Approach & technology

No answer without a source. And the source is verifiable.

A language model that sometimes makes things up is worthless for serious work. The systems of this project are built to prevent that — and to make it demonstrably true. This is how the system is put together.

The RAG pipeline

From question to verified answer

Six steps, one principle: the model may only speak about what it can substantiate.

  1. Step 1

    Question

    The user asks a question in natural language.

  2. Step 2

    Hybrid retrieval

    Vector (semantic) and keyword (FTS) search together surface the best candidate sources.

  3. Step 3

    Reranking

    A reranking model puts the truly relevant passages on top — noise drops away.

  4. Step 4

    Generation

    The language model formulates the answer, based solely on the retrieved sources.

  5. Step 5

    Citation verification

    Is every claim actually in the source? If not, it gets cut.

  6. Step 6

    Answer + source card

    The result comes with a clickable reference to the original.

Continuous feedback. An evaluation harness with a fixed set of checked questions (golden set) re-measures quality and the hallucination rate with every change. Regressions become visible before they hit production.

Principles

Four choices that decide everything

Citations by design

Sources aren't bolted on afterwards — they're built into the foundation. An answer without a verifiable source doesn't make the cut.

Data in a fixed place

For each part of the pipeline it is fixed where it runs and why. No promise, but an overview of which part sits where.

Layered AI = low cost

RAG and vector search are nearly free; a local model handles the heavy lifting; an expensive API is only called in where it's truly needed.

Measured, not hoped for

An evaluation harness with a golden set objectively measures quality and hallucination rate. We steer by the numbers, not by feel.

Under the hood

Mature, proven technology

No black-box platform, but a transparent stack kept in our own hands — and one that can be run on your own too.

pgvectorFTS / hybrid retrievalRerankingOllamaLlama · Qwen · GemmaCitation verificationFastAPIAstroCaddyCloudflaren8nschema.org · llms.txt
How we work

From idea to a running system

01

Explore

What is the goal, what data is there, and what can AI already do with it? Map first, then build.

02

Build

A working system — RAG, agent or model — with citations and evaluation built in from day one.

03

Run and measure

Monitored and maintained, with the golden set re-measuring on every change. No pilot that disappears into a drawer.

See it working

View a live system

The case-law portal runs in production: ask a question in plain language and see the source under every answer.