Merge pull request #137 from NillionNetwork/feat/nilai_nilcc_integration

feat: nilAI <> nilCC integration

Author: jcabrero

.github/workflows/cicd.yml

@@ -119,13 +119,12 @@ jobs:
          # Copy secret into .env replacing the existing HF_TOKEN
          sed -i 's/HF_TOKEN=.*/HF_TOKEN=${{ secrets.HF_TOKEN }}/' .env
 
+      - name: Compose docker-compose.yml
+        run: bash ./scripts/docker-composer.sh --dev -f docker/compose/docker-compose.llama-1b-gpu.ci.yml -o development-compose.yml
+
       - name: Start Services
         run: |
-          docker-compose -f docker-compose.yml \
-          -f docker-compose.dev.yml \
-          -f docker-compose.prod.yml \
-          -f docker/compose/docker-compose.llama-1b-gpu.ci.yml \
-          up -d
+          docker-compose -f development-compose.yml up -d
           docker ps -a
 
       - name: Wait for services to be healthy
@@ -149,11 +148,7 @@ jobs:
 
       - name: Stop Services
         run: |
-          docker-compose -f docker-compose.yml \
-          -f docker-compose.dev.yml \
-          -f docker-compose.prod.yml \
-          -f docker/compose/docker-compose.llama-1b-gpu.ci.yml \
-          down -v
+          docker-compose -f development-compose.yml down -v
 
   stop-runner:
     name: Stop self-hosted EC2 runner

.gitignore

@@ -175,3 +175,6 @@ prometheus/data/*
 
 private_key.key
 private_key.key.lock
+
+development-compose.yml
+production-compose.yml

