Whimbox: A Python AI Agent That Plays Infinity Nikki for You

What Whimbox Automates and Who It Serves

Whimbox is an AI agent for Infinity Nikki, a game that demands repetitive daily chores. The README lists a long set of automated tasks: mining, material farming, weekly bosses, daily wishes, star collection, battle pass claims, and more. It also handles small conveniences like auto-dialogue, auto-collection, auto-fishing, and skipping cleaning animations. The target user is a player who wants to offload these grinds. This is not a general-purpose tool. It is specific to one game and one resolution. The project acknowledges its inspiration from Genshin Impact assistants like GIA and BetterGI, so it fits the same niche of game-specific automation. Who is it for? A player with a mid-to-high-end PC who plays at 16:9 and is willing to accept the risk that the game's terms of service could cover this usage. The README is blunt about that risk: it says Whimbox does not modify game files or read memory, only screenshots and simulated input, but the user takes all consequences.

The Architecture: Backend RPC, No UI, and a Separate App

Whimbox 2.0 changed the project's shape. The repository you are looking at is now the backend only. It no longer provides a UI; it handles RPC services, large model calls, and tool calls. The actual interface lives in a separate project called whimbox_app. The README is explicit: to run it, you must use the app. The backend is Python-only, and the README says it supports only Python 3.12. The project structure shows a modular layout: an ability module for switching capabilities, an action module for picking up and fighting, an agent_workspace for managing the large model context, an API module for OCR and YOLO models, a map module for minimap recognition and big map operations, and a task module with scripts that the agent can call. This separation suggests a design where the app sends commands over RPC, the backend interprets them with a language model, and the task scripts execute the actual game actions via screenshot and input simulation. The README does not detail the RPC protocol, so you will have to inspect the code for that.

Getting It Running: Commands and Configuration

The README gives a specific path to run Whimbox. You cannot just clone and run this backend alone. The steps are: first, ensure you can run a full automation routine through the whimbox_app; second, in the routine configuration, check the option that says 'start Whimbox and automatically run the routine'; third, launch whimbox_app.exe from the command line. There is also an option to automatically close the game and Whimbox after the routine finishes. The backend itself has no UI, so all configuration happens in the app. For developers, the README only says it supports Python 3.12 and that the author will sort out the project framework later. That is a thin onboarding story. If you want to develop on it, you will need to read the source code directly. The 'future plans' section mentions a floating window for web pages, but that is not yet implemented.

The AI Mechanism: Natural Language, OCR, and Tool Calls

Whimbox's core is an AI agent that uses a large language model and image recognition. The README lists AI dialogue as a feature: you can use natural language to make Whimbox execute the listed functions. It supports 'SKILL' to some degree, and it can be remotely controlled via WeChat. The backend's API module includes OCR and YOLO models, which are the image recognition components. The agent_workspace module manages the context for the large model, meaning it tracks what the model sees and decides. The tool calls are the task scripts in the task module. The flow is likely: the app sends a natural language command, the backend's agent interprets it, maps it to a tool, and executes the tool via simulated mouse and keyboard. The README does not specify which LLM or YOLO models are used, so that remains unverified. The map module handles minimap recognition and big map operations, which is how it navigates the world.

Limitations: Hardware, Resolution, and Platform

The README states three hard limits. First, Whimbox only supports games running at standard 16:9 resolution. If your monitor is ultrawide or your game window is not 16:9, it will not work correctly. Second, image recognition consumes extra performance on top of the game, so only mid-to-high-end PCs are supported. The README says a cloud gaming version is planned, but it is not available now. Third, the project does not support all map types: auto pathfinding currently works only for the open world and the Star Sea, not for other instances. The release notes also reveal a recent bug: version 3.0.3 fixed a map mask issue that prevented login to the bilibili server and the international server. That suggests map-related features are fragile and need frequent patches. The README also warns that the game's terms of service are comprehensive, so using Whimbox could lead to account penalties. These are not minor caveats; they define the tool's usable envelope.

Alternatives: GIA and BetterGI as Precedents

Whimbox does not exist in a vacuum. The README credits two open-source projects: GIA (Genshin Impact Assistant) and BetterGI. Both are for Genshin Impact, not Infinity Nikki. GIA is a Python-based assistant that automates daily tasks and uses computer vision; BetterGI is a C# tool that offers features like auto-pickup and elemental sight. The key difference is the target game. If you play Genshin Impact, those projects are mature and have larger communities. Whimbox is specific to Infinity Nikki, so you cannot swap one for the other. The approach is similar: screenshot-based vision and simulated input, not memory editing. But the implementation is game-specific. The README also mentions openclaw and nanobot as the inspiration for the AI part, but those are not direct alternatives; they are general agent frameworks. For a player of Infinity Nikki, there is no direct alternative mentioned in the material, so Whimbox is the only option listed.

Maintenance, License, and Upgrade Cost

The project is under GPL-3.0, which means any derivative work must also be open source under the same license. That is a legal constraint you should respect if you plan to fork it. The repository shows active maintenance: the latest release 3.0.3 was pushed on 2026-08-28, with 3.0.2 and 3.0.1 in the same month. The release notes are in Chinese and mention bug fixes for map masks and support for a new ability. That indicates the project is being updated to keep up with game changes. The upgrade cost is low if you use the app, because you can just download the new release. For developers, the cost is higher: the README says the project framework is not yet documented, so you will have to read the source to understand how modules interact. The maintenance burden falls on the single maintainer, as the README invites contributors to join a QQ group. There is no mention of a changelog or migration guide, so upgrading might require trial and error.

Editorial conclusion

Whimbox is for players who want a hands-off assistant for daily grind in Infinity Nikki and are comfortable running an unofficial automation tool with account risk. Do not use it if you run the game at non-16:9 resolutions or on a low-spec PC, because the README states only standard 16:9 is supported and image recognition requires mid-to-high hardware. Before adopting it, verify that your game version matches the latest release (3.0.3 fixes a map mask login bug for both servers) and check the project's development status, since the README admits the framework is not yet documented. The backend no longer ships a UI, so you must also run the separate whimbox_app frontend. If you want a similar assistant for a different game, look at GIA or BetterGI, which inspired this project but target Genshin Impact. Whimbox is a niche tool for a single game, and its future depends on the maintainer's continued adaptation to game updates.

Official sources

1. Official documentation
2. Official README
3. Project repository
4. Release notes