[BUG] BYOK custom model unable to reach Azure AI Services endpoint despite curl working
Bug Description
Custom BYOK model using Azure AI Services endpoint (services.ai.azure.com) shows
"Unable to reach " error in droid, but the same endpoint works perfectly
via curl with valid HTTP 200 response.
Steps to Reproduce
-
Add a custom model to ~/.factory/settings.json with:
- baseUrl: https://.services.ai.azure.com/openai/v1
- model:
- provider: openai, azure, or generic-chat-completion-api
- Backend model: Kimi K2.6 (via Azure)
-
Switch to the model in droid → error:
"Unable to reach .services.ai.azure.com"
-
Same curl command to the same endpoint → HTTP 200, valid response
Evidence
- DNS resolves correctly
- TLS handshake succeeds (TLSv1.3)
- API returns valid chat completion with model "accounts/fireworks/models/kimi-k2p6"
- Other Azure endpoints on droid-orchestrator-resource.services.ai.azure.com work fine
- Tried all provider types: openai, azure, generic-chat-completion-api — same error
- Droid version: 0.122.0
- Platform: macOS
Expected vs Actual
- Expected: Model connects and works (as it does via curl)
- Actual: droid reports unreachable despite full connectivity
Root Cause (suspected)
Droid's pre-flight connectivity/health check fails for this specific Azure AI
Services resource, possibly due to TLS inspection, certificate chain differences,
or a connectivity check implementation that differs from standard curl behavior.
[BUG] BYOK custom model unable to reach Azure AI Services endpoint despite curl working
Bug Description
Custom BYOK model using Azure AI Services endpoint (services.ai.azure.com) shows
"Unable to reach " error in droid, but the same endpoint works perfectly
via curl with valid HTTP 200 response.
Steps to Reproduce
Add a custom model to ~/.factory/settings.json with:
Switch to the model in droid → error:
"Unable to reach .services.ai.azure.com"
Same curl command to the same endpoint → HTTP 200, valid response
Evidence
Expected vs Actual
Root Cause (suspected)
Droid's pre-flight connectivity/health check fails for this specific Azure AI
Services resource, possibly due to TLS inspection, certificate chain differences,
or a connectivity check implementation that differs from standard curl behavior.