Resolving Channel Failures in Containerized AI Agent Deployments

# 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.