Examples¶
Prompts that work on a plain install (pip install sweepx + a local LLM) —
nothing to download:
load the Marmousi benchmark model and run a forward — show the shot gather
here is vp_init.npy — run a 2-D acoustic forward and show me the shot gather
make an elastic model (vp/vs/rho) and run an Elastic forward
list the equations you can model
inspect obs.segy and tell me its geometry # needs sweepx[segy]
The web UI has one-click demo chips for similar requests:
The agent can also build and run FWI / LSRTM tasks through sweepx.tasks from
plain language (e.g. "run a 2-D acoustic FWI on Marmousi, 30 epochs, Adam lr 25").
examples/agent/demos/ has a script per workflow (forward + gather, elastic and
anisotropic wavefields, topography, FWI with Adam or L-BFGS, …), each sending one
natural-language request to a running LLM server; outputs/ holds what they
produce.