Installation

pip install sweep-agent          # the agent + the sweep solver

…or as part of the whole umbrella:

pip install sweepx               # sweep-solver + sweep-agent (+ future companions)

Either path installs the sweep-agent CLI, the tool registry, and the core solver (sweep). Natural-language forward modelling and shot gathers work out of the box — acoustic and elastic (vp/vs/rho), including bundled benchmark models (Marmousi, Overthrust). Python 3.9+.

Extras:

pip install "sweep-agent[ui]"       # Gradio web UI
pip install "sweep-agent[vllm]"     # vLLM backend (GPU)
pip install "sweep-agent[animate]"  # GIF export

You also need a local LLM

Chat needs any OpenAI-compatible endpoint:

• Ollama (Mac / CPU): ollama serve, then ollama pull qwen2.5:14b. Run sweep-agent chat --model qwen2.5:14b (auto-detected on :11434). Tool-calling needs a capable model — the default Q4 quants are noticeably weaker; prefer a -q8_0 tag, a larger model (qwen2.5:32b), or a full-precision endpoint.
• vLLM (GPU node): pip install "sweep-agent[vllm]", then sweep-agent serve-llm --model qwen2.5-14b-instruct. Full-precision — the most reliable for tool-calling.

Full FWI / LSRTM additionally needs sweep-tasks (the production runner) — not on PyPI yet, install from source. Forward modelling and inspection tools don't need it; an FWI tool called without it returns a clean {"error": "sweep_tasks is not importable"}.

macOS (Apple Silicon)

Runs end-to-end on M-series with MPS (CPU 26.7 s → MPS 5.5 s on a 256×384 / 8-shot / 1500-step demo). Use Ollama for the LLM; run_forward_sweep defaults to device="auto" (MPS → CPU). Do not set SWEEP_BUILD_CUDA — that's the Linux + NVIDIA path; macOS uses sweep's eager torch.