Screenshots 1
1 / 4
View all casesAI Application

Searching decades of agricultural advice with RAG

A knowledge base where advisers can query over 100,000 historical documents in seconds.

106,000+Documents indexed
200msAverage retrieval time
745kSearchable text chunks
See the results

15 minutes, no obligation · No sales pressure

ClientDLV Advies
IndustryAgriculture & Consultancy
ProductDLV Kennisbank
Read time2 min read
Searching decades of agricultural advice with RAG

DLV Advies

DLV Advies, a Dutch agricultural consultancy, needed to make an archive of over 100,000 documents accessible to its advisers. Users needed to ask questions in Dutch and receive accurate answers with citations. SevenLab built a custom RAG (Retrieval-Augmented Generation) application to search this archive.

Advisers spent time searching fragmented archives of PDFs, spreadsheets and emails. With over 745,000 text chunks, keyword search alone could not reliably find precedents or compare regional projects. The organisation needed a system that could interpret agricultural questions, search the archive at scale and link answers to source files.

What SevenLab built

SevenLab built a hybrid retrieval platform combining pgvector similarity search with trigram-based keyword matching. Gemini Flash handles reasoning and structured answers, while a custom Bun-based backend processes and ingests documents in real time.

Hybrid search engine

Combines vector similarity and keyword matching to search across 106,000 documents.

Verified citations

Every AI-generated response includes clickable sources that open the original document chunk for immediate verification.

Automated ingestion pipeline

Handles bulk imports from external volumes with AI-driven deduplication and screening.

Agentic reasoning steps

Real-time transparency showing the AI's classification, search strategy, and data synthesis process.

Measurable business impact

106,000+Documents indexed
200msAverage retrieval time
745kSearchable text chunks
100%Verifiable source citations

DLV Kennisbank reduced research time from hours to seconds by bringing the document archive into one searchable interface. Advisers can find and check past project information without searching separate folders.

The ability to query our entire historical archive in natural language has fundamentally changed how our advisors work. We can now find specific precedents and project details that were previously buried in thousands of folders, all while ensuring every answer is backed by our own data.

Head of IT

Share

Trusted by teams at

PO-Raad
Adjust
Aegir Marine
Chemgas
Contakt
DBF
DLV
Farmplus
Het Oranje Kruis
Kleentec
Kwast Wijnkopers
MedicInfo
Moco Museum
NHA
Omgevingsdienst Utrecht
RDW
Rho Adviseurs
Sectorinstituut Transport en Logistiek
Storteboom
Stratego
Verhoeven
Voetbal International
Weekamp Deuren
Woonbedrijf
ZON Fruit & Vegetables

Want results like these?

Tell us your challenge and we'll show you how we'd solve it, with a clear scope, timeline, and fixed price.

Talk directly with our AI specialists

15 minutes, no obligation
No sales pressure
Prototype in 7 days