README.md

@@ -87,7 +87,83 @@ docker compose -f docker-compose.yml \
 up -d
 ```
 
-### 2. Manual Component Deployment
+### 2. Using the Docker Compose Helper Script
+
+For easier management of multiple compose files and image substitutions, use the `docker-composer.sh` script:
+
+#### Basic Usage
+
+```shell
+# Generate a composed configuration for development
+./scripts/docker-composer.sh --dev -o dev-compose.yml
+
+# Generate a composed configuration for production
+./scripts/docker-composer.sh --prod -o prod-compose.yml
+
+# Include specific model configurations
+./scripts/docker-composer.sh --prod \
+  -f docker-compose.llama-3b-gpu.yml \
+  -f docker-compose.llama-8b-gpu.yml \
+  -o production-compose.yml
+```
+
+#### Image Substitution
+
+Replace default images with custom ones (useful for production deployments with specific image versions):
+
+```shell
+# Production example with custom ECR images
+./scripts/docker-composer.sh --prod \
+  -f docker-compose.llama-3b-gpu.yml \
+  --image 'nillion/nilai-api:latest=public.ecr.aws/k5d9x2g2/nilai-api:v0.1.0-rc1' \
+  --image 'nillion/nilai-vllm:latest=public.ecr.aws/k5d9x2g2/nilai-vllm:v0.1.0-rc1' \
+  --image 'nillion/nilai-attestation:latest=public.ecr.aws/k5d9x2g2/nilai-attestation:v0.1.0-rc1' \
+  -o production-compose.yml
+
+# Then deploy with the generated file
+docker compose -f production-compose.yml up -d
+```
+
+#### Script Options
+
+- `--dev`: Include development-specific configurations
+- `--prod`: Include production-specific configurations
+- `-f, --file <filename>`: Include additional compose files from `docker/compose/` directory
+- `-o, --output <filename>`: Specify output filename (default: `output.yml`)
+- `--image <old=new>`: Substitute Docker images (can be used multiple times)
+- `-h, --help`: Show help message
+
+#### Production Deployment Example
+
+For a complete production setup with custom images:
+
+```shell
+# 1. Generate the production compose file
+./scripts/docker-composer.sh --prod \
+  -f docker/compose/docker-compose.llama-3b-gpu.yml \
+  -f docker/compose/docker-compose.llama-8b-gpu.yml \
+  --image 'nillion/nilai-api:latest=public.ecr.aws/k5d9x2g2/nilai-api:v0.2.0-alpha-0' \
+  --image 'nillion/nilai-vllm:latest=public.ecr.aws/k5d9x2g2/nilai-vllm:v0.2.0-alpha-0' \
+  --image 'nillion/nilai-attestation:latest=public.ecr.aws/k5d9x2g2/nilai-attestation:v0.2.0-alpha-0' \
+  -o production-compose.yml
+
+# Or:
+./scripts/docker-composer.sh --prod \
+  -f docker/compose/docker-compose.llama-70b-gpu.yml \
+  --image 'nillion/nilai-api:latest=jcabrero/nillion-nilai-api:latest' \
+  --image 'nillion/nilai-vllm:latest=public.ecr.aws/k5d9x2g2/nilai-vllm:v0.1.0-rc1' \
+  --image 'nillion/nilai-attestation:latest=public.ecr.aws/k5d9x2g2/nilai-attestation:v0.1.0-rc1' \
+  -o production-compose.yml
+
+
+# 2. Deploy using the generated file
+docker compose -f production-compose.yml up -d
+
+# 3. Monitor logs
+docker compose -f production-compose.yml logs -f
+```
+
+### 3. Manual Component Deployment
 
 #### Components
 

caddy/Caddyfile

@@ -16,6 +16,7 @@
 		uri strip_prefix /nuc
 		reverse_proxy nuc-api:8080
 	}
+
 	handle_path /prod/nuc/* {
 		uri strip_prefix /prod/nuc
 		reverse_proxy prod-nilai-nuc-api:8080

caddy/Caddyfile.http

@@ -0,0 +1,37 @@
+# Disable automatic HTTPS and use HTTP only
+{
+        auto_https off
+        default_bind 0.0.0.0
+}
+
+# Use :80 explicitly to force HTTP-only behavior
+:80 {
+        handle_path /grafana/* {
+                uri strip_prefix /grafana
+                reverse_proxy grafana:3000
+        }
+
+        handle_path /nuc/* {
+                uri strip_prefix /nuc
+                reverse_proxy nuc-api:8080
+        }
+
+        handle_path /prod/nuc/* {
+                uri strip_prefix /prod/nuc
+                reverse_proxy prod-nilai-nuc-api:8080
+        }
+
+        handle_path /prod/grafana/* {
+                uri strip_prefix /prod/grafana
+                reverse_proxy prod-grafana:3000
+        }
+
+        handle_path /prod/* {
+                uri strip_prefix /prod/base
+                reverse_proxy prod-nilai-api:8080
+        }
+
+        handle {
+                reverse_proxy api:8080
+        }
+}

docker-compose.dev.yml

@@ -1,51 +1,99 @@
 services:
+  caddy:
+    env_file:
+      - .env
+    ports:
+      - "80:80"
+      - "443:443"
+      - "443:443/udp"
+    volumes:
+      - ./caddy/Caddyfile:/etc/caddy/Caddyfile
   api:
