Duplication in LLM Input/Output Fields: Expected Behavior?

·Sep 18, 2025 02:48 PM·

I noticed in https://github.com/Arize-ai/openinference/blob/main/spec/semantic_conventions.md there are both input.value and llm.input_messages , looking at a real example shows the input messages are duplicated in these two fields, is that expected? Similar observations for output.value and llm.output_messages is also noticed.

2 comments

· Sorted by Oldest

RunLLM
APP
·
·
Yes, this is expected. The OpenInference semantic conventions specify both `input.value` (a raw string or JSON of the input payload) and `llm.input_messages` (a flattened, indexed list of chat messages) for LLM spans. This allows for both a full, machine-readable input and a structured, human-readable breakdown. The same applies to `output.value` and `llm.output_messages` for outputs. This duplication ensures compatibility with various tools and UIs, and is reflected in both the spec and real-world traces (<https://github.com/Arize-ai/openinference/blob/main/spec/semantic_conventions.md|semantic conventions>, <https://github.com/Arize-ai/phoenix/blob/main/docs/section-learn/tracing/what-are-traces.md|example spans>, <https://github.com/Arize-ai/openinference/blob/main/spec/llm_spans.md|llm_spans.md>). Would you like more detail on how these fields are populated or used in Phoenix and OpenInference? Sources: - <https://github.com/Arize-ai/openinference/blob/main/spec/llm_spans.md|llm_spans.md> - <https://github.com/Arize-ai/openinference/blob/main/spec/semantic_conventions.md|semantic conventions> - <https://github.com/Arize-ai/phoenix/blob/main/docs/section-learn/tracing/what-are-traces.md|example spans>
💡 Hint: Mention RunLLM in the thread for followups.
Roger Y.
·
·
this is because the input.value and output.value are raw data structures that can vary from vendor to vendor, but llm.input_messages and llm.output_messages are extracted objects that follow a consistent structure.