README
¶
Azure OpenAI Examples
Chronos talks to Azure-hosted OpenAI models through the same model.Provider
interface as every other backend, so agents, graphs, tools, and teams work
unchanged. Azure differs from vanilla OpenAI in one way: you address a
deployment on a resource endpoint with an api-version, not a plain
model id.
Environment variables
All Azure examples read the same four variables:
| Variable | Example | Purpose |
|---|---|---|
AZURE_OPENAI_API_KEY |
abc123... |
Resource API key |
AZURE_OPENAI_ENDPOINT |
https://my-resource.openai.azure.com |
Resource endpoint |
AZURE_OPENAI_DEPLOYMENT |
gpt-4o-mini |
Chat model deployment name |
AZURE_OPENAI_API_VERSION |
2024-10-21 |
API version |
export AZURE_OPENAI_API_KEY=<your-azure-api-key>
export AZURE_OPENAI_ENDPOINT=https://<your-resource>.openai.azure.com
export AZURE_OPENAI_DEPLOYMENT=<your-deployment-name>
export AZURE_OPENAI_API_VERSION=2024-10-21
Every example degrades gracefully: if AZURE_OPENAI_API_KEY is unset it prints
the required variables and exits 0 instead of making a network call. This
keeps go run and CI safe with no credentials.
This example: main.go — chat and streaming
Basic single-agent chat, in both standard (full response) and streaming (token-by-token) modes.
# Standard mode — waits for the full response, then prints
go run ./examples/azure/main.go
# Streaming mode — prints tokens as they arrive
go run ./examples/azure/main.go -stream
It builds a provider with model.NewAzureOpenAIWithConfig and wraps it in an
agent.New(...).WithModel(provider).Build() agent.
More Azure examples
| Example | Demonstrates |
|---|---|
examples/azure_tools |
Multi-round tool calling (calculator + lookup) with the StopReasonToolCall loop |
examples/azure_rag |
RAG with model.NewAzureOpenAIEmbeddings, a knowledge.VectorKnowledge, and a self-contained in-memory storage.VectorStore |
YAML configs (no Go code)
Run the same setups declaratively through the CLI:
examples/yaml-configs/providers/azure.yaml— single Azure agentexamples/yaml-configs/azure-team.yaml— a sequential researcher → writer → editor Azure team
CHRONOS_CONFIG=examples/yaml-configs/providers/azure.yaml go run ./cli/main.go run "Hello"
CHRONOS_CONFIG=examples/yaml-configs/azure-team.yaml go run ./cli/main.go team run azure-content "Write about durable AI agents"
Documentation
¶
Overview ¶
Azure OpenAI example for Chronos.
This example demonstrates how to use Chronos with Azure OpenAI, in both standard (full response) and streaming (token-by-token) modes.
Prerequisites:
- An Azure OpenAI resource with a deployed model
- Go 1.24+
Set the following environment variables before running:
export AZURE_OPENAI_API_KEY=<your-azure-api-key> export AZURE_OPENAI_ENDPOINT=https://<your-resource>.openai.azure.com export AZURE_OPENAI_DEPLOYMENT=<your-deployment-name> export AZURE_OPENAI_API_VERSION=2024-12-01-preview
Run (standard mode — waits for full response, then prints):
go run ./examples/azure/main.go
Run (streaming mode — prints tokens as they arrive):
go run ./examples/azure/main.go -stream
Source Files