import path_setup  # noqa: F401 — adds project root to sys.path

%load_ext autoreload
%autoreload 2

import pandas as pd

df_ground_truth = pd.read_csv("data/ground_truth.csv")
ground_truth = df_ground_truth.to_dict(orient="records")

from ingest import load_faq_data, build_index

documents = load_faq_data()

len(documents)

2775

index = build_index(documents)

from evaluation.evaluation import evaluate, text_search

results = []

for num_of_results in [3, 5, 10]:
    print(
        f"Evaluating num_of_results={num_of_results},"
    )
    result = evaluate(
        ground_truth,
        lambda query, num_results=num_of_results: text_search(
            query,
            index,
            num_results
        )
    )

    results.append({
        "num_of_results": num_of_results,
        "hit_rate": result["hit_rate"],
        "mrr": result["mrr"],
    })

Evaluating num_of_results=3,



  0%|          | 0/13874 [00:00<?, ?it/s]


Evaluating num_of_results=5,



  0%|          | 0/13874 [00:00<?, ?it/s]


Evaluating num_of_results=10,



  0%|          | 0/13874 [00:00<?, ?it/s]

df_results = pd.DataFrame(results)
df_results.sort_values("mrr", ascending=False).head(10)

from evaluation.evaluation import text_search

evaluate(
    ground_truth,
    lambda query='', index=index: text_search(query, index, 10)
)

  0%|          | 0/13874 [00:00<?, ?it/s]





{'hit_rate': 0.5211907164480323, 'mrr': 0.30056532030908645}