Skip to content

v0.6.0

Choose a tag to compare

View all tags
@github-actions github-actions released this 11 May 12:58
· 267 commits to main since this release
v0.6.0
da8c1f2

Agentic SWMM v0.6.0

This is a happy and important release for Agentic SWMM: the aiswmm runtime now works as an interactive agent entrypoint, and the one-line installer path can run the Tecnopolo prepared .inp baseline through execution, QA, plotting, and audit.

The milestone is simple but meaningful: after installation, a user can open aiswmm, describe a SWMM task in natural language, and receive auditable local artifacts instead of only a chat answer.

Install

macOS and Linux:

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

Windows PowerShell:

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

Python package users can also install the CLI directly:

pip install aiswmm
aiswmm setup --provider openai --model gpt-5.5

Start the runtime

Launch the interactive runtime:

aiswmm

Good first prompts to type after startup:

inspect available skills, then tell me what local tools you can call

run tecnopolo_r1_199401.inp, audit it, and plot node OU2

The same workflow can also be run as a one-shot command:

aiswmm "run tecnopolo_r1_199401.inp, audit it, and plot node OU2"

What was verified

The v0.6.0 baseline test confirms that the runtime can:

  • start the aiswmm CLI and expose the interactive agent command surface;
  • inspect the local skill registry;
  • expose local tools for doctor checks, skill reads, SWMM execution, plotting, auditing, and file reads;
  • resolve the Tecnopolo prepared INP from examples/;
  • run SWMM and produce runner artifacts;
  • create QA, plot, provenance, comparison, experiment-note, command-trace, and final-report artifacts.

Expected evidence is written under runs/agent/interactive/..., with files such as:

05_runner/model.rpt
05_runner/model.out
05_runner/manifest.json
06_qa/
03_plots/
08_audit/
experiment_provenance.json
comparison.json
experiment_note.md
command_trace.json
final_report.md

Baseline screenshots

Agentic SWMM v0.6.0 runtime smoke test showing aiswmm help and mocked agent execution

Agentic SWMM v0.6.0 skill and local tool inspection output

Current limitations

This release is still an early runtime milestone. The interactive planner currently supports OpenAI API-key CLI calls. Prompt routing still needs refinement, and the modeling-memory loop still needs more hardening before it should be treated as a fully automatic learning loop.

The important result is that the skill layer and local tool layer are successfully exposed, and the runtime can follow the prescribed run, QA, plot, and audit workflow with reproducible evidence.

API key safety

Configure your OpenAI key during setup or through an environment variable such as OPENAI_API_KEY. Do not paste API keys into the aiswmm conversation. If a key is exposed in a prompt, revoke or rotate it immediately.

Assets 4
Loading