examples

tls-eino-demo

A minimal, runnable Eino demo that ships traces to Volcengine TLS through the callbacks/tls handler.

It builds a real compose.Chain (chat prompt template → chat model) and invokes it both non-streaming and streaming. The TLS handler is registered as a global Eino callback, so the agent.turn / llm.request spans are produced by the framework itself — this is what a production integration looks like. A mock ChatModel stands in for a real provider so the demo runs without LLM credentials; swap it for components/model/ark and the spans stay identical.

Run

export TLS_EXPORTER_ENABLED=true
export TLS_APP_NAME=tls-eino-demo
export TLS_LOG_ENDPOINT=tls-cn-beijing2.volces.com   # optional; derived from region if empty
export TLS_LOG_REGION=cn-beijing2
export TLS_LOG_TRACE_TOPIC_ID=<your-trace-topic-id>
export TLS_LOG_API_KEY=<your-tls-api-key>            # or TLS_LOG_AK + TLS_LOG_SK

go run .

On success the traces appear in the TLS LLM dashboard, grouped by the demo's session_id. If the TLS service rejects the credentials you'll see a 403 from the producer (e.g. InvalidAccessKeyId / API-key errors) — that means the pipeline works and only the credentials need fixing.

What it does

1. Resolves TLSConfig from the environment (defaulting the endpoint from the region when TLS_LOG_ENDPOINT is empty).
2. Creates the TLS handler and registers it via callbacks.AppendGlobalHandlers.
3. Attaches a session via tls.SetSession so traces group per session.
4. Runs Chain.Invoke and Chain.Stream; the framework fires callbacks that the handler converts to TLS spans and flushes on shutdown.