mirror of
https://github.com/katanemo/plano.git
synced 2026-04-26 01:06:25 +02:00
33 lines
1 KiB
YAML
33 lines
1 KiB
YAML
|
services:
|
||
|
|
archgw:
|
||
|
|
image: archgw:latest
|
||
|
|
ports:
|
||
|
|
- "10000:10000"
|
||
|
|
- "19901:9901"
|
||
|
|
volumes:
|
||
|
|
- ${ARCH_CONFIG_FILE:-./demos/function_calling/arch_config.yaml}:/config/arch_config.yaml
|
||
|
|
- /etc/ssl/cert.pem:/etc/ssl/cert.pem
|
||
|
|
- ./envoy.template.dev.yaml:/config/envoy.template.yaml
|
||
|
|
- ./target/wasm32-wasi/release/intelligent_prompt_gateway.wasm:/etc/envoy/proxy-wasm-plugins/intelligent_prompt_gateway.wasm
|
||
|
|
depends_on:
|
||
|
|
model_server:
|
||
|
|
condition: service_healthy
|
||
|
|
env_file:
|
||
|
|
- stage.env
|
||
|
|
|
||
|
|
model_server:
|
||
|
|
image: model_server:latest
|
||
|
|
ports:
|
||
|
|
- "18081:80"
|
||
|
|
healthcheck:
|
||
|
|
test: ["CMD", "curl" ,"http://localhost/healthz"]
|
||
|
|
interval: 5s
|
||
|
|
retries: 20
|
||
|
|
volumes:
|
||
|
|
- ~/.cache/huggingface:/root/.cache/huggingface
|
||
|
|
environment:
|
||
|
|
- OLLAMA_ENDPOINT=${OLLAMA_ENDPOINT:-host.docker.internal}
|
||
|
|
- OLLAMA_MODEL=Arch-Function-Calling-3B-Q4_K_M
|
||
|
|
- MODE=${MODE:-cloud}
|
||
|
|
- FC_URL=${FC_URL:-https://arch-fc-free-trial-4mzywewe.uc.gateway.dev/v1}
|