HumpBack / run.py
gregrsea's picture
v2: implement belief memory, rule-core emulator, API and tests; delete v1
fbb4123
Raw History Blame Contribute Delete
1.97 kB
"""Entry point: build the v2 belief memory from scratch."""
from humpback2.memory import Memory
from humpback2.evaluate import full_report, changing_facts_report
def build():
m = Memory("humpback2.db")
# 1. Plant a small, honest knowledge base.
facts = [
{"claim": "python is a programming language", "subject": "python", "source_type": "user", "confidence": 0.9, "volatility": "static"},
{"claim": "the human brain uses about 20% of the body's energy", "subject": "brain", "source_type": "user", "confidence": 0.85, "volatility": "static"},
{"claim": "price of apple is 10", "subject": "price of apple", "source_type": "user", "confidence": 0.9, "volatility": "fast"},
{"claim": "apple is a fruit", "subject": "apple", "source_type": "user", "confidence": 0.9, "volatility": "static"},
{"claim": "paris is a city", "subject": "paris", "source_type": "official", "confidence": 0.95, "volatility": "static"},
{"claim": "the capital of france is paris", "subject": "capital of france", "source_type": "official", "confidence": 0.95, "volatility": "static"},
{"claim": "what is 6 times 7", "subject": "6 times 7", "source_type": "user", "confidence": 0.8, "volatility": "static"},
]
m.learn_many(facts)
print(f"Planted {m.size()} facts.")
# 2. Train the change-detection and recall levels.
report = full_report(m)
print("Calibration:", report)
cfr = changing_facts_report(m)
print("Changing facts:", cfr)
# 3. Put the engine in front of a question.
print("\n--- Q: what is paris ---")
print(m.ask("what is paris")["answer"])
print("--- Q: what is the price of apple ---")
print(m.ask("the price of apple")["answer"])
print("--- Q: what is 6 times 7 ---")
print(m.ask("what is 6 times 7")["answer"])
print("--- Q: what is python ---")
print(m.ask("what is python")["answer"])
return m
if __name__ == "__main__":
build()