Algolia

• Paste your URL, click Start — that's it
• Multi-threaded crawler with intelligent JS rendering
• Three-tier rendering pipeline: curl-cffi, httpx, Playwright headless Chromium
• Auto-detects SPAs (React, Next.js, Angular, Vue, Nuxt, SvelteKit, Gatsby)
• Crawls 21 MIME types: HTML, PDF, DOCX, XLSX, PPTX, ODT, ODS, ODP, RTF, MSG, and more
• Robots.txt obedience, spider trap detection, sitemap following
• Configurable scope: domain, subdomain, path, or full web
• Scheduled recrawls (hourly, daily, weekly)

• No built-in crawler — you must write code to push records via their API
• Algolia Crawler exists as a separate paid product with limited features
• You structure your data as JSON and maintain push scripts
• When your content changes, you update and re-push — manually or via custom scripts
• No document format extraction — want PDF search? Build your own pipeline

• Vector mode: Normalized weighted sum of lexical + vector scores with tunable weights and log normalization
• RRF mode: Reciprocal Rank Fusion — two separate requests merged mathematically for the best of both worlds
• Solr mode: Lexical-first search with vector reranking — precision-focused
• 1024-dimensional multilingual embeddings (50+ languages) on title and description fields
• KNN cosine similarity on dense vectors
• Per-field boost weights, phrase matching multipliers, minimum match tuning
• Typos? The vector model understands what you meant, not just what you typed — semantic understanding makes traditional typo tolerance look primitive
• Plus spellcheck with "Did you mean?" suggestions on top of that

• "NeuralSearch" exists but it's a black box — no control over modes, weights, or normalization
• No user-tunable hybrid parameters
• No choice between search strategies
• Typo tolerance is good, but it only handles character-level errors — it doesn't understand meaning
• Cannot tune field weights, phrase boosting, or minimum match

• Streaming AI hints powered by a GPU-accelerated LLM
• Context-aware: sends top results (title + description + content) to the LLM
• Real-time Server-Sent Events streaming directly in the search UI
• Answers appear as the user searches — no extra clicks, no separate page
• Built on your own indexed content, not hallucinated from training data

• No built-in LLM integration
• To get AI summaries, you'd build your own RAG pipeline on top of Algolia
• That means another service, another API, another bill

• Full themed search page with dark and light modes
• Infinite scroll or traditional pagination
• Faceted navigation (language, locale, source, custom facets)
• OG image previews, favicons, content type icons
• Mobile-responsive out of the box
• Configurable via URL parameters — no code needed
• One-line embed code: drop an iframe and you're done

• Provides InstantSearch.js widget library for React, Vue, Angular
• YOU assemble the UI from components
• More flexible for developers, but far more work for everyone else
• No ready-to-embed, zero-code search page

• Query Analytics: Top queries, daily trends, query length distribution, CSV export
• Zero-Results Dashboard: Every zero-result query tracked by unique IP — find your content gaps
• Click Analytics with CTR: Track which results get clicked, click-through rates per query, detect low-CTR queries that need better results
• Bulk management: Select and delete junk/test queries across all tabs
• All included in every plan

• Analytics exists but is a paid add-on on higher tiers
• Click analytics requires additional client-side integration code
• No zero-result tracking out of the box
• You pay more to understand how your own users search

• Pin specific documents to the top for specific queries
• Exclude documents from appearing for specific queries
• Global wildcard rules that apply to ALL queries
• Visual elevation bar directly on the search results page
• One-click pin/exclude while browsing results — no context switching

• Has "Rules" for pinning and hiding results
• But the Rules UI is separate from search results — you can't pin while searching
• No global wildcard rules
• More cumbersome workflow for result curation

• Crawler automatically extracts prices from JSON-LD, microdata, and meta tags
• Price range slider in the search UI (no code needed)
• Sort by price (ascending, descending, or by relevance)
• Currency detection and display
• Works for e-commerce sites out of the box

• You must manually structure price data in your JSON records
• No automatic extraction from web pages
• Price faceting available but requires manual schema design

• HTML, PDF, DOCX, XLSX, PPTX, ODT, ODS, ODP, RTF, MSG (Outlook email), plain text, XML, RSS, JSON
• Full text extraction with metadata preservation
• Document reader lets users view extracted content inline without leaving search results
• The crawler handles everything — you don't convert, parse, or pre-process anything

• Only indexes JSON records you push via API
• Want to search PDFs? Build a PDF extraction pipeline yourself
• Want to search Word documents? Same story
• Every non-HTML format is your problem to solve

• VADER sentiment scoring on every crawled page (positive, negative, neutral, compound)
• Language detection via langid (50+ languages)
• Language and locale facets in the search UI
• All automatic — no configuration needed

• No sentiment analysis
• Basic language detection but nothing automatic or enriching

