Current Stage: Delta

Sweden's Search Engine

Built from scratch for Swedish language, geography, and trust.
Privacy-first. No ads. No tracking.

0
Documents Crawled
0
Domains
0
Municipalities
0
Pages Indexed
Scroll

Built from scratch for Sweden

Klar is a Swedish-first, server-based search engine — not a fork, not a wrapper, not a reskin. Every component from the crawler to the ranking engine was built specifically for Swedish language, geography, and digital sovereignty.

Unlike global search engines that treat Swedish as "English translated through ML," Klar understands Swedish natively — compound words, grammar patterns, municipalities, government agencies, and cultural context.

It runs on its own infrastructure: a Python/Flask API server, a SQLite instant-index engine with per-page NLP processing, a KDB segment-based search storage layer, and a single-page web frontend with Leaflet/OSM maps, 5 map modes, and OSRM routing.

10.7M
Index entries
1.4M
Unique terms
60
Entities detected
4
KCA algorithms
🔤

Swedish NLP

Compound splitting, lemmatization, å/ä/ö normalization, stop-word removal

🗺️

Geo Intelligence

290 municipalities, 21 counties, .se TLD boost, geo coordinate matching

🛡️

Trust System

3-tier domain authority, verified badging, credibility scores (0.0–1.0)

🧠

Intent Parsing

85+ NLU patterns, WikiAnswer, section extraction, entity detection

Instant Indexing

Per-page SQLite processing at crawl time — queryable immediately, no rebuild

🔍

5 Map Modes

Directions, nearby, weather, explore, location — with OSRM routing in Swedish

Sweden deserves better search

Global search engines weren't built for Swedish. Here's why Klar had to be built from scratch.

!

Google Doesn't Understand Swedish

Swedish compound words like arbetstillstånd (arbete + tillstånd) or riksdagsledamot (riksdag + ledamot) are treated as opaque tokens. Google translates Swedish through English ML models instead of understanding Swedish grammar natively. Klar's LI-Flow algorithm decomposes compounds in real time.

🔒

Privacy Is Not Optional

Global search engines track, profile, and monetize user data. Klar is GDPR-native by design: no tracking, no profiling, no ads, no data selling. 73% of Swedes want a Swedish alternative to Google. Klar is that alternative.

🇸🇪

Swedish Content Gets Buried

.se domains compete with international SEO spam in Google's global index. Official Swedish sources (Skatteverket, Försäkringskassan, 1177) get buried under generic .com results. Klar's SveaTrust system prioritises verified Swedish sources with a 3-tier domain authority model.

🎯

The Vision: Sweden's National Search Infrastructure

"Klar is more than a product — it's a platform reinvented from scratch for Swedish search and public trust. Not competing with Google. Competing with zero."

Built in Sweden, for Sweden. Deterministic, explainable, privacy-first.

Four proprietary KCA algorithms

Klar Custom Algorithms (KCA) are the four engines that power every search result. Each is handcrafted for Swedish language, geography, trust, and intent.

LI

LI-Flow — Linguistic Intelligence Flow

Swedish NLP, compound splitting, lemmatization, normalization

LI-Flow is Klar's language understanding engine. It decomposes Swedish compound words into their morphemes (e.g. öppettideröppet + tider), applies Snowball stemming, strips Swedish stopwords, normalizes å/ä/ö variations, and builds a token-level understanding that respects Swedish grammar patterns (V2 word order, definite suffixes).

Strategy
3-way compound splitter (longest-first, recursive, lookup)
Coverage
1.4M unique terms, 10.7M index entries
Normalization
å/ä/ö + ASCII equivalents (ae→ä, oe→ö, aa→å)
Stop Words
Multi-phase aggressive removal (conversational fluff → particles)
GS

GeoSveaRC — Geo-Swedish Regional Context

Municipality mapping, .se TLD priority, geo boost

GeoSveaRC gives Klar an understanding of Swedish geography that no global engine can match. All 290 municipalities and 21 counties are mapped with coordinates, populations, and administrative relationships. Results from the user's detected municipality get a +15 score boost. .se and .nu domains receive inherent weight over foreign TLDs.

Municipalities
290 complete with coordinates, county, seat
Counties
21 with seat coordinates and populations
Geo Boost
+15 score for results matching detected municipality
TLD Priority
.se / .nu weighted higher than foreign TLDs
ST

SveaTrust — Authority & Source Validation

Domain tiering, verified badging, credibility scoring

While Google relies on backlinks (popularity), SveaTrust relies on officiality (trust). Domains are categorized into three tiers: Tier 1 (gov.se, riksdagen.se) gets a critical authority boost; Tier 2 (1177.se, major universities) gets high weight; Tier 3 (general .se) gets standard scoring. Every search result includes a verified boolean and a credibility score (0.0–1.0).

