【Mojo入門】Pythonの書きやすさ×C++の爆速性能！次世代AI言語Mojoを完全解説｜基本構文〜AI開発実例まで

curl https://get.modular.com | sh
modular install mojo

mojo hello.mojo

fn main():
    print("Hello, Mojo!")

fn main():
    let numbers = [1, 2, 3, 4, 5]
    for n in numbers:
        print(n * 2)

fn square(x: Int) -> Int:
    return x * x

import math

fn dot_product(a: List[Int], b: List[Int]) -> Int:
    let mut result = 0
    for i in range(len(a)):
        result += a[i] * b[i]
    return result

struct Vector:
    values: List[Float32]

    fn dot(self, other: Vector) -> Float32:
        let mut result = 0.0
        for i in range(len(self.values)):
            result += self.values[i] * other.values[i]
        return result

struct DenseLayer:
    weights: List[Vector]
    bias: Vector

    fn forward(self, input: Vector) -> Vector:
        let mut output: List[Float32] = []
        for w in self.weights:
            output.append(w.dot(input))
        return Vector(output)

fn main():
    let input = Vector([0.1, 0.5, 0.9])
    let w1 = Vector([0.2, 0.4, 0.6])
    let w2 = Vector([0.3, 0.7, 0.1])
    let bias = Vector([0.0, 0.0])
    let dense = DenseLayer([w1, w2], bias)
    let output = dense.forward(input)
    print(output.values)

from python import Python

@python
def load_data():
    import pandas as pd
    return pd.read_csv("data.csv")

ai_search_mojo/
├── app.py                # Streamlit UI（Python）
├── data.csv              # 検索対象ベクトル
├── mojo/
│   ├── similarity.mojo   # 類似度計算ロジック（Mojo）
│   └── __init__.py       # Python ↔ Mojo ラッパー
├── embedding.py         # テキスト→ベクトル生成
├── README.md

curl https://get.modular.com | sh
modular install mojo

id,title,vector
1,レモンティー,"[0.12, 0.45, 0.33, 0.91]"
2,ダージリン紅茶,"[0.14, 0.42, 0.30, 0.88]"
3,緑茶,"[0.01, 0.11, 0.90, 0.12]"

fn cosine_similarity(a: List[Float32], b: List[Float32]) -> Float32:
    let mut dot_product = 0.0
    let mut norm_a = 0.0
    let mut norm_b = 0.0

    for i in range(len(a)):
        dot_product += a[i] * b[i]
        norm_a += a[i] * a[i]
        norm_b += b[i] * b[i]

    return dot_product / (math.sqrt(norm_a) * math.sqrt(norm_b))

struct SearchResult:
    id: Int
    title: String
    score: Float32

fn search(query: List[Float32], dataset: List[(Int, String, List[Float32])], top_k: Int = 3) -> List[SearchResult]:
    let mut results: List[SearchResult] = []

    for (id, title, vec) in dataset:
        let score = cosine_similarity(query, vec)
        results.append(SearchResult(id, title, score))

    results.sort(by = (a, b) -> b.score - a.score)
    return results.slice(0, top_k)

# ダミー構成（実装例）
def search_similar(query_vec):
    # Mojoでビルドされた.soファイルを呼び出す予定
    # 現状はPython模倣で返す
    return [
        {"title": "レモンティー", "score": 0.982},
        {"title": "ダージリン紅茶", "score": 0.977},
        {"title": "緑茶", "score": 0.544},
    ]

import streamlit as st
from embedding import get_embedding
from mojo import search_similar

st.title("Mojo AI 高速検索エンジン")

query = st.text_input("検索語を入力：")
if query:
    vec = get_embedding(query)
    results = search_similar(vec)

    st.subheader("検索結果：")
    for res in results:
        st.write(f"🔹 {res['title']}（スコア: {res['score']:.3f}）")

from sentence_transformers import SentenceTransformer

model = SentenceTransformer('all-MiniLM-L6-v2')

def get_embedding(text):
    return model.encode(text).tolist()

pip install streamlit sentence-transformers
streamlit run app.py