LLM Obs: Patterns coverage + Playground record cap

[thomaschin35]

What does this PR do? What is the motivation?

Three LLM Observability documentation pages are updated to reflect four merged backend PRs:

• PR modified APM retention policy to reflect recent updates #6284 – Added a coverage field to Patterns configuration, allowing users to associate datasets with a Pattern for offline coverage computation.
• PR correct CWS rules filter #11293 – Added a suggest_datapoints activity that automatically marks individual production interactions as suggested for dataset addition.
• PR Fix further reading link text on MwL #12143 – Raised the maximum dataset records processed per Playground-launched experiment from 1,000 to 10,000.

Changes

content/en/llm_observability/monitoring/patterns.md

• Added Coverage datasets (optional) sub-item to the Scope configuration step, explaining how to associate datasets for offline coverage measurement.
• Added Coverage column to the topic list table describing the coverage ratio and suggested datapoints count.
• Extended the "Find evaluation coverage gaps" section to explain that Patterns now computes offline coverage automatically when a coverage dataset is configured, including global ratio and per-dataset breakdown.
• Added a note to the "Export and act on interactions" section explaining that Patterns marks individual interactions as suggested to help close coverage gaps, and that they can be added to a dataset with Add to Dataset.

content/en/llm_observability/playground.md

• Added an alert-info callout after the "Save the experiment" step noting that Playground experiments process up to 10,000 dataset records per run and that only the first 10,000 are processed if the dataset exceeds this limit.

Merge readiness

• Ready for merge

AI assistance

Used Claude Code (Bits Code) to draft and apply all documentation changes based on PR descriptions provided.

Additional notes

Draft PR — pending review before merge.

---

PR by Bits - View session in Datadog

[datadog-staging]

View session in Datadog

Bits Code status: ✅ Done

CI Auto-fix requires auto-push.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: cd7f9191d4

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Reconcile the full-dataset claim with the cap

For datasets with more than 10,000 records, this new note contradicts the paragraph immediately above that says the experiment runs across all records in the dataset. Readers can come away thinking a saved experiment evaluates every row even though this callout says the run is capped, so please update the full-dataset sentence or combine it with this note to make the capped behavior unambiguous.

Useful? React with 👍 / 👎.

[buraizu]

@thomaschin35 thanks for the PR! Since this feedback hasn't been addressed, I'll add the WORK IN PROGRESS label for now. Could you please address this feedback, or provide a rationale for ignoring it, and then remove the WORK IN PROGRESS label to re-open it for documentation team review?

Let me know if you have any questions!

[github-actions]

Preview links (active after the build_preview check completes)

Modified Files

• https://docs-staging.datadoghq.com/dd/thomas.chin/patterns-coverage-playground/llm_observability/monitoring/patterns
• https://docs-staging.datadoghq.com/dd/thomas.chin/patterns-coverage-playground/llm_observability/playground