Please select the area the issue is related to
CLI
Please select the aspect the issue is related to
Aspect/AI (AI/LLM integration, MCP, AI readiness)
Description
The command-line deployment tool incorrectly removes a required authentication reference when deploying AI proxy configurations, causing deployments to either fail during validation or succeed but later return authentication errors at runtime. This issue is limited to the CLI workflow and does not affect the AI Workspace UI, which correctly includes the required authentication reference. Because automated tests expected this behavior and the build process performs only basic validation, the problem was not detected earlier. As a result, CI/CD deployments cannot reliably deploy AI proxies that use the standard API key authentication configuration.
Steps to Reproduce
Create an LlmProvider with an api-key-auth policy, create an LlmProxy referencing it with provider.auth.{type: api-key, header, value}, run ap ai-workspace apply --project-id , then GET /api/v0.9/llm-proxies/ — the stored provider.auth contains type and header but no value. Deploying that proxy to a gateway then fails validation.
Severity Level of the Issue
Severity/Blocker (Core functionality is broken and consumer is blocked. Need immediate attention)
Environment Details (with versions)
No response
Please select the area the issue is related to
CLI
Please select the aspect the issue is related to
Aspect/AI (AI/LLM integration, MCP, AI readiness)
Description
The command-line deployment tool incorrectly removes a required authentication reference when deploying AI proxy configurations, causing deployments to either fail during validation or succeed but later return authentication errors at runtime. This issue is limited to the CLI workflow and does not affect the AI Workspace UI, which correctly includes the required authentication reference. Because automated tests expected this behavior and the build process performs only basic validation, the problem was not detected earlier. As a result, CI/CD deployments cannot reliably deploy AI proxies that use the standard API key authentication configuration.
Steps to Reproduce
Create an LlmProvider with an api-key-auth policy, create an LlmProxy referencing it with provider.auth.{type: api-key, header, value}, run ap ai-workspace apply --project-id , then GET /api/v0.9/llm-proxies/ — the stored provider.auth contains type and header but no value. Deploying that proxy to a gateway then fails validation.
Severity Level of the Issue
Severity/Blocker (Core functionality is broken and consumer is blocked. Need immediate attention)
Environment Details (with versions)
No response