Command line

Installing ragX adds the ragx command.

ragx ask

Answer a question over a corpus, with a live trace of the agent:

ragx ask "Who founded Ashford University?" --corpus ./docs

Option	Default	Meaning
-c, --corpus	.	file or folder to search
-m, --model	from the environment	provider:model
-v, --variant	ragx-lexical	any variant
--steps	10	forager turns
--think / --no-think	the variant's	thinking mode
--live / --no-live	--live	show the live trace
--json	off	print the answer as JSON (implies --no-live)

--json prints the text, citations, grounded, confidence and usage, which is convenient for scripts:

ragx ask "Who founded Ashford University?" -c ./docs --json | jq .text

ragx search

Retrieval only, with no LLM call: the top passages for a query.

ragx search "Ashford University founder" -c ./docs -v ragx-fused -k 5

ragx ingest

Build the text mirror and, for dense variants, the embeddings ahead of time:

ragx ingest ./docs --variant ragx

ragx serve

Start an OpenAI-compatible server:

ragx serve ./docs --variant ragx-lexical --port 8000

ragx variants

List every variant with its ranker, stages and backfill.

ragx version

Print the installed version (also ragx --version).