Auto-Compaction (Azure OpenAI)

examples/compaction-azure-openai.rs demonstrates enable_auto_compaction end to end on Azure OpenAI. Azure's Responses API accepts the same context_management compaction entries as direct OpenAI and emits the same encrypted compaction items; appam's Azure URL construction and api-key authentication work unchanged.

Run

export AZURE_OPENAI_API_KEY="..."
export AZURE_OPENAI_RESOURCE="my-resource"
export AZURE_OPENAI_MODEL="gpt-5.4-mini"   # deployment name, optional
cargo run --example compaction-azure-openai

What this example actually configures

• LlmProvider::AzureOpenAI with your resource name and API version
• enable_auto_compaction(16_000) — comfortably above the compacted-window size
• One local tool: fetch_catalog (~4K tokens per section)
• An on_compaction stream hook and a post-run verification report

Key builder setup

let agent = AgentBuilder::new("compaction-demo-azure-openai")
    .provider(LlmProvider::AzureOpenAI {
        resource_name: resource_name.clone(),
        api_version,
    })
    .model(&model)
    .system_prompt("You are a star catalog librarian. ...")
    .enable_auto_compaction(16_000)
    .with_tool(Arc::new(fetch_catalog()))
    .max_tokens(4096)
    .build()?;

See the Context Compaction guide for how compaction works across providers.