# flinq

> flinq is the embedding API for the European construction industry. A cross-modal Vertical Foundation Model maps heterogeneous construction data (GAEB bills of quantities, IFC/BIM models, Ökobaudat material data, BKI/STLB cost references) into one shared vector space, exposed as an HTTP API. The document types it serves (bills of quantities, tenders, technical specifications) exist across Europe under different national formats; flinq started in and is currently strongest in Germany, with deepest support today for German formats such as GAEB. flinq is an API service, not a SaaS product: LV matching, early-phase cost prediction and RAG are examples you build on it. Available as a cloud API, optionally on-premise, GDPR-compliant and EU-hosted.

## Site
- [flinq · die Embedding-API für die Baubranche (DE, canonical)](https://flinq.ai/): German homepage.
- [flinq · the embedding API for the construction industry (EN)](https://flinq.ai/en/): English homepage.

## What flinq reads
- [What flinq reads](https://flinq.ai/#produkt): one cross-modal embedding model reads different kinds of construction data, each in its common file formats: bills of quantities (GAEB/PDF), BIM models (IFC), site photos (JPG/PNG), manufacturer products (CSV/PDF), cost data (CSV; Ökobaudat, BKI/STLB cost references) and standards (PDF). It maps them all into one shared vector space; semantically similar objects sit close together regardless of data type, format and source.
- [Using the API](https://flinq.ai/#api-nutzen): conceptual HTTP request and response for the embeddings endpoint. Runnable examples live in the docs.

## Why an embedding model (vs a generative LLM)
- [Why an embedding model](https://flinq.ai/#vorteil): advantages of a specialised embedding model over a generative LLM for search, matching and RAG. Up to 100x cheaper for the same task (embed once, then pure vector math, no model call per query); real-time answers (milliseconds vs seconds); reproducible, auditable results; no hallucination (retrieves real, existing objects); higher domain accuracy so less rework (BauSatz nDCG 0.78 vs 0.60); one integration powers many features (search, matching, RAG, dedup, clustering); cross-modal linking across formats; and EU/GDPR/on-prem so data stays in your environment. The cost comparison is versus running a generative LLM for the same task and depends on the model and query volume.

## Proof
- [BauSatz benchmark](https://flinq.ai/docs/benchmark): flinq small (0.6B) versus OpenAI, Qwen3-8B, Mistral, Gemma2 and BGE-M3 (zero-shot). Leads on six of eight tasks, including product retrieval (nDCG 0.78 vs 0.60) and IFC name matching (71.2 vs 49.7). Full, uncropped table on the page.

## Use cases (built on the API, not products flinq ships)
- [LV matching](https://flinq.ai/#anwendungen): match bill-of-quantities line items across projects and catalogues, even when wording differs.
- [Early-phase cost prediction](https://flinq.ai/#anwendungen): derive cost figures from semantically similar, already costed line items and reference projects.
- [RAG & knowledge management over construction data](https://flinq.ai/#anwendungen): retrieval-augmented generation for company-wide knowledge management (Wissensmanagement) over standards, bills of quantities and project documents.
- [More applications, filterable by customer type](https://flinq.ai/#anwendungen): further applications on the same embeddings, tagged by role — planners/architects (Planer), trades (Gewerke), construction firms (Bauunternehmen), wholesalers (Großhändler), manufacturers (Hersteller). Examples: surface the relevant DIN standards from your own collection of standards, product search and comparison, quantities and costs from the BIM/IFC model, bid comparison (Preisspiegel), calculation and calculation checking, image-to-product-or-position, equivalent alternatives on supply shortages, ETIM/eCl@ss classification, product-catalogue consolidation, visibility in tenders. Some cross-modal use cases (BIM/IFC quantities, image-to-product, ETIM/eCl@ss) are still in development. Full breakdown by customer type in llms-full.txt.

## Deployment
- [Deployment and trust](https://flinq.ai/#betrieb): available as a cloud API or on-premise; GDPR-compliant; hosted in the EU.

## Developers
- [Documentation](https://flinq.ai/docs/): developer docs with runnable examples for LV matching, RAG and cost prediction. (URL is a placeholder until launch.)
- [API reference](https://flinq.ai/docs/api): endpoints and parameters. (Placeholder.)
- [Console](https://console.flinq.ai/): the flinq console — log in, register and access the application.

## Company
- [Team](https://flinq.ai/#team): founders and backgrounds.
- [FAQ](https://flinq.ai/#faq): what flinq is, supported data formats, access, deployment, GDPR.
- [Contact](mailto:info@flinq.ai): flinq TGU - TTI GmbH, Nobelstraße 15, 70569 Stuttgart, Germany.