Phala Network

Phala provides Trusted Execution Environments (TEEs) for running Delibera's coordinator and worker agents in production. Code and data inside a TEE are shielded from the host operator.

How It Works

Phala uses Intel SGX enclaves exposed through CVM (Confidential Virtual Machine) containers. Delibera agents run as Docker containers inside these CVMs, ensuring that:

• Agent reasoning and vote content remain private during execution
• The host infrastructure cannot inspect memory or tamper with computation
• Attestation proofs verify the code running matches the expected image

Container Requirements

Phala CVM containers have strict requirements:

services:
  worker:
    image: leomanza/delibera-worker:latest
    platform: linux/amd64
    restart: always
    volumes:
      - /var/run/dstack.sock:/var/run/dstack.sock

Endpoint Provisioning

When deploying a new CVM container, Phala assigns a public HTTPS endpoint. This process takes 3-10+ minutes. The deployment script polls for readiness:

// watchForEndpoint() polls 40 times at 15s intervals (up to 10 min)
const endpoint = await watchForEndpoint(deploymentId);

Local Development

In LOCAL_MODE, TEE and contract interactions are skipped. The coordinator and workers run as plain Node.js processes:

Coordinator: port 3000
Workers:     ports 3001-300N (WORKER_COUNT=N ./run-dev.sh)