     platform: linux/amd64 # for macOS to force running on Rosetta 2
     ports:
       - "8080:8080"
+    env_file:
+      - .env
     volumes:
       - ./nilai-api/:/app/nilai-api/
       - ./packages/:/app/packages/
       - ./nilai-auth/nuc-helpers/:/app/nilai-auth/nuc-helpers/
-    networks:
-      - nilauth
   nuc-api:
     platform: linux/amd64 # for macOS to force running on Rosetta 2
     ports:
       - "8088:8080"
+    env_file:
+      - .env
     volumes:
       - ./nilai-api/:/app/nilai-api/
       - ./packages/:/app/packages/
       - ./nilai-auth/nuc-helpers/:/app/nilai-auth/nuc-helpers/
-    networks:
-      - nilauth
   attestation:
     ports:
       - "8081:8080"
+    env_file:
+      - .env
     volumes:
       - ./nilai-attestation/:/app/nilai-attestation/
       - ./packages/:/app/packages/
   redis:
     ports:
       - "6379:6379"
   postgres:
+    image: postgres:16
+    container_name: postgres
+    restart: always
     ports:
       - "5432:5432"
+    env_file:
+      - .env
+    volumes:
+      - postgres_data:/var/lib/postgresql/data
+    healthcheck:
+      test: ["CMD", "sh", "-c", "pg_isready -U ${POSTGRES_USER} -d ${POSTGRES_DB} -h localhost"]
+      interval: 30s
+      retries: 5
+      start_period: 10s
+      timeout: 10s
+
   nuc-postgres:
+    image: postgres:16
+    container_name: nuc-postgres
+    restart: always
+    environment:
+      - POSTGRES_HOST=nuc-postgres
     ports:
       - "5433:5432"
+    env_file:
+      - .env
+    volumes:
+      - nuc_postgres_data:/var/lib/postgresql/data
+    healthcheck:
+      test: ["CMD", "sh", "-c", "pg_isready -U ${POSTGRES_USER} -d ${POSTGRES_DB} -h localhost"]
+      interval: 30s
+      retries: 5
+      start_period: 10s
+      timeout: 10s
   grafana:
+    env_file:
+      - .env
     ports:
       - "3000:3000"
-
+    volumes:
+      - ${PWD}/grafana/runtime-data:/var/lib/grafana
+      - ${PWD}/grafana/datasources/datasource.yml:/etc/grafana/provisioning/datasources/datasource.yml
+      - ${PWD}/grafana/dashboards/filesystem.yml:/etc/grafana/provisioning/dashboards/filesystem.yml
+      - ${PWD}/grafana/config/grafana.ini:/etc/grafana/grafana.ini
+  prometheus:
+    volumes:
+      - ${PWD}/prometheus/config/prometheus.yml:/etc/prometheus/prometheus.yml
+      - ${PWD}/prometheus/data:/prometheus/data
   nilauth-postgres:
     image: postgres:17
     environment:
       POSTGRES_PASSWORD: postgres
     ports:
       - "30432:5432"
-    networks:
-      - nilauth
 
   nilchain:
     image: ghcr.io/nillionnetwork/nilchain-devnet:v0.1.0
@@ -55,8 +103,6 @@ services:
       - "26648:26648" # JSON RPC
       - "26649:26649" # gRPC
       - "26650:26650" # REST
-    networks:
-      - nilauth
 
   nilauth:
     image: public.ecr.aws/k5d9x2g2/nilauth:ff0d9198d1b8819527bc036a58f875c4046b6f21
@@ -69,17 +115,15 @@ services:
     ports:
       - "30921:30921" # main server
       - "30922:30022" # metrics server
-    networks:
-      - nilauth
 
   token-price-api:
     image: caddy:2
     ports:
       - "30923:80"
     command: |
       caddy respond --listen :80 --body '{"nillion":{"usd":1}}' --header "Content-Type: application/json"
-    networks:
-      - nilauth
 
-networks:
-  nilauth:
+
+volumes:
+  postgres_data:
+  nuc_postgres_data:

docker-compose.prod.yml

@@ -1,4 +1,34 @@
 services:
+  prometheus:
+    volumes:
+      - ./files/prometheus/prometheus.yml:/etc/prometheus/prometheus.yml
+      - prometheus_data:/prometheus/data
+  grafana:
+      env_file:
+        - .env.mainnet
+      volumes:
+      - ./files/grafana/datasource.yml:/etc/grafana/provisioning/datasources/datasource.yml
+      - ./files/grafana/filesystem.yml:/etc/grafana/provisioning/dashboards/filesystem.yml
+      - ./files/grafana/grafana.ini:/etc/grafana/grafana.ini
+      - ./files/grafana/nuc-query-data.json:/var/lib/grafana/dashboards/nuc-query-data.json
+      - ./files/grafana/query-data.json:/var/lib/grafana/dashboards/query-data.json
+  api:
+    env_file:
+      - .env.mainnet
+    volumes:
+      - ./files/nilai-api/config.yaml:/app/nilai-api/src/nilai_api/config/config.yaml
+  nuc-api:
+    env_file:
+      - .env.mainnet
+    environment:
+      - POSTGRES_DB="nuc-nilai-mainnet2"
+    volumes:
+      - ./files/nilai-api/config.yaml:/app/nilai-api/src/nilai_api/config/config.yaml
+  caddy:
+    env_file:
+      - .env.mainnet
+    volumes:
+      - ./files/caddy/caddyfile:/etc/caddy/Caddyfile
   attestation:
     deploy:
       resources: