# AI use cases for Trustworthy Agent Simulation (apromisedland/trustworthy-agent-simulation)

Repo: [apromisedland/trustworthy-agent-simulation](https://github.com/apromisedland/trustworthy-agent-simulation) | Use cases: 5 | Page: https://dedupedai.com/use-cases/apromisedland-trustworthy-agent-simulation/

An auditable, configurable LLM multi-agent town simulation built on AgentScope and Mesa, with replay, policy experiments and a local Streamlit dashboard.

This file holds every use case of this repo. Each one starts with a "## " heading and ends with its URL.

- Run the default town and inspect it in the dashboard: https://dedupedai.com/use-cases/apromisedland-trustworthy-agent-simulation/run-town-and-inspect-dashboard/
- Replay a saved run at a chosen tick: https://dedupedai.com/use-cases/apromisedland-trustworthy-agent-simulation/replay-past-run-at-tick/
- Compare policy interventions across shared seeds: https://dedupedai.com/use-cases/apromisedland-trustworthy-agent-simulation/compare-policies-across-seeds/
- Run the public-goods scenario plugin: https://dedupedai.com/use-cases/apromisedland-trustworthy-agent-simulation/run-public-goods-scenario-plugin/
- Run the smoke scenario with a real model and budget cap: https://dedupedai.com/use-cases/apromisedland-trustworthy-agent-simulation/connect-real-model-with-budget-cap/

## How to run the default town and inspect it in the dashboard with Trustworthy Agent Simulation

Repo: [apromisedland/trustworthy-agent-simulation](https://github.com/apromisedland/trustworthy-agent-simulation) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README

Run the 18-resident town offline and browse its map, contracts, cash and event audit.

**Who it is for:** A researcher or developer who wants to see how the default synthetic town behaves before changing anything.

**The problem:** Reading simulation logs by hand makes it hard to see locations, contracts, cash flow and who knew what.

**The outcome:** A committed offline run in runs/demo that you can browse in the local dashboard, with a location map, relationship network, agent state, contracts, cash distribution, economic indicators, event audit and timeline replay.

### Steps

1. Run `python3.12 -m venv .venv`
2. Run `.venv/bin/python -m pip install uv==0.12.20`
3. Run `.venv/bin/uv sync --frozen --extra dev --inexact`
4. Run `.venv/bin/python -m tass.cli run --config configs/town.yaml --output runs/demo`
5. Run `.venv/bin/python -m tass.cli dashboard`, then open http://127.0.0.1:8501 and check the map, contracts and event audit

### Prompt to paste into your agent

```text
In the apromisedland/trustworthy-agent-simulation repo, set up the environment as the README Linux section describes, then run configs/town.yaml into runs/demo. Report the number of ticks committed and the path of the output folder. Verify that runs/demo contains run.sqlite, events.jsonl, metrics.csv and run.json.
```

### Caveat

The default town is a synthetic research environment and its output does not show real-world economic or policy effects.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/apromisedland-trustworthy-agent-simulation/run-town-and-inspect-dashboard/

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

Repo: [apromisedland/trustworthy-agent-simulation](https://github.com/apromisedland/trustworthy-agent-simulation) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README

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

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

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

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

### 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 to paste into your agent

```text
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.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/apromisedland-trustworthy-agent-simulation/replay-past-run-at-tick/

## How to compare policy interventions across shared seeds with Trustworthy Agent Simulation

Repo: [apromisedland/trustworthy-agent-simulation](https://github.com/apromisedland/trustworthy-agent-simulation) | Difficulty: Advanced | Time: 45 minutes | Basis: From the README

Run the policy suite on fixed seeds and read means, errors and 95% intervals.

**Who it is for:** A policy researcher who wants to test whether a subsidy or source disclosure changes town outcomes, and needs to know if the difference is more than noise.

**The problem:** One simulation run cannot show whether a policy effect is stable, and single numbers hide the spread across runs.

**The outcome:** A batch output with runs.csv, summary.json and paired_effects.json covering no intervention, subsidy, source disclosure and their combination, using seeds 11, 22 and 33.

### Steps

1. Complete the install steps from the README Linux section so `.venv` exists
2. Run `.venv/bin/python -m tass.cli batch --config configs/town.yaml --output runs/policies --suite policy --seeds 11,22,33` (inferred, check the README)
3. Open runs.csv in the batch output for one row per independent run
4. Read summary.json for means and standard errors, and paired_effects.json for Student-t 95% intervals

### Prompt to paste into your agent

