Skip to content

Adam Greenwell

Frequently Asked Questions

Common questions about Lumenare Search.

Current release: 2.2.01

General Questions

Lumenare Search is a WordPress plugin that replaces the default search with live results, weighted relevance ranking, typo tolerance, analytics, Search Intelligence, filterable results and optional document indexing. Search & Answers combines question understanding, optional meaning reranking, Ask Site and cited answer cards in one settings area.

Is Lumenare Search free?

Yes, the plugin is free with no premium version, upsells or feature limitations. Optional AI services may charge for content analysis, query embeddings and generated answers.

Does it work with my theme?

Lumenare Search works with any properly-coded WordPress theme. It automatically replaces default search forms and integrates with the block editor. Custom styling can be added via CSS.

Will it slow down my site?

Performance depends on your content, hosting and enabled features. Lumenare Search limits work with:

  • Efficient database indexes
  • Cached live results and query embeddings
  • Batch processing for large indexing operations
  • Automatic new-content reindexing instead of repeated full rebuilds

Optional meaning reranking and AI-written answers add provider latency and possible charges. Test them on your own site before enabling them broadly; live typing requires its own explicit meaning-search preference.

Does it work with WooCommerce?

Yes. Enable "Products" in the post types setting to search WooCommerce products. Product categories and tags can be weighted via taxonomy settings.

Installation & Setup

  1. Go to Plugins → Add New in WordPress
  2. Search for "Lumenare Search"
  3. Click Install Now, then Activate
  4. Go to Settings → Lumenare Search → Index Management
  5. Click "Reindex All Content"

Do I need to configure anything?

Lumenare Search works out of the box with sensible defaults. For most sites, apply the recommended Setup preset and build the initial index. Advanced configuration is optional.

How long does indexing take?

Depends on content volume:

  • Small sites (< 100 posts): Seconds
  • Medium sites (100-1000 posts): 1-5 minutes
  • Large sites (1000+ posts): 5-30 minutes

Indexing runs in batches to prevent timeouts.

Can I index custom post types?

Yes. Go to Settings → Lumenare Search → General → Post Types and check any custom post types you want to include. Then reindex.

Search Functionality

What's the difference between ANY and ALL match modes?

ANY (OR): Returns results containing at least one keyword.

  • Search: "WordPress security" → Results with "WordPress" OR "security"
  • More results, broader discovery

ALL (AND): Returns results containing all keywords.

  • Search: "WordPress security" → Results with "WordPress" AND "security"
  • Fewer results, more precise

How does fuzzy matching work?

Fuzzy matching finds results despite typos using Levenshtein distance (edit distance). If a user searches for "computr" (missing an 'e'), the system finds "computer" in the index because it's only 1 character different.

The threshold setting controls how many character differences are allowed (default: 2).

What are stop words?

Stop words are common words excluded from indexing because they appear in almost every post and don't help identify relevant content. Examples: "a", "an", "the", "is", "and".

Removing these reduces index size and improves relevance.

How do synonyms work?

Synonyms expand search queries. If you configure "car, automobile, vehicle" as synonyms:

  • User searches "car"
  • System also searches for "automobile" and "vehicle"
  • Results containing any of these words are returned

Synonyms are applied at query time—no reindexing needed.

What is phrase boosting?

Phrase boosting increases relevance for specific multi-word terms. Without it, searching "Master System" might rank posts about any "system" higher than posts specifically about "Sega Master System."

With phrase boosting, exact phrase matches get significant score bonuses.

Relevance & Rankings

How are search results ranked?

Results are ranked by relevance score, calculated from:

  1. Keyword matches – weighted by location (title > excerpt > content)
  2. Taxonomy matches – bonus for matching categories/tags
  3. Phrase boost – bonus for exact phrase matches
  4. Freshness, trending, audio, CTR, post-type, and promotion boosts – optional ranking layers

Higher scores appear first.

Why are my results not relevant?

Common causes:

  1. Weights need tuning – Adjust title/content weights
  2. Missing stop words – Add common irrelevant words
  3. Missing synonyms – Add equivalent terms
  4. Wrong match mode – Try switching ANY/ALL
  5. Need phrase boosting – Add important multi-word terms

See Tuning Guide for details.

Do I need to reindex after changing settings?

Requires reindex:

  • Stop words
  • Post types
  • Indexed custom fields
  • Document indexing or document MIME type changes
  • Enabling a field that had a zero weight when content was indexed

Applied immediately:

  • Positive-to-positive field or taxonomy weight changes
  • Match mode
  • Synonyms
  • Search phrases
  • Fuzzy matching settings
  • Trending settings
  • Suggestion provider order
  • Query promotions
  • Audio, freshness, and click boost toggles

Meaning search also has separate prepared passages and embedding vectors. After content edits or changes to the analysis service, model, size, passage settings or enabled sources, save your settings and check Search & Answers → Advanced for pending preparation. Run Analyze site content for the selected configuration; a keyword reindex alone does not generate the needed vectors.

