FiberTree is a path-oriented database for storing and analyzing sequential decision paths. It helps you track, analyze, and optimize decision processes by recording decision paths (fibers) and their outcomes.
FiberTree is built around three key concepts:
Represents a single decision or action in a sequence. Can store any type of value and optional metadata.
A container for a sequence of moves forming a decision path. Each Fiber contains move sequences and related statistics (visit count, win rate, etc.).
A database that manages a collection of Fibers, providing functionality for adding, querying, analyzing, and visualizing decision paths.
pip install fbtreefrom fbtree import create_tree, Move
# Create a tree
tree = create_tree()
# Record a decision path
tree.start_path()
tree.add_move(Move("A"))
tree.add_move(Move("B"))
tree.add_move(Move("C"))
tree.record_outcome('win')
# Get statistics
stats = tree.get_statistics()
print(f"Path statistics: {stats}")
# Find the best continuation from a specific position
best_moves = tree.get_best_continuation([Move("A")], top_n=2)
for move_info in best_moves:
print(f"{move_info['move']} - Win rate: {move_info['win_rate']:.2f}")
# Save the tree
tree.save("my_decision_tree.json")- Statistical Analysis: Track win rates, visit counts, and other metrics
- Path Finding: Search for specific decision paths and find optimal continuations
- Move Frequency Analysis: Analyze frequency of moves at specific depths
- Path Diversity Analysis: Get insights into path distribution and branching factors
- Heatmap Generation: Create heatmaps for board games
- Tree Visualization: Visualize the decision tree structure as text or Graphviz diagrams
- Tree Pruning: Remove low-value or overly deep paths
- Tree Merging: Combine knowledge from multiple trees
FiberTree provides two storage backends:
# Default in-memory storage
tree = create_tree()- Fast, suitable for temporary analysis
- Data is lost when the program ends unless exported with
save - Ideal for smaller datasets
# SQLite persistent storage
tree = create_tree(storage_type='sqlite', db_path='my_database.db')- Data persists on disk
- Supports larger datasets
- Suitable for long-term data collection and analysis
FiberTree is particularly well-suited for:
- Game AI Development: Store and analyze game decision trees
- User Behavior Analysis: Track user navigation paths in apps or websites
- Sequential Decision Optimization: Industrial process optimization, medical diagnosis paths
- Knowledge Representation: Build decision knowledge bases and capture expert decision patterns
See the full documentation for complete API reference.
MIT