Resolving Channel Failures in Containerized AI Agent Deployments

When running local multi-container AI agents (such as OpenClaw instances for Hortense and Gwen), Docker network isolation can cause subtle proxy mismatches if internal container ports collide with host-exposed ports.

Lessons Learned

• ▸Port Mapping Consistency: Always align the internal gateway configuration with the container-exposed port.
• ▸Private Mesh Networking: Utilizing tailscale serve to proxy web interfaces over HTTPS eliminates open firewall exposures.
• ▸Ollama Host Binding: Ensuring the Ollama daemon listens on 0.0.0.0 or host.docker.internal allows containerized agents to query local GPU models cleanly.