Crate otel_genai_bridge 

Translate LLM telemetry attributes between OpenInference and the OpenTelemetry GenAI semantic conventions.

Two specs in flight:

• OpenInference — Arize’s instrumentation lib. Names like llm.model_name, llm.token_count.prompt.
• OpenTelemetry GenAI semantic conventions — the upcoming standard. Names like gen_ai.request.model, gen_ai.usage.input_tokens.

Phoenix wants OTel GenAI (Arize-ai/phoenix#10622), Traceloop wants the new keys (traceloop/openllmetry#3515), but most existing instrumentation emits OpenInference. Until backends converge, you ingest one and may want to read or emit the other. otel-genai-bridge is the smallest possible primitive that translates.

Quick start

use otel_genai_bridge::{to_otel_genai, to_openinference};
use serde_json::{json, Value};
use std::collections::HashMap;

let mut openinference: HashMap<String, Value> = HashMap::new();
openinference.insert("llm.model_name".into(), json!("claude-sonnet-4"));
openinference.insert("llm.token_count.prompt".into(), json!(1234));
openinference.insert("llm.token_count.completion".into(), json!(56));

let otel = to_otel_genai(&openinference);
assert_eq!(otel.get("gen_ai.request.model").unwrap(), &json!("claude-sonnet-4"));
assert_eq!(otel.get("gen_ai.usage.input_tokens").unwrap(), &json!(1234));
assert_eq!(otel.get("gen_ai.usage.output_tokens").unwrap(), &json!(56));

// Round-trip back to OpenInference:
let back = to_openinference(&otel);
assert_eq!(back.get("llm.model_name").unwrap(), &json!("claude-sonnet-4"));

Constants

OPENINFERENCE_TO_OTEL

OpenInference → OTel GenAI attribute name pairs.

OTEL_TO_OPENINFERENCE

OTel GenAI → OpenInference attribute name pairs (the inverse).

Functions

known_mappings

Pretty alias for the union of mappings (deduped).

to_openinference

Translate a flat attribute map from OTel GenAI to OpenInference.

to_otel_genai

Translate a flat attribute map from OpenInference to OTel GenAI.