We are introducing a unified forward proxy configuration for AI plugins to support network-isolated environments. In such environments, outbound connections to external services (LLM providers, embeddings, guardrails, etc.) are restricted and must go through a forward proxy. This feature ensures AI plugins can operate correctly under these constraints without impacting existing capabilities.
To support this:
- A
proxy_config field will be added at the plugin level to define proxy settings.
When configured, all outbound requests from the plugin (including LLM calls, embeddings, and other external integrations) will be routed through the specified proxy.
- The configuration supports both HTTP and HTTPS proxies, along with optional authentication and TLS verification settings.
DoD:
- TBD: a reference page with an overview of how forward proxy works
ERD
PR
#5264
We are introducing a unified forward proxy configuration for AI plugins to support network-isolated environments. In such environments, outbound connections to external services (LLM providers, embeddings, guardrails, etc.) are restricted and must go through a forward proxy. This feature ensures AI plugins can operate correctly under these constraints without impacting existing capabilities.
To support this:
proxy_configfield will be added at the plugin level to define proxy settings.When configured, all outbound requests from the plugin (including LLM calls, embeddings, and other external integrations) will be routed through the specified proxy.
DoD:
ERD
PR
#5264