hermes - 💡(How to fix) Fix [Feature Request]: Configurable gateway environment/proxy bypass for platform connectors [1 participants]

Summary

Please consider adding a supported way to configure gateway environment variables or per-platform proxy behavior, especially for persistent websocket connectors.

Some local/enterprise deployments run with system-level proxies, VPNs, or network filters. For websocket-based platform connectors, this can cause specific platform endpoints to be routed through the wrong proxy and disconnect repeatedly. Today, the workaround may require patching Hermes launchd/service generation so the gateway process starts with a custom no_proxy/NO_PROXY value.

Proposed capability

Expose gateway process environment and/or platform proxy behavior in config, for example:

gateway:
  env:
    no_proxy: msg-frontier.feishu.cn,localhost,127.0.0.1

Or platform-scoped configuration:

platforms:
  feishu:
    websocket:
      proxy: direct
      no_proxy:
        - msg-frontier.feishu.cn

The exact config format can vary. The key request is: deployments should be able to control gateway proxy behavior without editing generated service files or Hermes source.

Use cases

• Feishu/Lark websocket connections that must bypass a local system proxy
• Corporate VPN environments where only some endpoints should use the proxy
• Self-hosted deployments with separate proxy policies for model APIs vs messaging platforms
• Launchd/systemd users who need stable env injection across upgrades

Expected outcome

Gateway service generation should preserve configured environment/proxy settings across upgrades and restarts. Platform connectors should be able to opt into direct connections where needed.

Why this matters

Messaging gateways are long-running and sensitive to proxy behavior. Treating proxy bypass as configuration would prevent downstream deployments from patching service templates after every upgrade.