mosaic ask

Answers a question from your indexed papers: it retrieves the most relevant ones and asks the configured LLM to answer from them alone, citing them as [n].

Usage

mosaic ask "What causes anomalous electron transport in Hall thrusters?"
mosaic ask "What is still unexplained about the breathing mode?" --mode gaps
mosaic ask "PIC versus fluid models of the discharge" --mode compare --show-sources
mosaic ask "Which codes simulate the plume?" --mode extract --year 2015-2025
mosaic ask "wall effects on electron mobility" -q "hall thruster" -n 20
mosaic ask "anomalous transport" --from refs.bib -o answer.md

Behaviour

Mode	Answer
synthesis (default)	a state-of-the-art summary
gaps	open problems, contradictions, methodological limits
compare	methods, datasets, metrics and results side by side
extract	per paper: task, method, dataset, metric, key result

Retrieval returns rag.top_k papers (default 10; -n overrides it). --query and --from restrict it to the papers matching a query, or listed in a .bib/.csv file; --year filters them by year.

--output writes Markdown (.md: the question, the mode, the answer and the references) or JSON (.json).

Needs the index (mosaic index) and an LLM (mosaic config --llm-provider ...). Exit code 1 for an unknown mode, or when retrieval or the LLM call fails.

See also: tutorial 9, asking your library, AI recipes.

Options

mosaic ask [OPTIONS] QUESTION

Argument	Value	Description
QUESTION	TEXT	Question or topic to analyse

Option	Value	Default	Description
--mode	TEXT	synthesis	synthesis (default), gaps, compare, extract
--query, -q	TEXT		Pre-filter: restrict to papers matching this FTS query
--from	PATH		Pre-filter: restrict to papers from a .bib or .csv file
--year, -y	TEXT		Year or range filter (e.g. 2020-2024)
-n, --top	INTEGER		Override rag.top_k for this query
--output, -o	PATH		Write answer to file (.md or .json)
--show-sources			Print retrieved papers before the answer