This commit addresses GitHub issue #828 by fixing URL construction for Azure OpenAI's Responses API
and preventing malformed URLs from client configuration issues.
## Problems Fixed:
1. Host normalization: Prevents malformed URLs when client config includes paths/queries
2. API paradigm support: Properly handles Azure's next-gen v1 Responses API
3. API version consistency: Centralizes version management with env overrides
## Key Changes:
- Normalize Azure host URLs to origin only (strip path/query)
- Prefer server environment variables over client-provided hosts
- Add special handling for Responses API (/openai/v1/responses)
- Support both traditional (deployment-based) and v1 API paradigms
- Add configurable API versions via environment variables
- Include debug logging for API paradigm selection
## New Environment Variables:
- AZURE_API_V1: Enable next-gen v1 API explicitly
- AZURE_RESPONSES_API_VERSION: Control Responses API version
- AZURE_CHAT_API_VERSION: Control Chat Completions API version
- AZURE_DEPLOYMENTS_API_VERSION: Control deployments listing API version
## Testing:
Validated with Azure OpenAI endpoint showing:
- List Deployments: ✅ Works
- Chat Completions: ✅ Works (with correct params for GPT-5)
- Responses API (v1): ✅ Works with /openai/v1/responses?api-version=preview
- Responses API (traditional): ❌ 404 (Azure doesn't support this pattern)
The fix defaults to using Azure's recommended next-gen v1 API for Responses
while maintaining backward compatibility for existing deployments.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
Enrico Ros