Engine Azure gpt-4o-mini ★ (~15-45s) Azure gpt-4o (best · ~60-180s) GPU (cuttlefish) (local · ~30-90s)

Azure mini is the default and finishes fastest. Azure full is the most thorough but can take 1-3 minutes. GPU keeps everything inside the BNL fleet. Live progress shown in the right-hand reasoning panel.

Corpus slices

Three core legal slices are on by default. Enable Hague Convention, Norwegian Courts, Bufdir guidance, or DBN Resources for more targeted research.

Advanced controls

Sub-questions 4 How many angles the agent expands the question into before retrieval.

Chunks / sub-Q 6 How many corpus chunks the hybrid retriever pulls per sub-question.

Similarity floor 0.30 Minimum cosine similarity for uploaded-doc chunks to count as a match.

Sources kept 12 Top sources kept after dedupe + rerank to feed synthesis.

Temperature 0.15 Synthesis creativity. Keep low for grounded legal briefs.

Drop up to 5 case files here, or browse

PDF, DOCX, TXT — chunked + embedded in memory only, never stored.

Branching from sub-question

Prior brief summary

Add your notes / context

Question or pasted text

Ready

Pick slices, drop a case file or paste a question, then run. The agent will expand the question, retrieve from the corpus + your upload, rerank, and synthesise a cited brief.