What analytics are tracked?

  • Search queries – What users search for
  • Results count – How many results each query returned
  • Zero-result queries – Searches that found nothing
  • Click-through – Which results users click
  • Trending – Queries gaining popularity
  • Result impressions – Which results were shown for each query
  • Search Intelligence actions – Reviewed/reopened states, promotions, synonym drafts, remediation outcomes, and Search ROI

Is user data collected?

Search queries, result impressions, clicks and Search Intelligence action metadata are stored locally in your WordPress database. Optional user ID is stored for logged-in users.

Optional AI features send data to the selected provider: content analysis sends eligible public passages, meaning reranking sends uncached search queries for embeddings, and AI-written answers send the question and a bounded bundle of retrieved public passages. Embedding vectors are stored locally. Adding an API key does not enable these features; choose the analysis run, search behavior and answer style explicitly. Sources-only answers using local keyword or question understanding do not require a provider.

The trending system:

  1. Tracks all search queries
  2. Calculates trending score based on frequency, growth rate, and recency
  3. Queries meeting minimum search threshold become "trending"
  4. Related content receives relevance boost

Trending scores are calculated daily via WordPress cron.

Yes. Add the Lumenare Search Trends widget or Lumenare Search Trends block to display analytics-backed trending or latest successful searches.

Latest searches default to repeated successful queries so one-off private searches are not promoted publicly. Lower the minimum occurrence setting only when the site's analytics policy allows raw latest-query display.

Can I disable analytics?

Analytics are always collected (they're needed for trending), but you can:

  • Disable trending ranking in settings
  • Clear analytics tables via database
  • Ignore the Analytics and Search Intelligence tabs

Ask Site

What is Ask Site?

Ask Site lets visitors ask a question in plain language and receive a cited excerpt or an optional AI-written answer from retrieved public site content. It shares retrieval, profiles and evidence checks with search and results-page answer cards. Configure it under Settings → Lumenare Search → Search & Answers.

Do I need an AI provider to use Ask Site?

No. Sources and cited excerpts can use local keyword or question understanding without an AI provider. AI-written answers with citations is a separate opt-in requiring WordPress 7.0 or newer with AI Client support and an approved, configured text-generation connector. Selecting meaning reranking independently adds an embedding-provider requirement, even when the answer style uses sources only.

Why can't visitors see the Ask Site form?

Ask Site hides itself from visitors until its base readiness requirements pass: public Ask Site enabled, a healthy index with public content, a non-zero rate limit and a resolvable search profile. Administrators still see the form so they can diagnose it. Check Search & Answers → Placement, the status information and the assigned Search Profiles settings. Add the Lumenare Ask Site block or [lumenare_ask_site] shortcode where visitors should ask questions.

I selected AI Client mode but only get sources. Why?

Confirm AI-written answers with citations is selected and saved under Search & Answers → Answers, including any assigned profile override. The site needs WordPress 7.0 or newer with AI Client support, configured credentials and any required approval for a text-generation connector. An embedding model chosen for content analysis is a separate capability.

If generation is unavailable, fails or returns invalid citations, Lumenare uses a supported cited excerpt when available. Results-page answer cards show a no-answer notice when no supported fallback is available. If the Ask Site form shows no useful response after a provider failure, check the service status and site error log.

Can Ask Site expose private or password-protected content?

Private and password-protected content cannot supply answer passages. Password-protected posts may remain findable by title in ordinary search, but their protected text is not used as answer evidence.

What happens to questions Ask Site can't answer?

They appear in Search Intelligence as opportunities, where you can create a draft answer, look for existing content that covers the question, mark it reviewed, or dismiss it. Bot-like, numeric, non-English, and imported query noise is classified separately so the list stays actionable.

Can I show an answer above normal search results?

Yes. Enable Show answer cards on full search results pages by default under Search & Answers → Placement. This is independent of public Ask Site placement. Cards use the answer style and limits from Answers, with profile overrides available in Search Profiles.

Cards appear on the first page of supported Lumenare HTML searches and keep the ordinary result list and pagination. They do not appear in the live dropdown or on later results pages. A recognized question without enough evidence gets a clear no-answer notice; an ordinary keyword search can simply show its usual results.

Does meaning reranking make every question answerable?

No. It reorders a bounded set of existing keyword or question-aware hits using compatible public passages. It cannot find an article outside those hits. Sources-only answers require a supporting live passage; opted-in AI writing can also assess bounded retrieved passages with checked citations and supporting quotes. A title match or high similarity score alone does not establish an answer.

How do I prepare content without WP-CLI?

Open Search & Answers → Advanced → AI service and model for content analysis. Choose an approved service, click Refresh model list, choose an embedding model and click Save Settings. Then click Analyze site content. Keep the tab open while the progress moves from preparing posts to analyzing passages. You can Pause after the current batch and Resume analysis later, including after reopening the page.

