WordleBot is an AI-driven solver for the popular word-guessing game, Wordle. This repository contains code that integrates concepts from Q-learning, spectral clustering, and similarity-based filtering of candidate words to make strategic guesses and improve performance over time. Using a training loop, the bot learns which clusters of words to focus on, thereby refining its decision-making with repeated gameplay.

1. WordGraph

   • Maintains a pool of possible words and calculates a similarity graph between them.
   • Similarity scores are based on matched positions and shared letters.
   • As guesses are made and feedback is received, the WordGraph filters out invalid words, narrowing down the candidate pool.

2. Spectral Clustering

   • Groups words into clusters according to their similarity scores, helping the bot focus its guesses on specific “neighborhoods” of similar words.
   • The cluster centroids are used as strategic guesses to cover maximal ground efficiently.

3. Q-Learning (QLearner)

   • Employs Q-learning to determine which clusters to guess from.
   • State: The distribution of words across clusters.
   • Action: Selecting a cluster from which to guess.
   • Reward: Derived from feedback on how close the guess is to the target word.

4. Grader

   • Simulates Wordle’s feedback: correct letters in correct positions, correct letters in wrong positions, and letters not in the word.
   • Integral to the feedback loop that informs the QLearner and WordGraph.

5. Training (Trainer)

   • Automates running multiple games for training.
   • Collects performance metrics and prints results (average attempts, best/worst performance).
   • Facilitates hyperparameter tuning (cluster_count, alpha, gamma, exploration rates).

6. Visualization & Reporting

   • Provides plotting functions to visualize performance over training sessions.
   • Offers performance metrics to track learning progression over time.

• WordGraph: Manages words, calculates similarities, and filters based on feedback.
• SpectralClusterer: Clusters words using their similarity graph and identifies cluster centroids.
• QLearner: Implements Q-learning logic for selecting clusters.
• Grader: Provides Wordle-like feedback evaluation.
• WordleBot: Integrates all components to play the game end-to-end.
• Trainer: Runs multiple games, trains the bot, and records performance.
• Utility Functions (e.g., run_training_with_parameters, plot_performance, report_training_performance): Streamline experimentation, reporting, and visualization.

• Python 3.7+
• Install required packages:

  pip install numpy scipy scikit-learn matplotlib