RDTvlokip Search
Featured- Search Engine
- Privacy
- Transparency
- Web
- From-Scratch
Independent search engine focused on privacy and transparency. From-scratch crawler (RDTvlokipBot), adaptive BM25 + PageRank algorithm (24 scoring signals), 5,021,846+ indexed pages, hosted in Germany — zero ad trackers.
Key figures
- 5,021,846+
- Indexed pages
- 24
- Ranking signals
- 321,717+
- Unique domains
- Allemagne 🇪🇺
- Hosting
Why I built this
The gap left by Qwant
Qwant uses Bing under the hood. It's not an independent engine — it's a Microsoft interface with a French flag.
From-scratch philosophy
A real crawler, a real index, a real algorithm. Relying on Bing for an "independent" engine is a contradiction.
RDTvlokipBot, a crawler that respects the web
Full compliance with robots.txt (RFC 9309) and the Content-Signal: search=no extension. Adaptive per-domain delay, circuit breaker after 5 consecutive failures, exponential retry backoff, quality filter (≥50 words, deduplication, canonical respect). Transparent, reverse-DNS-verifiable User-Agent — never impersonates a browser.
24 signals, a real ranking algorithm
Adaptive BM25 whose weights shift with query intent (navigational, informational, mixed), PageRank computed via sparse matrices (scipy), French synonyms (~200 groups) and NumPy-vectorized stemming — plus an in-house BERT-like model (~20M parameters, PyTorch) trained to disambiguate ambiguous queries, like "python" (language or animal).
Concrete digital sovereignty
70-80% of European cloud controlled by AWS/Azure/Google. Digital dependence is a real vulnerability — RDTvlokip Search is my concrete response.
Much more than a search engine
Time, geographic, and anti-tracking filters (82 signatures detected), advanced search operators (site:, intitle:, inurl:, filetype:), URL inspection page, interactive domain graph, RSS aggregator across 66 French-language sources, encrypted @rdtvlokip.fr mail, encrypted text editor, proof-of-no-tracking page: transparency and privacy are baked into every feature.
The challenge
5,021,846+ pages indexed alone, at 20, on a Hetzner VPS + GTX 1080 Ti. Proving it's achievable without a team of 50 or millions of euros.
Availability
Live availability is loading…
Full status page