```text
In the apromisedland/trustworthy-agent-simulation repo, run the policy batch suite with seeds 11,22,33 into runs/policies. Report the mean and 95% interval for each condition from the batch outputs. Verify that a single-seed result is not presented as an interval, since the README says one run does not yield one.
```

### Caveat

The README says baseline policies need not reproduce real model behavior, so these results are not evidence of real policy validity.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/apromisedland-trustworthy-agent-simulation/compare-policies-across-seeds/

## How to run the public-goods scenario plugin with Trustworthy Agent Simulation

Repo: [apromisedland/trustworthy-agent-simulation](https://github.com/apromisedland/trustworthy-agent-simulation) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README

Run the second plugin scenario to see how a scenario can extend the town without runner changes.

**Who it is for:** A developer who wants to test a different scenario on the same runner before writing their own plugin.

**The problem:** Adding a new scenario to a simulation usually means changing core code, which is risky to audit.

**The outcome:** A run of the public-goods plugin in runs/public-goods, viewable in the dashboard alongside the default town.

### 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/public_goods.yaml --output runs/public-goods` (inferred, check the README)
3. Run `.venv/bin/python -m tass.cli dashboard`, then open http://127.0.0.1:8501 and select the public-goods run

### Prompt to paste into your agent

```text
In the apromisedland/trustworthy-agent-simulation repo, run configs/public_goods.yaml into runs/public-goods using the project interpreter. Report the output folder contents and the final tick. Verify that the run completes without editing any file under src.
```

### Caveat

The README describes the public-goods plugin only as a demonstration of scenario extension, not as a validated model of public goods.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/apromisedland-trustworthy-agent-simulation/run-public-goods-scenario-plugin/

## How to run the smoke scenario with a real model and budget cap with Trustworthy Agent Simulation

Repo: [apromisedland/trustworthy-agent-simulation](https://github.com/apromisedland/trustworthy-agent-simulation) | Difficulty: Advanced | Time: 60 minutes | Basis: From the README

Send real provider requests in LLM mode while staying under the default request and token budget.

**Who it is for:** A researcher who wants to test real model behavior in the town and must keep API spending bounded.

**The problem:** Model-driven runs call a paid provider, so credentials must be set safely and usage must be capped and auditable.

**The outcome:** An LLM-mode smoke run in runs/llm-smoke with recorded model requests and responses, and a budget that stops the run at the last committed world state if exhausted.

### Steps

1. Complete the install steps from the README Linux section so `.venv` exists
2. Run `cp .env.example .env`, then set TASS_API_KEY, TASS_BASE_URL and TASS_MODEL in `.env` (inferred, check the README)
3. Run `.venv/bin/python -m tass.cli run --config configs/smoke.yaml --mode llm --output runs/llm-smoke` (inferred, check the README)
4. Review the recorded requests and budget use in the run database before running anything larger

### Prompt to paste into your agent

```text
In the apromisedland/trustworthy-agent-simulation repo, check that TASS_API_KEY, TASS_BASE_URL and TASS_MODEL are set in .env without printing the key, then ask the user to confirm before running configs/smoke.yaml in --mode llm into runs/llm-smoke. Report the number of requests and tokens used against the default budget of 500 requests, 1,000,000 input tokens and 100,000 output tokens. Verify that the run stops at the last committed state if the budget is exhausted.
```

### Caveat

Cost figures are estimates only, shown when prices are configured, and are not provider billing data.

Risk, credentials: The API key must come from .env or environment variables and should never be pasted into YAML, the dashboard or shared logs.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/apromisedland-trustworthy-agent-simulation/connect-real-model-with-budget-cap/