Back to workbench

ArchSight Solver v1.9.1

Five-minute quickstart

中文快速开始 | English

Choose one path. The Web path needs no installation. The GitHub Release wheel provides the CLI and MCP server. Docker provides the complete Web/API runtime. A source checkout is for development and contribution.

1. Try the workbench

Open solver.archsight.cn, select a built-in template or public benchmark, run the analysis, and inspect diagnostics plus the current result source. From 成果导出 you can download a DOCX/XLSX result or a verification package.

No account is required. The public site is a demonstration environment; save important work as a local .slv project file.

2. Install the CLI and MCP server from a GitHub Release

Requirements: Python >=3.13.

After the workflow completes, download archsight_solver-1.9.1-py3-none-any.whl from the v1.9.1 GitHub Release, then install it into a virtual environment:

python -m venv .venv
# Linux/macOS: source .venv/bin/activate
# Windows PowerShell: .\.venv\Scripts\Activate.ps1
python -m pip install ./archsight_solver-1.9.1-py3-none-any.whl

Create create-request.json (a source checkout also includes examples/verification-package/create-request.json):

{
  "payload": {
    "analysisType": "beam",
    "beamType": "simply_supported",
    "loadType": "uniform",
    "spans": [6],
    "q": 12,
    "E": 206,
    "I": 85000
  },
  "evidence": {
    "source": "five-minute-quickstart"
  }
}

Create and replay a verification package:

archsight-solver-tool verification_package_create --input create-request.json --pretty > created.json
python -c "import json; d=json.load(open('created.json',encoding='utf-8')); json.dump({'package':d['package']},open('verify-request.json','w',encoding='utf-8'),ensure_ascii=False)"
archsight-solver-tool verification_package_verify --input verify-request.json --pretty

Success means the second command returns status: "pass", integrityValid: true, and replayMatched: true.

Start the MCP server with:

archsight-solver-mcp

The installed distribution includes runtime schemas, benchmark data, templates, materials, supports, sections, and MCP documentation. It does not depend on a repository checkout or current working directory.

3. Run the complete Web/API image

Download archsight-solver-v1.9.1.tar.gz and SHA256SUMS from the GitHub Release, verify the archive, and load it:

sha256sum --check SHA256SUMS --ignore-missing
docker load --input archsight-solver-v1.9.1.tar.gz
docker run --rm -p 127.0.0.1:6240:6240 archsight-solver:release

Open http://127.0.0.1:6240. The single image serves the frontend and /api. The public Release archive is the direct container distribution path. Callers with GitHub Packages access may instead authenticate and use the immutable workflow image:

docker login ghcr.io
docker run --rm -p 127.0.0.1:6240:6240 ghcr.io/archsightlabs/archsight-solver:v1.9.1

Do not use latest as release evidence. Official production deployment uses its independently managed Alibaba Cloud Container Registry target rather than GHCR.

4. Install the framework-free Host Client

Download archsight-solver-host-client-1.9.1.tgz from the GitHub Release:

npm install ./archsight-solver-host-client-1.9.1.tgz
import { SolverHostClient } from "@archsight/solver-host-client";

const client = new SolverHostClient({
  iframe: document.querySelector("#solver") as HTMLIFrameElement,
  solverOrigin: "https://solver.example.com",
});

The package has zero runtime dependencies and implements Host Protocol 1.0. Exact-origin allowlists, session ID, nonce, parent-window checks, CSP frame-ancestors, save ownership, and read-only behavior remain the host's responsibility. See the Host Client guide.

5. Run from source

Requirements: Python >=3.13, uv, and Node.js >=22.22.0.

git clone https://github.com/ArchSightLabs/archsight-solver.git
cd archsight-solver
uv sync --frozen
npm --prefix frontend ci --include=optional

Start two terminals:

uv run python app.py
npm --prefix frontend run dev

Main local gates:

uv run python -m pytest backend/tests -q
npm --prefix frontend run lint
npm --prefix frontend run test:unit
npm --prefix frontend run build

Next