​
Docker Image

The Podium Docker image includes the coordinator, gateway, and Chronicle: 
cd podium
./build_image.sh staging  # or production
The build script:
  1. Checks for Chronicle source at parent directory level
  2. Copies Chronicle into build context
  3. Builds multi-stage Docker image (Rust compilation + runtime)
  4. Optionally deploys to ECS with --deploy

​
Running with Docker

docker run -p 8090:8090 -p 9090:9090 \
  -e CHRONICLE_ENABLE=true \
  -e CHRONICLE_S3_BUCKET=my-chronicle-cas \
  -e AWS_REGION=us-east-1 \
  -e S3_BUCKET=my-deployment-registry \
  -e VALKEY_URL=redis://valkey:6379 \
  -e ADMIN_API_KEY=pod_admin_... \
  -e SECRETS_KEY=... \
  --privileged \  # Required for FUSE
  podium:latest
--privileged or --cap-add SYS_ADMIN --device /dev/fuse is required for Chronicle FUSE mounts inside containers.

​
Production Architecture

                    ALB
                     │
              ┌──────┴──────┐
              ▼              ▼
         Gateway-1      Gateway-2
              │              │
              └──────┬───────┘
                     │
                  Valkey
                     │
         ┌───────────┼───────────┐
         ▼           ▼           ▼
    Coordinator  Coordinator  Coordinator
    (ECS Task)   (ECS Task)   (ECS Task)

​
AWS Components

ComponentServiceNotes
GatewayECS FargateStateless, horizontally scalable
CoordinatorECS EC2Needs FUSE (privileged), stateful
Service DiscoveryElastiCache (Valkey)Coordinator registration
Deployment RegistryS3Agent bundles
State ReplicationS3 (Litestream)Agent databases
CAS StorageS3Chronicle content-addressable storage
ObservabilityCloudWatch (OTLP)Traces + metrics

​
Terraform

Podium includes Terraform modules for AWS deployment covering ECS services, ALB configuration, Valkey cluster, S3 buckets, IAM roles, and security groups.

​
Scaling

  • Gateway: Stateless — scale horizontally behind ALB
  • Coordinator: Stateful — scale by adding instances; Valkey distributes new agents to least-loaded coordinator
  • Agents: Scale within a coordinator up to resource limits; scale out by adding coordinators