# AI use cases for Open ETA (OpenMOSS/OpenETA) Repo: [OpenMOSS/OpenETA](https://github.com/OpenMOSS/OpenETA) | Use cases: 5 | Page: https://dedupedai.com/use-cases/openmoss-openeta/ OpenETA is a Python embodied task agent that runs a continuous observe, reason, act, verify loop across simulators and real robots. This file holds every use case of this repo. Each one starts with a "## " heading and ends with its URL. - Check a LIBERO evaluation manifest before spending compute: https://dedupedai.com/use-cases/openmoss-openeta/validate-libero-eval-manifest/ - Run one bounded task through a Planner and the simulator: https://dedupedai.com/use-cases/openmoss-openeta/run-single-bounded-task/ - Watch simulator episodes on a live Dashboard: https://dedupedai.com/use-cases/openmoss-openeta/watch-simulator-live-dashboard/ - Turn an episode log into a replay video: https://dedupedai.com/use-cases/openmoss-openeta/replay-episode-as-video/ - Run LIBERO episodes in parallel with human recovery: https://dedupedai.com/use-cases/openmoss-openeta/run-parallel-libero-evaluation/ ## How to check a LIBERO evaluation manifest before spending compute with Open ETA Repo: [OpenMOSS/OpenETA](https://github.com/OpenMOSS/OpenETA) | Difficulty: Intermediate | Time: 15 minutes | Basis: From the README Confirm a parallel-evaluation manifest is well formed without calling a model or MCP service. **Who it is for:** A robotics researcher preparing a LIBERO benchmark run who wants to catch manifest errors before any model calls or simulator time. **The problem:** A bad manifest can waste a long parallel run, and it is hard to tell whether the file is valid without starting the agent stack. **The outcome:** The manifest passes validation, so the researcher knows the episode list is usable before connecting a Planner. ### Steps 1. Run `uv sync --extra dev` in the Open ETA repository folder. 2. Run `uv run openeta-batch --manifest examples/parallel_libero_eval.json --validate-only`. (inferred, check the README) 3. Read the validation output and fix any reported manifest problem before the next step. ### Prompt to paste into your agent ```text Use the OpenETA repo to validate the manifest at with `uv run openeta-batch --manifest --validate-only`. Report whether validation passed and list any errors, without connecting to a model or MCP service. ``` ### Caveat The README says validation runs without a model or MCP service, so it confirms the manifest only and does not show that any task succeeds. 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/openmoss-openeta/validate-libero-eval-manifest/ ## How to run one bounded task through a Planner and the simulator with Open ETA Repo: [OpenMOSS/OpenETA](https://github.com/OpenMOSS/OpenETA) | Difficulty: Intermediate | Time: 45 minutes | Basis: From the README Send a single instruction to the Open ETA agent with a fixed turn limit against a LIBERO scene. **Who it is for:** A developer testing a new OpenAI-compatible Planner model who needs a quick end-to-end check of the perception and action loop. **The problem:** Testing a Planner against a physical or simulated task usually means wiring many services together before seeing any behavior. **The outcome:** The agent runs one task for up to a set number of turns, and the developer can see how the loop behaves on that scene. ### Steps 1. Run `uv sync --extra dev` in the Open ETA repository folder. 2. Run `bash scripts/setup_envs.sh libero` to install the LIBERO simulator backend. 3. Run `uv run python -m sim.mcp_server` and keep it running. 4. Run `uv run openeta --once "inspect the current scene" --max-turns 8` after configuring the Planner with `/provider` in the TUI if needed. ### Prompt to paste into your agent ```text In the OpenETA repo, run `uv run openeta --once "" --max-turns 8` with the LIBERO simulator MCP service running. Report the final status and whether the environment reports task completion, and do not treat a finished process as task success. ``` ### Caveat The README says a finished process does not imply task success, so check the trusted environment receipt or checker output rather than the exit alone. Risk, credentials: The Planner endpoint needs its own API access settings, so keep those keys out of logs and shared session files. 