• "Did you mean?" spellcheck suggestions
• Edge n-grams for instant prefix matching (autocomplete)
• ASCII folding for accent-insensitive search (cafe = café)
• Stemming and synonym support
• And on top of all that, vector search that understands meaning regardless of exact spelling

• Typo tolerance is solid (one of Algolia's strengths)
• But it's a black box — no tuning available
• No vector-level semantic understanding of typos

• Exclude specific URL patterns from search results via search.xml config
• Regex-based exclusion patterns
• Combined with Query Elevation for full result curation

• No URL exclusion mechanism — you'd remove records via API calls
• Content control is code-driven, not configuration-driven

• Fixed monthly pricing — search all you want, no per-query charges ever
• No per-record-per-month storage fees
• All features included: crawling, hybrid search, AI Hints, AI Reader, analytics, elevation, search UI
• Free tier: index up to 100 pages with full features — hybrid search, AI, everything. No credit card games.
• Paid plans: from ~$20/month for small sites up to ~$200/month for unlimited corporate enterprise search
• Built-in rate limiter (per minute / per hour) — bots hit the wall, not your invoice
• Your bill this month is identical to next month — even during traffic spikes or bot attacks

• Charges per search request: $0.50 per 1,000 searches after a free tier of 10,000/month
• Charges per record stored per month
• Analytics, AI, and advanced features are paid add-ons on top
• At 100,000 searches/month you pay $45/month just in query fees — before records, before AI, before analytics
• A traffic spike or bot attack = an enormous overnight bill. Algolia's response: "Usage is usage. Not our problem."
• Algolia's pricing page is deliberately confusing — good luck calculating your actual cost before you're committed

• Data lives on dedicated Solr clusters
• No vendor lock-in — standard Apache Solr under the hood
• Master-replica architecture with automatic failover
• Full Solr API access to your index
• Can migrate to self-hosted Solr at any time — your schema, your data, your rules

• Proprietary engine — your data is in their cloud, in their format
• Migration out requires rebuilding everything from scratch
• No standard API compatibility with anything else
• You're locked in the moment you integrate

• POST JSON payloads with up to 50 documents per batch
• Or upload a .json file — ideal for large batches from CMS exports or data pipelines
• URI-based document identity: every document needs a URI, and the document ID is always md5(uri). Same URI = same document. Resubmit to update.
• Automatic dedup protection: duplicate URIs already in the queue are rejected before processing
• Rich text extraction: set rtf: true and pass a URL to a PDF, DOCX, PPTX, ODT, XLSX, or RTF — Opensolr extracts the text for you
• Full Solr error reporting per document — know exactly which document failed and why
• Returns doc_ids array in every response for tracking
• All the same features: dense vectors, hybrid search, AI hints, query elevation, analytics, the full search UI
• Complete PHP, Python and cURL examples in the documentation

• JSON record push via API — this is their only method of getting data in
• No file upload — you must always construct and send JSON programmatically
• No built-in rich text extraction — want to index a PDF? Extract it yourself first
• No URI-based dedup — you manage document identity and deduplication in your own code
• No integrated queue or job status — you build your own pipeline
• Per-record-per-month charges on top of everything else

• Crawl your website automatically
• Push JSON via the Data Ingestion API
• Use the standard Solr API to add, update, delete documents directly
• Mix all three on the same index

• Use the ready-made embeddable search UI
• Build your own UI against the standard Solr API
• Query via HTTP from any language: PHP, Python, JS, Go, Java — anything
• Full Solr syntax: facets, highlighting, spellcheck, sorting — all exposed

• Standard Apache Solr under the hood
• Export your schema and data at any time
• Migrate to self-hosted Solr with zero query changes
• Algolia? You rebuild from scratch on proprietary APIs

• Hybrid vector + keyword search across all your content
• AI Hints — streaming LLM answers from your own indexed content
• AI Reader — full-screen AI document summaries inline
• Faceted navigation: content type, author, language, date, price, custom fields
• Analytics dashboard built into Drupal admin
• Query Elevation: pin and exclude results per query, directly from search results
• Multi-language search with automatic locale filtering — adapts to any multilingual site
• Embeddable UI or fully customizable native Twig templates

• Indexing done by the Opensolr Web Crawler — zero Drupal-to-Solr field mapping code
• No Search API module required — completely standalone
• Never hits Solr's 1,000-field limit (pages indexed as URLs with ~40 fixed fields)
• Auto-detects Drupal's current language for locale-filtered results out of the box
• Drupal 10 + Drupal 11 compatible, PHP 8.1+
• Available on opensolr.com/drupal-ai-search and drupal.org/project/opensolr_search

Paste any website URL into the RAG in 60 Seconds sandbox. The Web Crawler fetches your pages, Opensolr generates the vector embeddings, and you get keyword + vector hybrid search plus a RAG answer on your own content — live, on a temporary index, with the keyword-only, vector-only and hybrid results shown side by side. No account, no OpenAI key, nothing to install.