A practical configuration guide for connecting OpenCode (an AI coding assistant) to Azure-hosted models, including Azure OpenAI GPT models, Azure Cognitive Services, and Azure AI Foundry base models like Kimi. The post covers the specific differences in provider configuration: which domain to use (openai.azure.com vs cognitiveservices.azure.com), when to include the /v1 route suffix, and why resourceName must be omitted for Foundry models. It also notes that Codex models require the azure-cognitive-services provider rather than the standard azure or azure-foundry-base-models providers.
Sort: