Cohere

Install:

shell

pip install "nullrun[cohere]"

cohere_client.py

import nullrun
import cohere

nullrun.init(api_key="nr_live_...")

client = cohere.Client()
resp = client.chat(
    model="command-r-plus",
    message="Hello",
)

Patched via the per-vendor extractor in nullrun.instrumentation.auto. The extractor handles both Cohere v1 (prompt_tokens / completion_tokens) and v2 (input_tokens / output_tokens) response shapes, including the field fallback when one of the two shapes is missing.

Cohere streaming does not include usage in the streamed chunks — only the non-streaming response carries it. Tracked events for streamed Cohere calls will report tokens=0.

Test coverage

Per NullRun's testing policy (see the SDK README), Cohere patches have extractor unit tests only — no full transport-to-track integration test. The extractor shape is exercised by isolated tests, but no multi-roundtrip end-to-end test confirms the cost actually flows through /api/v1/track. This is documented per §8 of the source-of-truth positioning — verify behaviour against your real workload before relying on it.