cohere-observability

v2026.09.24

Instrument Cohere requests, streams, retrieval, tools, limits, quality, and cost with low-cardinality telemetry and safe traces. Use when monitoring a Cohere service. Trigger with "Cohere observability", "Cohere metrics", or "monitor Cohere".

GitHub
Install command
npx skhub add jeremylongshore/cohere-observability
Markdown
SKILL.md

Cohere Observability

Overview

Expose enough evidence to separate application, retrieval, model, capacity, and provider failures without logging credentials or customer content.

Prerequisites

  • The target repository, runtime, environment, and accountable owner
  • An approved Cohere team and key for any live verification
  • Current quality, security, privacy, capacity, and change-control requirements

Tool Discipline

Use Read, Glob, and Grep to inspect code, configuration, and evidence. Use WebFetch only for current Cohere primary documentation. Use Write or Edit only when the user requested implementation and the exact target files are known; never write credentials or customer content.

Current Contract

  • Measure queue, retrieval, provider, first-event, completion, and tool durations separately.
  • Label metrics with bounded dimensions such as endpoint, operation, resolved model, environment, outcome, and retry class.
  • Keep prompts, documents, embeddings, keys, and raw tool arguments out of telemetry by default.
  • Pair operational metrics with sampled quality, citation, and safety evaluations.

Authentication

Use an environment-specific key injected from an approved secret manager. Never print, persist, commit, or place CO_API_KEY in an example. Confirm access with the least costly bounded operation appropriate to the task, and treat key creation, rotation, revocation, role changes, and production-capacity requests as owner-approved actions.

Instructions

  1. Define service objectives and an allowlist of telemetry fields before instrumenting.
  2. Create counters for requests, outcomes, retries, throttles, circuit transitions, and tool decisions.
  3. Create histograms for queue, retrieval, provider, first-event, completion, and end-to-end duration.
  4. Record usage and cost units without tenant names or unbounded request IDs as metric labels.
  5. Build alerts for availability, sustained throttling, queue age, latency, quality, and spend.
  6. Test redaction, trace propagation, stream cancellation, and incident correlation.

Approval Boundaries

Do not expose or rotate keys, change Cohere Team roles, accept commercial terms, enable sensitive production data, increase spend or capacity, switch production models, send a support bundle, or execute model-proposed side effects without the accountable owner's approval. Keep diagnosis read-only unless implementation was requested.

Output

Return the resolved API and model contract, files or settings inspected, evidence collected, validation result, remaining risk, owner, and rollback or next action. Redact keys, authorization headers, prompts, retrieved documents, embeddings, customer identifiers, and unrestricted environment output.

Error Handling

ConditionResponse
Cardinality spikeRemove unbounded labels and aggregate request identifiers in logs only.
Prompt in traceRedact it and review retained telemetry as a data incident.
Missing terminal eventMark the stream incomplete and count a distinct outcome.
False provider alertSplit local queue/retrieval time from Cohere duration.

Examples

Use this compact handoff shape to keep the selected scope, validation evidence, and operational result reviewable.

Input:

service=rag-api; backend=otel; content-logging=off; quality-sampling=1%

Expected handoff:

metrics=bounded; traces=redacted; slo=defined; alerts=tested

Resources

Discovery
Tags

No tags published for this skill.

Version
Latest version metadata

Version

v2026.09.24

Published

Sep 24, 2026

Category

Uncategorized

License

MIT

Source path

skills/.curated/cohere-observability

Default branch

main

Latest commit

e5a6c3b

Tree SHA

c2dc8e8