT_PORT:-7687}:7687"
123+
volumes:
124+
- neo4j_data:/data
125+
healthcheck:
126+
test:
127+
- CMD-SHELL
128+
- 'PORT_HEX="$(printf ''%04X'' 7687)"; cat /proc/net/tcp /proc/net/tcp6 2>/dev/null | grep -q ":$${PORT_HEX} "'
129+
interval: 10s
130+
timeout: 3s
131+
retries: 120
132+
start_period: 10s
133+
restart: unless-stopped
134+
environment:
135+
NEO4J_AUTH: ${NEO4J_USERNAME:?missing}/${NEO4J_PASSWORD:?missing}
136+
NEO4J_dbms_default__database: "neo4j"
137+
138+
postgres:
139+
image: gzdaniel/postgres-for-rag:16.6
140+
command: ["sh", "-c", "service postgresql start && sleep infinity"]
141+
# ports:
142+
# - "5432:5432"
143+
volumes:
144+
- postgres_data:/var/lib/postgresql
145+
healthcheck:
146+
test:
147+
- CMD-SHELL
148+
- 'PORT_HEX="$(printf ''%04X'' 5432)"; cat /proc/net/tcp /proc/net/tcp6 2>/dev/null | grep -q ":$${PORT_HEX} "'
149+
interval: 5s
150+
timeout: 3s
151+
retries: 120
152+
start_period: 10s
153+
restart: unless-stopped
154+
environment:
155+
# The custom image featuring pre-installed AGE and pgvector extensions, including a pre-configured administrator account
156+
POSTGRES_USER: "rag"
157+
POSTGRES_PASSWORD: "rag"
158+
POSTGRES_DB: "rag"
159+
160+
vllm-embed:
161+
image: vllm/vllm-openai:latest
162+
runtime: nvidia
163+
command: >
164+
--model ${VLLM_EMBED_MODEL:-BAAI/bge-m3}
165+
--port ${VLLM_EMBED_PORT:-8001}
166+
--dtype float16
167+
--api-key ${VLLM_EMBED_API_KEY}
168+
${VLLM_EMBED_EXTRA_ARGS:-}
169+
environment:
170+
NVIDIA_VISIBLE_DEVICES: ${NVIDIA_VISIBLE_DEVICES:-all}
171+
NVIDIA_DRIVER_CAPABILITIES: ${NVIDIA_DRIVER_CAPABILITIES:-compute,utility}
172+
ports:
173+
- "${VLLM_EMBED_PORT:-8001}:${VLLM_EMBED_PORT:-8001}"
174+
volumes:
175+
- vllm_embed_cache:/root/.cache/huggingface
176+
ipc: host
177+
healthcheck:
178+
test:
179+
- CMD-SHELL
180+
- 'PORT_HEX="$(printf ''%04X'' ${VLLM_EMBED_PORT:-8001})"; cat /proc/net/tcp /proc/net/tcp6 2>/dev/null | grep -q ":$${PORT_HEX} "'
181+
interval: 5s
182+
timeout: 3s
183+
retries: 120
184+
start_period: 10s
185+
deploy:
186+
resources:
187+
reservations:
188+
devices:
189+
- driver: nvidia
190+
count: all
191+
capabilities: [gpu]
192+
restart: unless-stopped
193+
194+
vllm-rerank:
195+
image: vllm/vllm-openai:latest
196+
runtime: nvidia
197+
command: >
198+
--model ${VLLM_RERANK_MODEL:-BAAI/bge-reranker-v2-m3}
199+
--port ${VLLM_RERANK_PORT:-8000}
200+
--dtype float16
201+
--api-key ${VLLM_RERANK_API_KEY}
202+
${VLLM_RERANK_EXTRA_ARGS:-}
203+
environment:
204+
NVIDIA_VISIBLE_DEVICES: ${NVIDIA_VISIBLE_DEVICES:-all}
205+
NVIDIA_DRIVER_CAPABILITIES: ${NVIDIA_DRIVER_CAPABILITIES:-compute,utility}
206+
ports:
207+
- "${VLLM_RERANK_PORT:-8000}:${VLLM_RERANK_PORT:-8000}"
208+
volumes:
209+
- vllm_rerank_cache:/root/.cache/huggingface
210+
ipc: host
211+
healthcheck:
212+
test:
213+
- CMD-SHELL
214+
- 'PORT_HEX="$(printf ''%04X'' ${VLLM_RERANK_PORT:-8000})"; cat /proc/net/tcp /proc/net/tcp6 2>/dev/null | grep -q ":$${PORT_HEX} "'
215+
interval: 5s
216+
timeout: 3s
217+
retries: 120
218+
start_period: 10s
219+
deploy:
220+
resources:
221+
reservations:
222+
devices:
223+
- driver: nvidia
224+
count: all
225+
capabilities: [gpu]
226+
restart: unless-stopped
227+
228+
volumes:
229+
milvus_data:
230+
milvus-etcd_data:
231+
milvus-minio_data:
232+
neo4j_data:
233+
postgres_data:
234+
vllm_embed_cache:
235+
vllm_rerank_cache:

0 commit comments

Comments
 (0)