lifein-ai/bazaar-helper

BazaarHelper is a local decision assistant for the video game The Bazaar. It reads the live state of a match along with card data, event data, and community-built loadout configurations, then works out how good the current event, shop, and visible cards are for the player. It shows explainable suggestions in a local web interface, meaning it tells you why it recommends something rather than just giving a single answer. The tool reads a game_state.json file that a companion plugin exports from inside the game while you play. It matches that state against the hero and loadout you are currently using, and analyzes the current event, shop, and reward options to show which cards fit your build, how likely a card pool is to produce what you need, and whether a purchase or reroll is worth your current gold, health, and reputation. It can also show which key cards a loadout still needs. An optional feature calls DeepSeek, a Chinese AI model, to turn the rule-based recommendation into a plain-language strategy explanation. The rule system does all the actual analysis first, the AI only summarizes and explains, it does not make the recommendation itself. Setup involves running a Python web app locally, alongside a BepInEx plugin (a modding framework for Unity games) that exports the game's internal state to a JSON file the tool can read. The project ships with a command line interface for manual testing, a release packaging script that builds a standalone Windows executable, and a test suite. Data files covering cards, events, skills, and community loadouts live in a data folder and can be updated with included import and audit scripts. The README notes that the browser-based web interface is being phased out in favor of running as a local API service for an in-game overlay, and that new features should go into the API rather than the browser pages.