DEDUPED Hot AI repos, turned into use cases Get the Monday 5

How to replay a saved run at a chosen tick with Trustworthy Agent Simulation

Load a committed snapshot from a finished run and inspect the world at one tick.

  • Research
  • Intermediate
  • 30 minutes
  • From the README
  • Any agent

Who it is forA reviewer who needs to check what the agents knew and did at a specific point in an earlier simulation.

The problemAuditing a multi-agent run requires seeing the exact world state at one moment, not a summary of the whole run.

The outcomeA replayed view of the world at tick 10 from runs/demo, read only from committed snapshots with no model calls.

What you need first

  • Python 3.12 or 3.13 installed, with access to the python3.12 command on Linux
  • A terminal opened at the repository root
  • Internet access to install locked dependencies with uv
  • For real model runs only: an API key and a Chat Completions compatible endpoint with tool calling

Steps

  1. Complete the install steps from the README Linux section so .venv exists
  2. Run .venv/bin/python -m tass.cli run --config configs/town.yaml --output runs/demo if runs/demo does not exist yet
  3. Run .venv/bin/python -m tass.cli replay runs/demo --tick 10 (inferred, check the README)
  4. Compare with tick -1, which is the initial state, by changing the tick value

Prompt for your agent

In the apromisedland/trustworthy-agent-simulation repo, replay the run at runs/demo for tick 10 using the project interpreter in .venv. Report the state summary for that tick. Verify that the replay command does not call any model or contact the network.

Caveat. Replay reads only committed snapshots, so it cannot show decisions that were never committed.

Use this in your AI agent

Ask your agent

Using the Deduped MCP, get the setup steps for apromisedland/trustworthy-agent-simulation and "Replay a saved run at a chosen tick"

Add Deduped to Claude Code once

claude mcp add --transport http deduped https://dedupedai.com/mcp

Cursor, ChatGPT, others

Share: X LinkedIn Hacker News Email Copy link

Third-party repos we don't own. AI-written use cases can be wrong. Disclaimer