Repository navigation
Document opt-in full-text analytics and the 0.9 engine - #789
Conversation
There was a problem hiding this comment.
Code Review
This pull request updates the documentation to introduce opt-in Full-text Analytics for Harper v5.4.0, detailing configuration options, metrics, and fields. It also documents the upgrade of the full-text search engine to @harperfast/fulltext version 0.9.0, explaining checkpoint reuse and new query capacity settings. The review feedback recommends avoiding the <VersionBadge> component mid-sentence in lists, suggesting plain text instead, and restructuring a paragraph in the analytics overview to improve readability and resolve grammatically disjointed phrasing.
🚀 Preview DeploymentYour preview deployment is ready! 🔗 Preview URL: https://preview.harper-documentation.harperfabric.com/pr-789 This preview will update automatically when you push new commits. |
🚀 Preview DeploymentYour preview deployment is ready! 🔗 Preview URL: https://preview.harper-documentation.harperfabric.com/pr-789 This preview will update automatically when you push new commits. |
🚀 Preview DeploymentYour preview deployment is ready! 🔗 Preview URL: https://preview.harper-documentation.harperfabric.com/pr-789 This preview will update automatically when you push new commits. |
🚀 Preview DeploymentYour preview deployment is ready! 🔗 Preview URL: https://preview.harper-documentation.harperfabric.com/pr-789 This preview will update automatically when you push new commits. |
🧹 Preview CleanupThe preview deployment for this PR has been removed. |
⊙ Problem
Operators need configuration and interpretation guidance for the optional full-text analytics added by Add opt-in native full-text analytics. The existing full-text overview and release notes also describe the older engine version.
🤔 Is this documentation needed before enabling the feature?
Yes. Exact cumulative counters, unavailable observations and worker-specific series require different interpretation from averaged request metrics. Keep collection opt-in and explain its costs without claiming a production overhead bound.
💡 Solution
Document how to enable collection dynamically, read the
fulltextmetric, and calculate rates without losing integer precision or crossing reset boundaries. Update the engine guidance to fulltext 0.9.0 while retaining the existing migration limits.🔧 Changes
Product and architecture tour
Configure, interpret, then upgrade
Start with configuration, then interpret observations and check upgrade requirements:
analytics.fullText, a minimalenabled: trueexample, implied defaults, validated cadence bounds, optional histograms and reader-byte observations. Live configuration notes cover override retirement after environment overlays and the union of recording requests during operator export.storage.fullText.querycapacity settings:maxQueryTermsdefaults to 64 (maximum 1024);maxQueryClausesdefaults to 256 (maximum 4096). Changes require a full Harper process restart;restart_servicealone does not apply them.✅ Verification
npm run buildpassed and generated 426 documents for the final wording.npm run format:writeandnpm run format:checkpassed.Docs for: Add opt-in native full-text analytics.
Depends-on: HarperFast/harper#3946
🤔 Coordinate pending reference changes
Document fulltext 0.8 query policy and fuzzy controls overlaps engine/query wording; preserve 0.9 and both features. Assemble the v5 reference and open release notes from harper and harper-pro moves these pages. If that cutover lands first, port this docs PR into the owning code repository rather than restoring deleted reference files.
Comment generated by kAIle (GPT-6)
Related PRs: #777 overlaps (engine/query guidance), #788 overlaps (reference migration)
Release target: v5.4 — documents the new development-line analytics surface
Complexity: medium
Deferred findings
b7fa803b170f80d5nit documentation —reference/configuration/options.md— docs say internal consumers can keep collection active; no core caller exists — filed Implement full-text generation lifecycle, rebuild, eviction, backup, and status harper#2510Review-Coverage: authored=codex; ran=cursor-composer,claude; adjudicated=domain(claude); declined=gemini,cursor-grok,muse; rounds=4; full=2 @ 549b795
Review-Attention: skim ~2m (decisions: doc-the-loss-vs-fix-core, query-knobs-in-this-pr, engine-bump-bundled, dual-fulltext-config-entries) @ 549b795