Stream OpenDunes request traces to LangSmith for LLM debugging and evaluation.
Not yet available. This destination is on the roadmap.
LangSmith is LangChain's tracing and evaluation platform. Connecting it to OpenDunes via Broadcast gives you a logged trace for every request — useful for debugging prompt regressions, running offline evaluations, and curating fine-tuning datasets.
OpenDunes sends each request to LangSmith as a run: the input messages, model response, model slug, token counts, DA cost, and latency. The run type is llm. Any metadata you attach to the chat completion request is forwarded as run tags so you can filter traces by session, user, or experiment.
Once available, configure the LangSmith destination from your dashboard under Broadcast. You will need:
Runs arrive in LangSmith within seconds of the request completing.
Back to Broadcast overview.