Tier 1
.gov.se, .mil.se, .riksdagen.se — critical authority boost
Tier 2
Official .se (news, healthcare, universities) — high weight
Tier 3
General .se and .nu domains — standard scoring
Verified Badging
60 entity domains with verified flag + credibility scores
OM

OmbudSemantic — Intent Mapping

Conversational parsing, entity extraction, answer boxes

OmbudSemantic acts as the user's proxy to translate long conversational queries into precise search intent. It aggressively strips conversational fluff ("jag skulle vilja veta", "kan du berätta"), identifies core entities (products, places, agencies), cross-references with GeoSveaRC and SveaTrust, and matches against 126 curated Q&A entries plus 85+ NLU patterns for instant answer boxes.

NLU Patterns
85+ across 10+ intent types (DEFINITION, PROCEDURE, COMPARISON, etc.)
Answer Data
126 curated Q&A entries + WikiAnswer for any Wikipedia topic
Entity Extraction
60 entities across 11 categories with word-boundary regex
WikiAnswer
Section extraction for PROCEDURE queries, 24h cache, 63/63 tests

Three-layer HSSE design

The Hybrid Sovereign Search Engine (HSSE) architecture separates data acquisition, intelligence, and delivery into independent layers.

L1

Data Acquisition — The Gatherer

Crawler + Crawler State Store

Discovers, fetches, and monitors the Swedish web (.se, .nu domains). Supports intra-domain parallelism (1–25 workers per domain), force re-crawl mode, sitemap discovery, and incremental dedup. Currently covers 197 domains with ~59,317 crawled documents.

197 domains 59K+ documents Thread-safe domain workers --force mode
L2

Intelligence & Storage — The Brain

SQLite Instant-Index + KDB Segments + Knowledge Graph

Every crawled page is NLP-processed immediately (stemming, compound splitting, entity detection, quality scoring) and written to SQLite tables — queryable at crawl time with no index rebuild. KDB layer builds binary search segments for production-grade BM25 search. Swedish Knowledge Graph provides 290 municipalities, 21 counties, and government agency context.

Per-page NLP Incremental IDF FTS5 + BM25 KDB merge daemon
L3

Retrieval & Ranking — The Delivery

Ranking Pipeline + API + Web UI

Multi-factor ranking: BM25 (25%) + domain authority (15%) + freshness (15%) + content density (10%) + structure (10%) + Swedishness (5%) + entity boost + geo boost. Domain diversity cap (max 3 per domain), foreign language demotion, and "Menade du?" spell correction. Flask API serves search results, answer boxes, autocomplete, and geo endpoints. Web UI features 5 map modes, dark/light theme, and responsive design.

7-factor ranking Verified badging Credibility scores 5 map modes

Platform statistics

Real metrics from the Klar search engine as of Delta stage.

59,317
Documents Crawled
197
Domains Covered
37,658
Pages in SQLite
1.4M
Unique Terms
10.7M
Index Entries
290
Municipalities
60
Entities
126
Curated Q&A
85+
NLU Patterns
95.7%
Pre-Alpha Tests
100%
Delta Checks
15+
Python Modules

Everything a modern search should be

From instant answer boxes to interactive maps, Klar delivers a complete search experience built around Swedish needs.

Instant Indexing

Pages are NLP-processed and queryable at crawl time. No batch rebuild — the SQLite instant-index engine processes each page immediately through the full Swedish pipeline.

💬

WikiAnswer NLU

85+ natural language patterns extract instant answers from Swedish Wikipedia. Section extraction for PROCEDURE queries. ENTITY guard prevents false positives. 63/63 tests pass.

🗺️

Smart Maps

5 map modes: directions, nearby, weather, explore, location. OSRM turn-by-turn routing in Swedish. Nominatim geocoding. Leaflet/OSM — free, no API key needed.

🛡️

Verified Badging

Every search result includes verified (bool) and credibility (0.0–1.0) fields. Official Swedish sources get SveaTrust badges. Green checkmark for verified domains.

🔍

Smart Autocomplete

/api/suggest endpoint with keyboard navigation. Dynamic suggestions from navigational map, answer data, and index vocabulary. "Menade du?" spell correction always on.

🔤

Compound Splitting

Three strategies (longest-first, recursive, lookup) decompose Swedish compounds in real time. öppettider → öppet + tider. 1.4M unique terms in the index.

Klar vs Google — in Sweden

How Klar compares to Google.se across the dimensions that matter for Swedish users.

