
Badges
Activity
Challenge Categories
Challenges Entered
Build an LLM agent for five real-world games
Latest submissions
| Participant | Rating |
|---|
| Participant | Rating |
|---|
-
overfit_small Orak Game Agent Challenge 2025View
Orak Game Agent Challenge
Question Regarding Pre-Creating a Save File for the Competition
3 days agoHello,
I have a question regarding the competition setup.
In the Orak GitHub documentation (Orak/docs/setup_pokemon.md at release ยท krafton-ai/Orak ยท GitHub), it is mentioned:
โRecommended: Before running the full automation, launch the game once and create a save file with your desired player and rival names. This helps avoid early-game menu handling issues.โ
Based on this note, I would like to ask whether, in the context of the competition, it is permitted to pre-create a save file in advance for the purpose of avoiding the name input process when launching the game.
In other words, is it allowed to generate and use a save file beforehand solely to skip the initial player/rival name selection, without affecting gameplay or logic?
Issue: gRPC Session Expired (120s)
About 16 hours agoDuring game execution, an error occurs where the authentication session expires and the server rejects requests when the execution time exceeds 120 seconds.
โฉโฉโฉ
File โ/workspace/orak-2025-starter-kit/.venv/lib/python3.11/site-packages/grpc/_channel.pyโ, line 996, in _end_unary_response_blocking raise _InactiveRpcError(state) # pytype: disable=not-instantiable ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ grpc._channel._InactiveRpcError: <_InactiveRpcError of RPC that terminated with: status = StatusCode.UNAUTHENTICATED details = โSession expired (timeout: 120s)โ debug_error_string = โUNKNOWN:Error received from peer {grpc_status:16, grpc_message:โSession expired (timeout: 120s)โ}โ
โฉโฉโฉ
When a timeout occurs, the code must be resubmitted and executed again. In this process, the game server continues execution from the state (obs), while the agent is reinitialized and loses all existing history and internal state. As a result, the agentโs decision-making flow is disrupted, causing issues in the overall workflow.
We would like to inquire about the following points:
Is it possible to extend the timeout limit for game execution to allow more leeway?
In the hidden test environment, are these execution-time constraints and agent state reinitialization issues taken into consideration in the system design?
We would appreciate your clarification on these matters.
Thank you.