Compatible analyses are reused; missing or changed passages are sent to your selected provider. Service charges may apply. Opening the page, saving settings and Refresh status do not start this work. Analysis does not train an AI model or enable search, answer writing or placement switches.

Why is Analyze site content disabled?

Check the message beside the controls. The service needs credentials and Lumenare approval, the model must support embeddings, and the selection and analysis settings must be saved. After fixing setup, click Refresh status. A temporary service backoff or another running batch may require waiting.

A database-read error is separate from connector approval. If an older build logs Illegal mix of collations during content analysis, upgrade to 2.2.00, which fixes that comparison without requiring database collation conversion. See the Troubleshooting Guide for persistent errors.

Do stored analysis counts prove meaning search is ready?

No. Stored counts include earlier configurations. Check pending work for the selected service and model and run analysis again after content edits or changes to the service, model, analysis size, passage settings or enabled sources. Meaning reranking accepts only compatible vectors whose passages still match current public content. Visitor requests do not repair stale passages or backfill missing vectors.

Can a PDF look correct but stop content analysis?

Yes. Its visible letters and its extractable text are separate. Damaged or missing text mappings can produce scrambled words while the page still looks correct. Readable PDFs with valid text mappings are supported; scanned, image-only files need OCR first. Also check missing uploads, permissions, encryption and file-size limits. Use Test Parse in Media Library, repair or replace the same attachment if needed, then Resume analysis. See the full PDF troubleshooting steps, including how to intentionally omit PDF content.

Document IDs in an AI-service error may identify every source in the submitted batch, including healthy PDFs. They are not proof of damage. Follow the error category: resolve timeouts, rate limits, quota, credentials or model issues with the service before changing files.

Technical Questions

What database tables are created?

Lumenare Search creates these tables as needed:

  • wp_lumenare_search_index – Keyword index
  • wp_lumenare_search_chunks – Prepared passages for optional meaning search
  • wp_lumenare_search_vectors – Locally stored embedding analyses
  • wp_lumenare_search_analytics – Search query logs
  • wp_lumenare_search_trending – Trending topic scores
  • wp_lumenare_search_clicks – Click-through tracking
  • wp_lumenare_search_result_impressions – Results shown for known searches
  • wp_lumenare_search_import_runs – Resumable third-party imports
  • wp_lumenare_search_suggestion_terms – Imported/popular suggestion seeds
  • wp_lumenare_search_opportunity_actions – Search Intelligence workflow state
  • wp_lumenare_search_query_promotions – Reversible exact-query boosts
  • wp_lumenare_search_event_queue – Durable telemetry queue
  • wp_lumenare_search_security_logs – Security filtering logs
  • wp_lumenare_search_ability_logs – Abilities API audit logs when enabled

Does it use external APIs?

Local keyword search, question understanding and sources-only answers can run without external APIs. Optional AI work uses the site's WordPress AI integrations:

  • Refresh model list requests service metadata without sending site content.
  • Analyze site content sends eligible public passages to the saved embedding service and stores the returned vectors locally.
  • Meaning reranking sends uncached queries for embeddings; live typing requires an explicit opt-in.
  • AI-written answers send the visitor's question and bounded retrieved public passages through the WordPress AI Client.

API keys and connector installation alone do not enable these features. Content analysis, search behavior, answer style and public placement are separate choices, and provider charges may apply.

Is it compatible with caching plugins?

Yes. Search results are cached in WordPress transients. Compatible with WP Super Cache, W3 Total Cache, LiteSpeed Cache, and similar plugins.

Verify that your caching setup respects search parameters, filters and profiles, and does not serve shared cached responses for live AJAX requests.

Does it work with multisite?

Yes. Each site in a multisite network has its own search index and settings.

What PHP version is required?

PHP 7.4 or higher. PHP 8.x is supported, with PHP 8.3 compatibility improvements included in the current release.

What WordPress version is required?

WordPress 6.2 or higher. The current release is tested up to WordPress 7.1.

Troubleshooting

Search isn't working at all

  1. Check plugin is enabled in settings
  2. Verify index exists (check Index Management section)
  3. Check browser console for JavaScript errors
  4. See Troubleshooting Guide

I get no results for queries that should match

  1. Check post types setting includes your content
  2. Verify content is published (not draft)
  3. Check stop words aren't filtering your terms
  4. Try "Any (OR)" match mode
  5. Reindex content

How do I reset everything?

  1. Deactivate the plugin
  2. Delete plugin data (optional: delete wp_lumenare_search_* tables)
  3. Reactivate the plugin
  4. Rebuild index

Where can I get help?

WordPress.org Support Forums

Feature Requests

Can you add [feature]?

Feature requests are welcome! Post them on the WordPress.org support forums with:

  • Description of the feature
  • Use case (why you need it)
  • How you envision it working

Is there a premium version coming?

No. Lumenare Search is and will remain 100% free.

Search & Answers

Back to Documentation Index →