Dimension Google.se Klar
Swedish language understanding Generic multilingual NLP Native compounds + grammar
Compound word handling Probabilistic guesses 3-strategy lexicon decomposition
Privacy (no tracking) Tracks, profiles, sells data Zero tracking, GDPR-native
Swedish-only index Global index buries .se content Swedish-first, .se prioritised
Ads Ad-funded results No ads, no sponsored content
Municipality awareness None 290 municipalities mapped + geo boost
Government source verification None 3-tier SveaTrust + verified badging
Answer boxes in Swedish Generic featured snippets WikiAnswer NLU + 126 curated Q&A
Map integration Google Maps (strong) Leaflet/OSM, 5 modes, OSRM routing
Speed (Sweden) Fast Good — improving (current: Delta)

Who is Klar for?

Klar serves anyone who needs search results that understand Swedish language, institutions, and context.

🇸🇪

Swedish Citizens

Need a search engine that understands Swedish language, institutions, and context — not translated from English.

🏛️

Public Sector & Agencies

Seek verified official content, real-time government updates, and trusted source prioritisation.

🔒

Privacy-Conscious Users

Expect GDPR-native search without tracking, profiling, or data monetisation.

📚

Students & Researchers

Value clarity, accuracy, and trusted sources over algorithmically optimised clickbait.

💼

Businesses & Publishers

Want fair discovery within a Swedish-first ranking model that rewards quality over SEO manipulation.

Current stage: Delta

From Gamma to Official Release — Klar's staged rollout to become Sweden's search engine.

Delta Stage

NLU-driven map intelligence — current
95.7%
Pre-Alpha tests (88/92)
100%
Delta checks (43/43)
  • NLU-aware map intelligence with 5 map modes
  • OSRM routing + Nominatim geocoding
  • 60 entities, 11 categories (retail added)
  • Phase A search quality (retail öppettider)
  • SQLite instant-index engine (37K+ pages)
  • KDB binary segment storage + merge daemon
  • KSD dashboard — SSE, resource monitor, import controls
  • 3,725 knowledge graph entries (municipalities + counties)
  • All 15+ Python modules compile clean

Full Release Pipeline

Progress through each milestone
Gamma
Complete
Pre-Alpha
Complete
Alpha
Complete
Pre-Delta
Complete
Delta
Current
Pre-Beta
Next
Beta
Upcoming
Official Release
Target
Next: Pre-Beta — Misinformation detection, expanded coverage, personalisation

Frequently asked questions

Everything you need to know about Klar Search Engine.

What is Klar Search Engine?

Klar is a Swedish-first, server-based search engine rebuilt from scratch — not a fork of an existing engine. It indexes 59K+ Swedish pages using 4 proprietary KCA algorithms (LI-Flow, GeoSveaRC, SveaTrust, OmbudSemantic) for native Swedish language understanding, geo context, trust verification, and intent parsing. No ads, no tracking, GDPR-native.

How is Klar different from Google?

Google treats Swedish as English translated through ML models. Klar understands Swedish natively — compound words, grammar patterns, 290 municipalities, Swedish government agencies, and cultural context. Klar also has zero tracking, no ads, verified source badging, and a .se-first ranking model.

Is Klar a browser extension or download?

Neither. Klar is a server-based search engine with its own infrastructure: a web crawler, SQLite/KDB indexing engines, a Flask API server, and a single-page web frontend. You access it through a web browser at its server address (port 5000). It is not a browser, extension, or download.

Is Klar free?

Yes. Klar is completely free to use with no ads, no premium tiers, and no tracking. There is no paid plan now or planned for the future.

Does Klar respect my privacy?

Absolutely. Klar is privacy-first by design: no user profiling, no cross-session tracking, no behavioral data collection, no data sold to third parties. Full GDPR compliance is baked into the architecture.

What are the KCA algorithms?

Klar Custom Algorithms (KCA) are four proprietary engines: LI-Flow (linguistic intelligence — compound splitting, stemming, normalization), GeoSveaRC (regional context — 290 municipalities, .se priority, geo boost), SveaTrust (authority validation — 3-tier domain system, verified badging, credibility scoring), and OmbudSemantic (intent mapping — conversational parsing, entity extraction, answer boxes).

What stage is Klar in?

Klar is currently in Delta stage — the fifth milestone on the path to official release. Delta achieved 100% on 43 verification checks and introduced NLU-driven map intelligence with 5 map modes, OSRM routing, and retail entity support. Next stage is Pre-Beta, focusing on misinformation detection, expanded crawl coverage, and personalisation features.

How many pages does Klar index?

Klar has crawled approximately 59,317 documents from 197 Swedish domains. Of those, 37,658 pages are indexed in SQLite with 1.4 million unique terms and 10.7 million index entries. The remaining documents will be imported through the incremental pipeline as the crawler continues to run.

Built for Sweden's Future Search

Klar is more than a product — it's a platform reinvented from scratch for Swedish search and public trust. Follow our progress as we build the search engine Sweden deserves.

Back to Top