Skip to content

Latest commit

 

History

History

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

README.md

ACE - Sim

Multi-provider LLM simulation chains for validating ideas and reviewing tasks before implementation.

ACE Logo

Gem Version Ruby License: MIT

Works with: Claude Code, Codex CLI, OpenCode, Gemini CLI, pi-agent, and more.

Experimental — its already proof its worth, but need some more work.

Getting Started | Usage Guide | Handbook - Skills, Agents, Templates

ace-sim executes preset-driven simulation chains (sequential step runs per provider) across one or more providers via ace-llm, then optionally synthesizes suggestions and revised source artifacts for follow-up work. Use /as-sim-run to launch simulations from inside a coding agent.

How It Works

  1. Select a simulation preset (validate-idea or validate-task) and provide a source file, with context assembled by ace-bundle.
  2. The simulation engine runs each step (draft, plan, work) sequentially — each step's output feeds as input into the next, building on prior reasoning through ace-llm.
  3. After all steps complete, a final synthesis stage gathers feedback from every stage to propose improvements, surface questions, and produce a revised source artifact — feeding better specs back into ace-task or sharper ideas into ace-idea.

Use Cases

Validate ideas before committing to implementation - run validate-idea to compare model reasoning across providers and stress-test assumptions from a single source file.

Review task specs before coding starts - run validate-task to inspect plan/work outputs across providers and iteration counts, keeping ace-task specs sharp before delivery begins.

Compare provider behavior under the same workflow - use repeated --provider and --repeat options to evaluate consistency and convergence in ace-llm simulation outputs.

Synthesize recommendations from simulation runs - enable --synthesis-workflow and --synthesis-provider to produce actionable suggestions, then feed results into ace-review for follow-up review.

Testing Model

ace-sim follows the restarted fast / feat / e2e contract:

  • ace-test ace-sim - default deterministic fast loop (test/fast).
  • ace-test ace-sim feat - deterministic feature-contract lane (only when test/feat exists).
  • ace-test ace-sim all - run all deterministic lanes.
  • ace-test-e2e ace-sim - scenario workflow validation under test/e2e.

Getting Started | Usage Guide | Handbook - Skills, Agents, Templates | Part of ACE