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/openmoss-openeta/run-single-bounded-task/ ## How to watch simulator episodes on a live Dashboard with Open ETA Repo: [OpenMOSS/OpenETA](https://github.com/OpenMOSS/OpenETA) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README Start the simulator MCP service and open its Web Dashboard to observe runs as they happen. **Who it is for:** A robotics engineer debugging agent behavior who wants a visual view of the simulator state during a run. **The problem:** Debugging an agent loop from text logs alone makes it hard to see where a grasp or placement went wrong. **The outcome:** The engineer has a live Dashboard in the browser showing the simulator side of the episode. ### Steps 1. Run `uv sync --extra dev` in the Open ETA repository folder. 2. Run `bash scripts/setup_envs.sh libero` to install a simulator backend. 3. Run `uv run python -m sim.mcp_server` to start the MCP service, REST API, and Web Dashboard. 4. Open http://localhost:8765/ in a browser to view the live Dashboard. ### Prompt to paste into your agent ```text In the OpenETA repo, start `uv run python -m sim.mcp_server` and confirm the Dashboard responds at http://localhost:8765/. Report whether the page loads and which simulator backend is installed. ``` ### Caveat The README describes the Dashboard as a live view from the simulator MCP service, and it does not say which simulator backends appear on it beyond the one you install. 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/openmoss-openeta/watch-simulator-live-dashboard/ ## How to turn an episode log into a replay video with Open ETA Repo: [OpenMOSS/OpenETA](https://github.com/OpenMOSS/OpenETA) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README Rebuild the timeline of a finished episode and export it as an MP4 for review. **Who it is for:** A researcher reviewing a failed or uncertain rollout who needs to see what the agent observed and did, step by step. **The problem:** Reading raw episode logs makes it hard to judge whether an action changed the world as the agent claimed. **The outcome:** The researcher gets replay.mp4 showing the reconstructed timeline of the episode. ### Steps 1. Run `uv sync --extra dev` in the Open ETA repository folder. 2. Locate the episode log folder from a previous run, such as an episode directory under outputs/. 3. Run `uv run openeta-replay EPISODE_DIR --output replay.mp4` with EPISODE_DIR replaced by that folder. ### Prompt to paste into your agent ```text In the OpenETA repo, run `uv run openeta-replay --output replay.mp4` and report the output path and whether the file was created. Verify the timeline matches the episode's recorded turns. ``` ### Caveat The README says replay rebuilds the timeline from episode logs, so it only shows what those logs recorded and cannot add missing evidence. 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/openmoss-openeta/replay-episode-as-video/ ## How to run LIBERO episodes in parallel with human recovery with Open ETA Repo: [OpenMOSS/OpenETA](https://github.com/OpenMOSS/OpenETA) | Difficulty: Advanced | Time: 120 minutes | Basis: From the README Execute a LIBERO manifest across ten concurrent isolated episodes and save results to JSON. **Who it is for:** A research team measuring a Planner on the LIBERO task suite who needs many episodes finished in a reasonable time with isolated runs. **The problem:** Running LIBERO tasks one at a time is slow, and parallel runs can interfere with each other if episodes are not isolated. **The outcome:** The team gets outputs/parallel-libero.json with per-episode results from isolated parallel runs. ### Steps 1. Run `uv sync --extra dev` and `bash scripts/setup_envs.sh libero` in the Open ETA repository folder. 2. Run `uv run python -m sim.mcp_server` and keep it running. 3. Run `uv run openeta-batch --manifest examples/parallel_libero_eval.json --concurrency 10 --approvement reviewed_autonomy --output outputs/parallel-libero.json`. (inferred, check the README) 4. Review the output JSON and check each episode's trusted environment receipt before reporting a success rate. ### Prompt to paste into your agent ```text In the OpenETA repo, run the parallel LIBERO evaluation with `uv run openeta-batch --manifest examples/parallel_libero_eval.json --concurrency 10 --approvement reviewed_autonomy --output outputs/parallel-libero.json`. Report the number of episodes completed and the success count backed by environment receipts, and flag any episode without a receipt. ``` ### Caveat The README says process completion does not imply task success, so report success only from environment receipts, rewards, checkers, or retained rollout evidence. Risk, credentials: Ten concurrent episodes each call the Planner endpoint, so check that the API key and rate limits cover that load before starting. 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/openmoss-openeta/run-parallel-libero-evaluation/