In the DeepSeek Harness plugin ecosystem, for CST Studio Suite automation and simulation tasks, developers often face risks such as uncontrolled execution, difficulty resuming from checkpoints, and lack of budget control. The CST_AI plugin provides a safe and recoverable solution in DSH workflows through declarative task descriptions, pre-execution validation, and budget control.

Plugin Positioning

This is a workflow plugin maintained by aomk20201110-commits, focused on safe automation and analysis for CST Studio Suite. It is released under the MIT License and is currently open source on GitHub.

Core Features

The plugin provides complete task lifecycle management capabilities:

  • Five public operations: supports plan (planning), run (execution), status (status), inspect (inspection), and resume (resumption).
  • Solve policy and budget: supports solve_policy (such as FORBID / ALLOW_ONE / ALLOW_UP_TO_N / REUSE_ONLY) and budget limitations to prevent unauthorized solving.
  • Durable run storage: supports checkpoint resumption through atomic checkpoints and an append-only event log.
  • Parameter scanning: supports defining scan points in spec.points[], with each point independent but sharing the budget.
  • Structured errors and validation: provides machine-readable error classification; the plan stage has zero CST contact, validating before execution.
  • Result reuse: supports signature and artifact hash verification, enabling results to be reused without re-solving.

Installation and Enablement

Install the plugin using the DeepSeek Harness command line:

dsh plugin --profile web add github:aomk20201110-commits/CST_AI#<40-char-commit>

Typical Usage

  1. Declare tasks using TaskSpec JSON.
  2. Run plan first to verify zero CST contact and estimate budget.
  3. Use run to execute only the side effects listed in the plan.
  4. Use status or inspect to query run state from a new process.
  5. Use resume to continue after interruption.
  6. You can also use cst_cli.py directly as the command-line entry point.

Environment Requirements

  • Operating system: Windows only.
  • Core environment: CST Studio Suite 2026 and its built-in Python 3.12.x interpreter.
  • Runtime: PowerShell 5.1 or later.
  • Optional dependency: If using the DSH tool adapter, Node.js 18+ must be installed.

Notes

  • No Python package dependencies: The plugin uses only the standard library and does not require installing any external packages with pip.
  • Handling missing data: If data is unavailable, it returns null and a validity code instead of fabricating values.

Summary

CST_AI provides DeepSeek Harness with controlled and recoverable CST simulation workflow capabilities. For more details, see the community directory or the project homepage.