Technical details
Asking needs your own API key from an AI provider. It stays in this tab: click API key.
bring your own API key
Assistant steps
The first start takes about 30 s: it loads Python, the engine and its data (Neurosynth, Neurosynth + tracts or NeuroStore, as the selector at the top says), which the browser keeps for the next visit.
Examples and relations
Examples click one to put it in the box
Relations and rules click one to add it at the cursor, ⓘ for its documentation
Commands click one to type it, ⓘ for what it does
Engine and version
The engine here leaves out the precomputed analyses lift, relative_risk, uniformity and ratios: they need several GB, more than a browser tab is given. Add ?engine=full to the address to build them on a machine with the memory; the terminal command has them. The first query using query_similarity downloads the SPECTER2 query model and the embeddings of the studies, about 310 MB.
Neurosynth + tracts (?engine=neurosynth_iit) adds to the studies of Neurosynth the whole-brain tractogram of the IIT Human Brain Atlas v3, 15,000 of its 37,980 streamlines chosen at random and thinned, indexed by the atlas's Desikan–Killiany wmparc (labels of FreeSurfer's colour LUT), and tract_querier's bundles (uf, af, cst…) as Datalog rules. The tractogram and the wmparc are those of the IIT Human Brain Atlas (Illinois Institute of Technology; Qi, Arfanakis et al.), under the IIT Human Brain Template License (non-commercial educational and research use, with attribution); the colour LUT is FreeSurfer's.