Skip to content

v0.6.1

Choose a tag to compare

View all tags
@Zhonghao1995 Zhonghao1995 released this 12 May 04:10
· 237 commits to main since this release
v0.6.1
16a2a44

Agentic SWMM v0.6.1 （Alpha）

This release hardens the aiswmm runtime introduced in v0.6.0. The focus is a cleaner local agent experience: fewer confusing folders, fewer demo crashes, faster MCP startup after caching, and a smoother run-to-plot continuation flow.

Install

macOS and Linux:

curl -fsSL https://aiswmm.com/install.sh | bash

Windows PowerShell:

irm https://aiswmm.com/install.ps1 | iex

For a pinned install:

curl -fsSL https://aiswmm.com/install.sh | AISWMM_INSTALL_REF=v0.6.1 bash
$env:AISWMM_INSTALL_REF = "v0.6.1"
irm https://aiswmm.com/install.ps1 | iex

Highlights

  • The interactive runtime now consistently presents itself as aiswmm, including the terminal prefix and executor banner.
  • Interactive output folders are flatter and easier to inspect, using date-and-case paths such as runs/YYYY-MM-DD/HHMMSS_tecnopolo_run.
  • The Tecnopolo prepared .inp demo can be run, audited, inspected for plot options, and continued into plotting from the same active run folder.
  • Follow-up plot selections such as Total_inflow J2 MACAO_94_23 now use the previous run directory instead of restarting input selection.
  • MCP tool schema discovery is cached under ~/.aiswmm/mcp_schema_cache, with timeout protection so slow MCP servers do not block the audited CLI path.
  • Recursive search patterns generated by the planner, such as **.inp, are normalized to valid patterns such as **/*.inp.

What users can try

Launch aiswmm and ask:

How many inp file you have？

Then run the prepared demo:

examples/tecnopolo/tecnopolo_r1_199401.inp run it

After the run finishes, choose a plot:

Total_inflow J2 MACAO_94_23

The expected evidence folder contains the run-local SWMM files, audit note, final report, and generated plot artifacts.

Evidence boundary

v0.6.1 is a runtime-quality release. It verifies the prepared-INP run, audit, plot-selection, previous-run continuation, MCP schema-cache, and search-normalization paths.

Sensitivity and uncertainty skills are present, but fully natural-language batch uncertainty requests still need additional routing and configuration generation before they should be claimed as complete end-to-end automation.

Assets 4
Loading